Techstrong TV January 28, 2026

Hi, everyone. Welcome back here to Tech Drunk tv. I'm really excited to introduce you to our next guest.

His name is Israel Mazen. Israel is the co-founder and CEO of a company called Mexico. We're going to hear all about it, but first, let's hear all about Israel.

Israel, welcome to Tech Drunk tv. Thank you for having me here. It's a pleasure.

Israel, I mentioned you're the co-founder and CEO of Mexico, but you know, that's relatively recent. Talk to us kind of about your arc, about, you know, how you got here today. Yeah.

So thank you. I have, uh, me and my team, uh, more than 30 years experience in building software companies, especially in cyber. Uh, mm-hmm.

We had companies that were public in the nasdaq. We acquired companies, we did exits, and, uh, actually we came together to, uh, four years ago to, uh, build this, uh, to build Mexico. So, uh, we have, as I said, a long time experience in building worldwide software companies, and this is, uh, why we came back together to the game and, uh, to build another, uh, large independent company.

This is our goal in Mexico. Absolutely. So you get the gang back together?

Yes, yes. All day. We have all the founders that we were many years together, uh, we gather back together, yes.

Excellent. Um, you know what, as long as you mention them, maybe they're watching this and they'd like to hear their name mentioned. Who's, who's on the founding team here with you?

Okay, so, uh, with me, we have Ellie Macia that we, me is the CTO and the brain. Mm-hmm. Mm-hmm.

Uh, of all the products. Mm-hmm. I have Gidon Hasam that is also the co-founder and COO and, uh, customer success, head of customer success, uh, or magazine that my brother also CRO, long, long time with us in this, uh, journey.

And then, uh, a few other, the join us that working with us 30 years, Dov Gal, the CFO, also working with us, uh, in the last 30 years, including, uh, public companies, private companies, uh, Zach, that is the head of r and d, all of them working with me in the last 30 years. Beautiful. You know, as you, like, you, I've, I've also been an entrepreneur 30 plus years.

It's a long time. Right. And, um, I'm mostly in cyber as well, right?

Until I, about 10, 12 years, well, almost 13 years ago now, I stopped all that. I started writing and speaking, and somehow wound up from a blog became What's Today, text Drunk. But I also, I surround myself with people that I've worked with, 2025, you know, 30 years.

I, I know how that is. But you can't just say, Hey, let's get together. We'll have, you know, we'll have a couple of drinks and we'll think about what we want to do.

There's gotta be a, a problem that calls to you that says, Hey, this is, this is a problem. And it's not just a problem We've seen ourselves. It's a problem I know other people are having, and let's, let's build something that solves this problem.

It's not enough just to build the big company or a successful company, that at the heart of it, at its core, there's gotta be some problem that you're passionate about trying to solve for everyone. What, what is that passion? What is that challenge?

At Mexico? This is true. When we, uh, gather again to do this company, we saw that there are a lot of scams that, uh, uh, around the world that all of us can fail, fail to these scams that actually try to steal your credential, try to scam you in a fake sites, fake social media, social engineering, and actually try to, uh, steal your identity.

And then do a lot of stuff like stealing money, stealing goods, stealing, uh, frequent flyers. It depends. But the main thing that we saw that happening significantly also to our family members and friends, that trying to actually, uh, get our credential, steal our identity.

Uh, I think that everyone, I believe that you see it, everyone see it happening every, every minute, every second, all over the world. So this was something that we saw and we said how it's, uh, not solved, uh, really in, in good way. So after we understand it, we talked with more than a hundred potential clients to see if they think that they solved it, why they didn't solve it.

And this is what we understand that the current solution that they had before we developed and started to sell our product didn't really solve their issues. They start maybe part of it, maybe give them some insight, but didn't solve the problem. So this is why we gathered again and we said, we must actually solve this problem.

You know, Israel, they say God works in funny ways. Just last week, just last week, friend of mine reaches out and tells me that they got a, a, a funny Instagram message from someone from me. I said, I didn't send you anything.

Well, let me see what you got. Sure enough, they got a, a follow request from a new, a new Instagram account. Shimmel Allen, no picture nothing.

Shimmel Allen, I look it up. Sure enough, he is follow, or they are following 85 of my friends, they're following, and you know how it works, right? I follow you.

You say, oh, that's Shimel. I'll follow him back. Yeah.

12 of my friends or connections followed him, followed this account back. So I, I reached out to those. The nice thing is on Instagram, I could see who, you know, who followed who.

So I reached out to the ones who followed, and I said, Hey, I see you're following this account, Shimel Allen, I want you to know it's not me and you should block and report it, but let me know, did they reach out to you? What's the scam? 'cause this has gotta be a scam, right?

As you said, yes. And sure enough, about half of them had gotten messages once they were following about clicking a link for the Bill and Melinda Gates Foundation. Another one was about a book or something.

They all different sort of scams ams, right? With, with the, with the idea is that these people should click links. So they were using me as a trusted source to, to, to really get people who are connected to me.

Now, I wrote to Instagram, you know, I went on Instagram. I said, someone's imitating me. They're imitating me that it's a scam.

You know, you get the automated message back. I'm very sorry, this doesn't violate our community standards. There's nothing you could do.

They gimme a suicide prevention hotline. If I'm that upset about it. I wrote an article last week about it.

These companies don't care meta and, and, and, and Google and what have you. They're not, you know, they talk a good game, but they don't help you when something like this happens. How do you help?

Yeah. When something like this happens, You describe exactly what's happened to many organizations in banks or airlines or retail or hospitality, that you get fake, uh, social media advertisement or profile or fake link to a site that look like the their site. You, and, but it's not, it's a fake.

And actually what they're doing, they get your, uh, actually identity and your credentials, and then they can do a lot of stuff. Transfer money, uh, buy a vacation, uh, with your credentials and many other stuff. Uh, steal your credit card number.

So this is the man you describe exactly what's happening in many type of vectors, like fake sites, fake social media, fake profile, what we are doing and what we developed, that it's a completely different on the other solution that were in the market. That we develop a solution that can in real time where to our customers base, uh, that it's, uh, banks, large banks, or retail airlines. Um, so it's give them in real time when it's happened and they send to you as their user some like fake link like this, or fake, uh, social media or whatever we know to detect it in real time.

And even when attackers start to prepare an attack before we attack you, we have a way to, to start detect it. And we put it as an event, an additional event, until we know that it will be an attack for sure. So we can actually alert to our customers, this users can fail to these scams, can be thousand user, 10,000 user, a hundred thousand users or more users.

And then we can, uh, also, uh, protect them by sending fake, uh, credentials to the hacker or they, or, uh, stop their, uh, devices or whatever, uh, that they cannot access, uh, to this account. So, and many other protection capabilities. So our main, uh, difference or advantage that we know to detect it, our solution is, uh, preemptive and in real time, and we can detect all of this, uh, that happening to our customers in real time.

And, uh, protect the users. We can, I think, prevent more than 70 or 80% of account takeover that happening to the users of the organizations. Israel, you mentioned you founded Mexico four about four years ago.

Yes. But, uh, recently you guys announced a series a, uh, round of financing, a healthy series, a $37 million. Usually you see a series a, a year, two years, you know, into the company.

I'm gonna imagine because of the, the history of you and your team, you kind of bootstrapped itself, funded for, for a couple years before you went out to raise money here. But anyway, congratulations on the raise. Tell us a little bit about who, what, why, why now.

Okay. So yes, we raised 37. Our goal was to waste 25 million, but we had very high demand for the round.

So we, uh, increased it to 37. And, uh, we did it now because we are going, uh, three times year over year in, in our a RR. And we want, we will plan to continue growth three times year over year.

And, uh, it was the right timing because we planned for the next few years the same growth. So we need to put the infrastructure, uh, expand our sales team, support team, uh, customer success team. Also, we are planning to developing more and more products on the same platform.

So we expand, also expanding our, uh, development and QA team. And so this is, uh, it was the right time for us to raise another round. Uh, and also the investors that invested first account investors, couple ventures and venture guide that was in the seed round, invested in this round, and the new investors, one of them is our customer, customer of US National Bank of Canada, that they saw the strengths of the product and the company.

So they invested the, they actually one of the leader in this round. And then we have, uh, PAG P'S group that it's, uh, the family office of Steve pca. It's a guy that, and the, uh, is the board of, uh, he is in the board of, uh, garner.

He's a senior advisory board of Bank Capital. He was the owner of both of Celtics till August, very well known that can help us a lot, already, help us to bring customers and potential partners. And the other one, it's a very wealthy family with many, many businesses in Latin America and Dominican Republic, uh, the Leon family.

And, uh, they can help us. They already introduce us to their businesses and, and potential clients for us. So it's bringing us not just money, but potential for many more customers.

Strategic investors. Yes, exactly. Strategic investors.

Yes. Yeah. That's excellent.

Excellent, excellent. Um, you know, look, today everybody is with AI and AI and AI and more ai. How is AI changing the game when it comes to digital impersonation, account takeover, fraud, et cetera?

Yes, it's very, uh, very good. Uh, question, because what's happening today that with using ai, using AI kits, uh, phishing as a service, scam as a service, what's happening that everyone also almost can do as to these scams? You don't need to be sophisticated hecker.

And also you can do a lot of them in very short time because you create them through ai. So it means that there are many more attacks. I can tell you that the estimation that until, uh, by 27, the aggregate amount of losses from this type of scams, the account take over, and Ford will be over $300 billion over the world.

It's a huge losses. So the, the, it's changed completely the, because everyone can do today these scams and very fast with ai. So what we are doing also, and we develop also solution, and we continue to develop it, that will know to detect all of these AI scams that coming from, uh, these fish kits or these scam kits or AI generated by AI to detect it immediately and then protect against them.

But it's make you much easier to do these scams and much faster. Absolutely. Absolutely.

Um, so I, I wanna make clear, right, your MCO sells enterprise customers, right? Organizations who are susceptible to this. There's a huge problem, like me as a cons as a single person, as a consumer, any plans maybe to offer something to help people, you know, just regular people, not organizations.

Yeah. So for now, we are more B2B, uh, player, and, uh, we are selling especially to organizations, but I can tell you that there's more and more organizations will use solutions like us and our solution. So you as a customer will be protected.

So let's assume that most of the organization will have something like this, then you'll be protected because, uh, you cannot actually, when you access, you get protected. For now, we are planning to stay in B2B, uh, play. Yes.

Now you guys can do a great job, right? And have a great product. At the end of the day, though, some of the responsibility has to be on the social media, the email providers, the exes, the, the, the Facebook meta, whatever, uh, you know, apple, Google, all of these companies that, that, you know, they, they just don't seem, I mean, they talk a good game, but they, the, the actions don't match the words.

Sometimes in, in taking this as a very serious problem, you're in a great spot because you represent, you'll be having all these companies that you and they're customers of yours, you'll maybe have some leverage. What, what can we do at that, on that side of the equation? Yes, I think to try to make it better.

Yeah. I think first regulators are more and more serious now about it. So they, you'll seek, uh, countries in uk, Europe, now, starting in the US Australia, that the regulators will force the organizations to protect their customers.

And if not, they give them penalties and they need to give the money back and penalties. So we see it more and more that regulators are, start being more serious and care about this. Uh, and this is what I think will happen because I think that, uh, to protect the real users, the end users, uh, uh, or regulation, uh, regulators will force the, uh, organizations to protect their users.

And this what will mean that they must, uh, uh, deploy solutions that can protect their users. I think it'll happen more and more, uh, in the next few years. I will tell you, in the case with me last week, I don't think they counted on most of my friends being cybersecurity people.

So when they reached out to my friends, my friends saw it was a scam, and they all, you know, did what they had to, they gave 'em a good time about it. But if not, you know, that was me. I, I feel bad for other people, Israel, for companies out here watching this saying, you know what?

This is a problem. We'd like to maybe check it out. What, what's the on-ramp?

What's the best way to engage with mem Mexico? Okay, so, uh, we actually, uh, selling directly. So they can of course contact us through our website.

com, all the contacts, you know, they can contact us and they'll get answer immediately. Uh, we have people around the world that's supporting them. We have channels, uh, that's selling us like Deloitte in EMEA and Latin America, local channels in Italy, Spain, Latin America.

So, uh, when, if someone want to approach us, we, they can approach us through our contact information on the site, and then someone immediately will be in touch with them. com. Yes.

com. M-E-M-C-Y-C-O. Very good.

Israel, first of all, congratulations on the oversubscribed series, a more importantly, thank you. You're welcome. More importantly, look, I could tell from personal, tell you from personal experience, this is a real problem.

It's a problem we need to solve. It's a problem we gotta take seriously. So I, I wish you lots of success, you and the team in solving this problem.

Thank you so much. I, I agree with you. And, uh, I believe that more and more customers will use our solution and then the customer, you end user will like you and other will be much more protected and do safer, you know, transactions in the internet.

Absolutely. Alright, Israel, ma Israel Mazen, co-founder, CEO of Mexico here on Textron tv. We're gonna take a break.

We'll be back with more in a little bit. I am Mitch Ashley of the Futurum Group. I'm Scott Roan with solutional.

We're here today to give you an overview of the Nokia Data Center Fabric reliability study, a survey that addresses some key issues in modern data center networking. We ran a future and research survey of a hundred IT infrastructure leaders from large enterprise IT organizations with the goal of understanding how data center network reliability is decided that's delivered and measured today and into the future. Mitch and I want to cover three main takeaways in this video.

First, reliability is the number one decision criterion. Second, operational challenges, especially human error, still drive incidents. And third teams claim meaningful automation, AI ops, adoption.

And we wanna unpack that a little bit. Yeah, three important messages. Number one, though, reliability is not a nice to have.

It anchors the design, the design, the operations, and ultimately in the business outcomes. Resilience is the end game. Yeah.

Not a huge surprise, right? That reliability was the top priority. Um, there's some interesting supporting stats that go around that.

Mitch, can you talk us through 'em? I think first of all, 86% of the respondents ranked reliability as a top decision criterion. So it wasn't just, uh, just above the midpoint.

It was well almost, you know, get, you don't get 86% in, in responses for very many questions. Uh, and the things that, that it was set on top of were the ease of integration operations. We know those are also challenges.

So why does this matter? A single hour, hour downtime is a widely expected to hit service levels and also revenue. So 47% foresaw a major service disruption risk, a 68 expected direct revenue loss.

So it's a big deal. 74% of organizations said they had greater than one incident of an outage in the past 12 months. So it's not a rare occurrence.

Um, when we see that many organizations saying they're having at least one out one outage a year, and that can be due to hardware failures, human error, those are common top causes. So now regarding human error, human, let's talk a little bit about that. We saw that amongst, um, multiple operational challenges, um, that drive those, uh, that drive the outages and incidents that we're seeing.

What, um, more is underneath those statistics, Mitch? It's a significant factor. I mean, the, the respondents rated it 80% said that human error impacts service 17 point half percent, called it a frequent top cause, things like that.

So it's certainly just more than a factor. It's an important aspect of when there is an outage, but it's also more than that. Um, there are often other failures that come alongside with human error at some point in that process.

That can be things like hardware or software failures. So how do we address this? When we asked the respondents, 35% said that they emphasized strict process and training.

25% said focus on resilience and recovery. Recovery, and only 12% said they aim to eliminate errors via automation. Meaning we know that errors are gonna happen, but we have to be able to handle those, respond to those we wanna resilient architecture, implementation, and also as well as the implementation or the automation that we're doing.

So, you know, teams are struggling to meet the evolving needs of the business because we know those are under constant change and also limit the scope or, or run extra planning cycles. Those are things that, that they're struggling with. Oftentimes, they'll even postpone important tasks due to confidence levels, uh, when they're not sure if that's something they're ready to implement or if this is the right timing to do that.

Last but not least, of course, skills always come up, but it was a significant gap. 54% said that that was skill gap was an issue, and several in incited cited that state versus desired monitoring limits were a factor as well. So on the implementation and use of automation in AIOps and the actual adoption, um, versus interest in automation and AIOps adoption, what did you find in the, in that bucket of responses?

Well, they, they said that here's what they're using today. Uh, a 67% said that they're using automated monitoring. 50%, actually 58% said they're using infrastructure's code.

I particularly found that interesting. And of course that's, uh, you know, followed by things like ticketing, auto failure over, but ai, ML based incident prediction was pretty significant at 54 4%. So I think this says that we're investing in IML as part of the, the solution set, but also I think we know that, you know, tooling does not necessarily equal positive outcomes.

Only 36% reported dedicated AIOps tooling as of now. And many are advanced practices are still in the maturing stages. So, you know, adoption is both planned and underway, but separating tool and use from operational reality gains is still key.

Yeah, that separation, uh, and, you know, that fine, fine grain understanding of are we just interested in automation and AI ops versus we're really, you know, going full force. We're gonna see that journey continuing, I think for years with many enterprises really just getting started in earnest. Definitely tracks.

I agree with you So much that you covered in, uh, in this survey. We're only touching the tops of the trees here. Where can people go to get the full report and, and plow through this and understand the fuller picture?

com and download the report from there. There's a section for analyst reports and the latest analysis that we've done. We'll also include a link with a video to make it easy to go right to the report.

It's free, download it, you've got, you'll in seconds, you'll be looking at some really compelling and interesting information. Definitely agree, Mitch, thanks for the pointer and for the readout. You bet, Scott.

Thank you. Hello and welcome to the latest edition of the Techstrong AI Leadership Insight series. I'm your host, Mike Baard.

Today we're with Dr. Arun sub Manian, who's the CEO for Articulate, and we're having a little chat about how AI will go into various vertical industry segments as we kind of develop more domain specific languages. Doctor, and welcome to the show, Mike.

Thank you so much for having me. All right. So what will it take to accomplish this?

'cause I think if I look at most of the solutions that people are using today, their general purpose, they're horizontally oriented, and yet in each vertical industry segment there's different nomenclature, different terms, different workflows. So how do we kind of flip this on its head as it were, and make this more vertical industry friendly? So actually, if you think about it, like the models have gotten significantly better, right?

So the general purpose models, uh, today I would say have gone from being a middle school student to a very competent high school student, sometimes even, uh, a competent, uh, uh, college graduate. However, the nuances of individual domains, as you said, are completely lost. And, um, like, what I mean by that is you can take, um, um, say even a high school student, give them a textbook on something very complicated.

Tell them that the answer to the question lies in the textbook and go find it. As long as they know how to go look at the indices, they can go do the table of contents, figure out where mostly things are, they can go retrieve something for you, maybe even the right answer. But then when they do that, they don't know whether this is the right context.

They certainly don't know whether it's the right answer and the person consuming it has no idea whether the depth of the information has been considered before giving you an answer. And that's what we mean by domain specificity versus generalist giving you an answer, right? Mm-hmm.

And, um, the, the flippant example I can give you is if you have a question about, uh, say brain surgery, you go ask a brain surgeon. The brain surgeon looks at a textbook, analyzes, uh, the information you're giving them, and then gives you an answer. A high school student looks at the same textbook, gives you the same answer, which answer would you prefer?

And hands down, it's a very obvious question, but when we go and ask a general purpose model that is designed to understand pretty much everything from a 5-year-old asking to give you a, a nursery rhyme to a, uh, financial analyst going and doing deep financial analysis of a particular investment thesis, the answers are moral or less plausible. But there anything but specific and most of the time, anything but accurate. So that delta is what we are trying to to navigate, right?

I'll give you another example that may sound very und. So take an example of a case where you wanna understand, um, a table, but then not just understand the table, but you want as a financial analyst, or say you are a safety engineer in a plant somewhere, and you wanna know that if this system looks at a table, it reproduces the table precisely every single time you look at the table. Right Now, there's no computing system on the planet that will give you 100% reproducibility.

999% repeatable, you needed to have tested your algorithms, your systems at least 10 million times across a variety of different tables. That's what we mean by going from a general purpose system to a system that works in an enterprise in a responsible way. Mm-hmm.

How does that kind of get worked through? And I'm asking this question because a lot of the answers that you get from AI are probabilistic. Yes.

And you know, that generally means they're right. Some percentage of the time that's less than a hundred. Yes.

And a lot of the things in a vertical industry workflow are deterministic, right? Yes. They're supposed to be done the same way every time exactly the same way.

And precisely, um, how do I marry these two things together to come up with something that is, you know, where one plus one equals five or better? Yes. So actually it is not really as much of, uh, a conflict as, uh, it seems on the outset, right?

So for example, when you ask a language model to make a prediction, it's really only making a prediction of a probability. There is no certainty there at all. And each time you ask it, depending on what, uh, is going on in the system, it's going to give you a slightly different answer.

However, you almost never put just a model to get you to an outcome. In fact, it's always a system that gets you an outcome. Even today when somebody uses, uh, Chad GPT or, uh, say cloud or Gemini three, they're not really just hitting a model unless you're hitting a model API, if you're using the application, you're hitting a system.

So you ask a question, the system interprets your question before it hands it to a model. And after the model gives it an output, there is a series of deterministic tools that interpret the outcome before you ever see an answer. So it is a combination of deterministic tools with probabilistic models that make the overall system deterministic.

And many times the final answer is something that is retried many times over. Mm-hmm. And the user may never actually see the retries in between.

Right. And are we essentially also using multiple AI agents to check the work of each of the AI agents or the models or whatever it may be, so that we're not overly dependent upon the single guess of one? Right.

That's absolutely essential because, uh, like, so we've, uh, developed a system called Model Mesh, which has now evolved into Asian mesh. And this was developed more than three years ago from a concept standpoint, at a time when it was not at all obvious to people that you needed more than one model to get you to an outcome. In fact, the the overall narrative was it is a model war, one model or two models are going to win.

And we went into this knowing that you need a combination of different models to work together at runtime to get you to an outcome. And second, even at runtime, your judge cannot be the, the same family as your player. That's what is the main differentiator.

Because even if you say multiple agents, one is generating an answer, another agent is actually critiquing it, the agents cannot come from the same lineage of models. If they do, you're more or less going to go to the average. Mm-hmm.

You at least need the judge to have an independence beyond just saying an instruction set. Right. And that one, we've seen it multiple times over.

And in fact, there is a, a recent paper from Stanford that quantified it. They called it the semantic collapse, which is once you get to about 10,000 documents, if all you're doing is vector search, almost all documents looks similar to all of documents. I respect to what question you're asking.

Hmm. That's why most of these kinds of question answer systems work really well. When you have a few hundred documents, you go to a few thousand, they start getting fuzzy.

And then beyond a few thousand documents, they're completely fuzzy. Hmm. And that's the reason why, Um, as you kind of think about this for a minute, we've had domain specific languages for a long time, but they were, shall we say, arcane and only a few people could, uh, master them.

And now it seems like we're gonna put a natural language interface in front of them, which that makes them more accessible. Yeah. So does that mean we'll see a, an explosion in r and d around domain specific languages?

'cause I think a lot of people didn't go build those because they were like, well, the only handful of people could use it, but now maybe everybody can use 'em. So actually it is the, the other way around, right? So where before you needed a domain specific language to describe a set of actions and activities related to the domain and only the advanced practitioner.

So that domain would even use that. Today, pretty much the interface is natural language. And the translation layer is what is missing.

Because say you ask an aerospace engineer about, um, what is Lyft compared to a manufacturing engineer? What is Lyft? It's the same English word means two completely different things.

And the context might also be relevant in terms of what you're talking to them. But knowing that context and knowing the domain makes a difference between giving you the right answer versus giving you the absolute wrong answer. Right.

And to your question of do I need to learn, uh, the domain at all? Do I need to know the nuances of the domain? That's where the unlock is.

Because today the finance person who doesn't understand the industry walks into a new company, they will have to learn the lingo of the company and the industry before they can actually like function properly. Tomorrow. When you have these domain specific models properly deployed, they can ask the same question, the answer will be contextualized to wherever they are.

Mm-hmm. That's really what is important. Mm-hmm.

So what is your best advice to the leaders of various vertical industry segments, whether it's manufacturing or finance, about how to infuse AI into their workflows? I think everybody's kind of having the same issues. They're like, we love the idea and the concept and we see the potential, but when it comes to the, uh, the actual execution, everybody seems to struggle.

Yes. So I'm, I would say more than advice, I would say I would have three recommendations, right? So first and foremost is we are well past the question of is AI useful?

If you are still doubting, if AI is useful, you really have to look at people who have actually made it useful for themselves. Now you can ask the question, I can see them finding use, I can't see it myself. Right?

That's where the gap is. And that's the gap that you are, um, highlighting. The second point to that one is if you accept the fact that we do live in a different world and you don't have to take it from, uh, it's a self-serving answer for me because I'm running an AI company, you can take it from the fact that we live that daily.

We can't operate if we don't find use from ai. That's number one. The second one is the general purpose.

AI systems that are out there are necessary, but not sufficient to get you to an outcome when you are running a deep industry. What I mean by that is, take a time when email didn't exist. Take a time when intranet in a company didn't exist the first time a few companies started implementing those, those were actually significant differentiators in terms of how they operated.

They got more productivity than their competitors. But very quickly, everybody caught up. Today, nobody would say having email is a differentiator, but everybody would say, not having email is a blocker.

That's where we are going with general purpose ai. So not having AI in your company in a safe way that everybody can use would significantly be a blocker. Mm-hmm.

Okay. Excuse me. But the third one is you just having AI is not gonna make a differentiation.

What is going to be different is what do you do with your own know-how, with your own data to improve your own operations that only you can do? And that's where domain specificity comes in. We call that hyper-personalization.

How do we take AI systems that are either domain specific or general purpose, make it specific to you and take advantage of that. Now, without that, everybody kinda sort of looks the same. Like today, if somebody sends you a one page document that looks well written, not that much of a differentiator, unfortunately, unless you can very clearly see the writing is original.

And unfortunately, we are very quickly going to a point where original writing is very rare. Hmm. So as you kind of put all this together though, um, who's gonna take the lead on these projects?

'cause to your point, everybody and his brother is gonna have the same kind of capability Yes. On the tooling. Yes.

So who should be at the forefront of these initiatives? So mainly business leaders who actually have to show business outcomes, right? So this is not a, I have a technology, let me try to find a use for it.

This is about saying, look, I have a business goal today. I cannot meet that business goal. Most of that business goal is about increasing your revenue, increasing your margins, whatever that might be.

Because you have to separate things into bottom line and top line. And I look at, uh, how leaders have to implement this as bottom line has a ceiling. So you can only get so much productivity outta the system.

Top line actually is significantly improving the business. You cannot just do one, you'll have to do both. And it is about where the businesses at that point in time, in terms of what you do first.

But not doing either will be a significant disadvantage. Mm-hmm. Now, I'll give you some examples, right?

Take a simple case of, okay, if you are not somewhere like giving your, uh, people the ability to automatically summarize meetings automatically, like identify action items automatically send that into your systems today. That's a significant disadvantage because not only that you are having somebody in the, the meeting figuring out how to take notes and then figure out where the, the action items come from. But also the way to track it manually is always gonna be slower than what these systems can do today.

Today it's trivial. Whether you use Zoom, whether you use teams, anything, it's nearly trivial to be able to use it. But the number of companies who don't use it still surprise me in a safe way.

Right? That's table stakes. What is that one thing that you see organizations doing today that just makes you shake your head a little bit and say, folks, we need to just be a little bit smarter than that.

I, So if it's one thing, it's still a significant portion of, uh, enterprises, especially operating like we are still in 2024. What I mean by that is they're assuming that the world is very similar to what it was in 2024 and 2025. Um, I can tell you, working in an environment where every two days, what we thought was our significant differentiator becomes stable sticks.

It, it feels like every day is something new. But even in very traditional industries, like we operate in manufacturing, we operate in energy, we operate in oil and gas in aerospace, where traditionally a change like this would've taken 20 years to go through. We are seeing customers change in a matter of weeks.

And they're not changing because it's a fad. They're changing because they can get to an outcome that they can measure that is fundamentally different, right? And the number of, uh, enterprises that we even show saying, look, this is what is happening out there, and we can actually show you evidence to look at that from a distance and go, but that's not me.

Uh, I, I won't get affected by that. I will continue to operate the way I'm operating. The, the tide for that is going to change significantly in 2026.

Alright, Cool. Just because the tools are out there, Right? Folks?

You heard it here. Change is coming. It's already here.

The only issue is figuring out how to operationalize this whole new set of technology in a way that gives you something that looks like a competitive advantage. But don't assume that whatever you created is gonna be there for long. 'cause everybody else is gonna do the same thing really quickly.

Hey, doctor, thanks for being on the show. Thanks a lot, Mike. Thank you so much for having me.

All right. ai Leadership Insight series. You can find this episode and others on our website.

We invite you to check all those out. Until then, we'll see you next time. One of our predictions for AI in 2026 was that sovereign AI would become increasingly important.

And this was emphasized by AWS at their European Sovereign Cloud launch in Potsdam. This week, we expect to see an increasing focus on digital sovereignty in the coming year, and many are looking at using specially developed AI models. But news came out last week of Apple and, uh, Google tying up for next generation Siri based on the Gemini model.

So perhaps this signals a trend toward leveraging external foundational models. We'll also consider the push and pull between infrastructure like data center and power versus distributed inferencing. On this episode of utilizing AI featuring Nick Patients and Brad Shiman of theum Group, Welcome to utilizing ai, the podcast focused on practical applications of artificial intelligence from the Futurum group.

Every Wednesday, we explore news and use cases of the ways in which AI is transforming enterprise IT and the industries it serves. I'm your host Steven Foskett, president of the Tech Field Day Business unit here at the Futurum Group. And joining me today, we have two of our fantastic futurum analysts.

Let's meet who's on the panel today, Nick? Hi. Yeah.

I'm Nick Patience on the AI platform's practice lead at at futurum. Um, and I, I look at everything that's to do, to do with enterprise ai and, uh, hi everyone. Brad Shiman.

I also work at futurum, where I am also a practice lead for a different research practice around data intelligence, analytics, and infrastructure. We kicked off this year of the utilizing AI podcast with a look at some of the predictions for 2026, and we don't wanna revisit those quite yet, but I think that it's worth looking into some in more detail now that, uh, things have started and we're starting to attend events. So, Nick, let me turn it over to you to talk a little bit about what you've learned, uh, this week's, uh, at at events as well as, uh, what you're thinking in terms of sovereign ai.

Sure. Yeah. Thanks David.

So one of the, one of my predictions, um, for the year and sort of a key part of our research agenda for AI platforms, there was, um, I think seven things in all, um, was about sovereign ai. And, and so I think it's, it's, it's something that was kind of a, I guess a niche geopolitical, um, idea and sort of peculiar in some ways to continental Europe. Uh, and now as I'm sure many of our viewers and listeners know, it's, it's expanded, uh, in importance to more or less every country, uh, in the world.

And there's two way, there's, there's the kind of national sovereign ai, uh, issues, um, but which are, which are very important, but, but probably, um, slightly too grandiose in thinking for, for kind of enterprise ai, um, users. But down at the kind of enterprise level, I think it's, it's, it's, uh, it's gonna be a key, um, a key topic. And, and this, um, and I was recently at this, uh, AWS launch of this European Sovereign Cloud.

Uh, the launch was in Germany. Uh, the cloud is in, uh, Germany, and it's an entirely separate cloud, um, from, from AWS It's not a, it's not a region, um, such as is entirely separate cloud that will be, um, uh, staffed by European Union citizens. Um, and essentially, uh, although it's owned by the parent company, obviously that's, that's unavoidable, uh, is run as a separate, um, unit essentially.

And, you know, the, I the idea is that it can give, uh, companies and organizations, um, you know, absolute control over their data, um, who has access to it, who doesn't have access to it, where it, where it resides, um, and so on and so forth. And I think, so I think it's, and then sort of specialized security and, and, and all sorts of other things. So I think it's exem exemplifies one of the things we're gonna be looking at this year.

And I guess the, the, the interesting thing I think from it is that that's a hyperscaler doing that. And when you think of sovereignty, um, you know, an American hyperscaler is, is both in a very, you know, an interesting but challenging position. Um, so there's also a lot of, you know, there's a lot of organizations, uh, and not just in Europe.

This is not just a Germany thing anymore. This is to say this is a, um, to more or less a global thing, but there's lots of organizations that are looking at, um, cloud and thinking, you know, maybe I want to, you know, have, have more control by having more on premises. And so yeah, there is repatriation going on and there's obviously a lot of data that, um, that never, that never actually obviously left OnPrem.

And so it's kind of, it's a little bit of a dance being carried out here by, by the hyperscalers. And they're getting certifications from various countries to say that they meet security requirements and compliance requirements, um, while also trying to grab a bit of that business and say, well, you don't need to have, you don't need to kind of go to a server server and storage and networking companies and get stuff and, and deploy it in your data centers and go to software companies. You can do it all with us.

And so I think it's gonna be, um, a dominant theme in 2026 and 20 and beyond. Um, and yeah, I just thought the AWS approach was really interesting and they roped out, you know, Matt Garman was there, the CEO, um, various politicians. It was a pretty, pretty big affair.

And they, they put a lot of effort into it. And so it kind of shows you how important, um, AWS thinks it is. And, and I'm sure the others will, the other hyperscale as well as well.

Yeah. And I, I would add to that, Nick, that, um, it's, it's not just about the infrastructure itself, but also about the software that runs on it. And we're seeing, as you mentioned, uh, and rightfully so, that this is a global concern.

And it's not just about legislative compliance, it's, it's about autonomy and being able to anticipate, you know, the fragility and, uh, you know, chaos that sometimes seems to, to make daily headlines for companies right now. And, um, so you're seeing in Europe and other, uh, regions that have some coordination amongst the nations, uh, efforts to, to basically free themselves from, uh, any sort of, uh, obligations they may have to external parties. And one of the, um, the big sort of unseen bits of fallout from the tariff wars that are going on right now regards software and, um, what sort of obligations a country may have to US-based firms doing business with the United States.

And so if for instance, you know, you are running Microsoft Office or Google, um, office workspace, sorry if they keep naming it different things, but, uh, anyway, if you're running that you, you sort of have obligations in terms of how you manage data and what data that company can see for your employees. And so we are starting to see companies not just locked down or take control, I should say, of the infrastructure, but uh, of the further up the stack for the software that they're running on that infrastructure. Yeah, totally.

And a key part of what, um, AWS are doing at the European sovereign Cloud level is also all the, all the access and identity management stack is also within, within this cloud. And so it's because have any other way it data you access to, um, you know, the, the systems, the software and the data. Um, and so it's, uh, it's, and the, and the metadata, I mean, it's down at that level of, of granularity of, of kind of control, um, that they're, that they're talking about.

I wonder if I could ask a question about it. Um, what about the models themselves? Is there any, um, thought of creating, um, regional, uh, sovereign models that, uh, are restricted in terms of, uh, training data or, uh, fine tuning, uh, that would avoid sort of, um, regional biases or embrace regional biases?

Biases? Sorry. Oh, that's true.

Yeah. I think certainly in terms of u using European, you know, in training data, obviously there's the link, the list, there's the language aspect to all this obviously. Um, but yeah, there are, there are definitely, um, you know, there are definitely, you know, specific, um, yeah, specific models.

Um, there already has been, you know, there's been, you know, specific ones in, in Greek and stuff, some from Singapore and, and Japan and, and all over. But, uh, so I think it kind of speaks to that general, um, um, you know, interest in, in, in sovereignty, um, all, all over. As I said, it's, I think for, for, for kind of, for organizations, you kind of think about it as a sort of strategic autonomy because obviously they're, you know, no matter how big you are as a, as a, as an organization, um, you know, you're not, you're not, you know, you're not the nation.

And so there's kind of two different ways of looking at it's, there's the, there's the, uh, national way of looking at it, which is, you know, we're gonna control the supply chain, we're gonna do everything ourselves. Um, which is somewhat illusion, you know, a bit of an illusion, um, even for the US and China because obviously they're both interdependent on each other, um, in different ways. So obviously China, um, is kind of more advanced in the production of energy and has taken open source models that the approach from the US obviously, um, you know, designs, you know, the world's most advanced chips, but they're made in Taiwan.

So there's, you know, all these kind of, I, the idea that can be completely independent, um, is I think an illusion, but from an enterprise point of view, is this kind of str strategic autonomy kind of lens. I, I like to think about, um, how, how, how, how it should be looked through. And, you know, if you think about models themselves as, you know, a representation of patterns, you know, in the training data that, um, that model itself is literally, you know, a knowledge base, a database, a, you know, map of the institutional knowledge of, uh, an organization or a community or, uh, something even larger.

And, you know, it's, you know, very much about being able to represent the, the way that, uh, not just the language itself as, as Nick you mentioned, but also the nuances of the way that people that speak that language natively think about themselves, think about the rest of the world, you know, from their context. And I, I thought, you know, this when I heard this, what I'm gonna describe in a second that this person was out of their mind, uh, back in 2020, uh, a CEO of a certain company that, that focused on, um, you know, diffusion models and creating images, uh, said everybody, every country, every city, every person will have their own model that will represent them. And you know, when you think about the data science that goes into training and fine tuning a model, especially at a frontier scale, it's a brute force effort that is very demanding in time and money, uh, and GPUs.

Uh, and the fact that we're starting to see companies, and I think we talked about it on this podcast a little bit ago with AWS and what they were doing with their, um, Azure Forge capability to, to customize models. We're starting to see even the frontier model makers, uh, give tools to the enterprise to take control of that representation of their own company, their in a mu in a much friendlier, uh, more accessible manner than having to hire a bunch of data scientists. You know, it's interesting, Brad, that that brings to mind an announcement that we heard of last week or the week before about, um, well, it wasn't really much of an announcement.

Essentially, um, apple and Google have jointly, uh, let it be known to Jim Cramer of all people that, uh, apple is gonna use Gemini to build their next generation Siri. Now it remains to be seen what exactly this means. And, uh, certainly we'll be watching for that at WDC this year.

But, uh, now almost two years ago at ww DC Apple, uh, previewed a very expansive vision of what they would do. It sounds like Apple then spent a year plus, uh, sort of spinning their wheels trying to develop their own model before finally throwing in the towel and deciding to use the pretty incredible Gemini from Google instead after, um, sounds like a bake off. Um, given all of this, what does this say in terms of using in-house models, uh, versus, uh, external foundational models, Brad?

Yeah, it's, it's funny, the timing is, is kind of interesting in that, um, as we were just discussing with AWS that, um, you know, frontier model makers are making it easier for companies to take ownership of a foundational model and fine tune it for use internally. And this partnership, uh, is very much, you know, an extension of that idea because yes, uh, apple has a longstanding relationship with OpenAI, which is what they announced, you know, with what Steven, you, you, uh, mentioned with their early we're gonna remake Siri idea. And, um, you know, that was basically a handoff to open ai, uh, to, to, you know, sort of obfuscates the identity or identifiable information about the user and pass that to, to, to open ai.

And with what we know about Google Gemini and its open source rel near relatives, I guess you call them in the Gemma family of models, uh, that you can literally just take the, what's great about the Gemini family and distill it down to a smaller model that still maintains a lot of the capability of the larger model, and then fine tune that thing to meet very specific use cases, not just running in the cloud, but running on the device itself. And we all know that Apple's been spending quite a bit on its system, on a chip design to, to be able to, to run AI effectively on, on their own platforms. And if you look at their mlx, uh, compiler ex, what do you call it, sorry, execution environment for their AI endpoints, it's pretty impressive what the, the size of model that it can run on some, you know, on a laptop, for example.

So I think it's at a good time, it's a good timing for Apple to, to make this, uh, joint announcement, not a press release, not not anything of any, any note, uh, in our usual circles because, you know, we call it an official leap. Yeah, right. It's, it was called literally an, an official joint statement.

My goodness. A, any anyway, you know, if, if, if Apple is serious about preserving user privacy, um, then this is a good partnership to execute on that, to go beyond what they had, you know, like, like you said, Steven tried internally and perhaps didn't succeed with, or ostensibly didn't succeed with. I think it's, yeah, it's interesting 'cause they, they were, so, apple was so far ahead when they bought the original technology from SRI didn't they Stanford, uh, research Institute, wait, I can't, I can't remember when it was the nineties or two thousands or something, I dunno, a long time ago.

Um, and then, you know, it just sort of, you know, sounded quite impressive when there were no alternatives and then suddenly when the world alternatives didn't, um, to be completely blunt, I mean, I'm a I'm an Apple user, you know, love, love, love the kind of vertical integrated stack and everything and everything like that. Um, but it's, um, I don't use it, um, Siri because it wasn't, it was no particularly functional for me, but, so this should make quite a lot of difference. But as you say, the privacy aspect, that was, that was the pitch, wasn't it?

That was Apple's pitch. That is why we're different. And it's, I wonder, I think you're right from a technical point of view, this, this will enable 'em to maintain that.

I wonder how important that still is for users, um, given the amount of, you remember the kind of early days of, of, of generative ai we're like, do not put things, do not upload things into that. Um, well, you know, guess what? I think everybody's doing a lot of that.

Um, so I, I, I do wonder how important that is. But, but, but also, or even, even if it's possible, right? Yeah.

Was it yes. Is there not, did we see, last week there was some legislation put forward in the European Union to, uh, basically, um, scrape data before it's ever encrypted in apps like WhatsApp or, or Signal. Alright.

Yeah, yeah, yeah. And I think the, the other, that's another places too. Yeah.

The sensitivity or the lack of publicity they want they were seeking has probably got more to do with the, um, the judgment against, uh, Google, um, from last year. Totally. 20 24, 20 25, wasn't it?

Which as far as I'm, I read, you know, that bar's Google from entering maintaining exclusive agreements that last more than one year. So this presumably is not an exclusive agreement. So if Apple were to go back to OpenAI or philanthropic and, and, you know, they could sign another deal.

Um, so, which is, which is interesting. I mean, that's the, uh, the judge didn't say, you know, specifically this company with this company and, you know, and these kind of deal just said, any, any exclusive deal that lasts more than one year. Um, so, so I think it's, um, you know, that that's probably got something to do with, um, some of the, the lack of, uh, shouting this from the, uh, from the rooftop plus the money involved, which I know they didn't announce, but there's numbers out there.

Well, right. What is the money situation? Because we, we do know that Apple, um, pays Google, or sorry, Google pays Apple quite a bit of money, uh, to make their search engine the default engine in Safari.

Mm-hmm. And that was part, that was the reason for the, um, part of the reason for the anti trust, wasn't it? Exactly.

And then is Apple is gonna pay Google what billion a year, um, for this. Yeah. Which is so, um, I'm sure they're completely separate.

Yes. I'm sure they're totally separate transaction, but the Gemini just shows reinforces the importance of Gemini, didn't it? The little old, um, DeepMind folks in, uh, here, in here in London.

Um, you know, behind all that, you know, it's, uh, now actually, uh, turning into absolute real money, um, along with Google Cloud's performance last year, and, um, and, and what we think it'll be doing this year as well. It's, uh, it's a real business, these, uh, these AI models. Yeah.

And I think that's actually the most important thing. In fact, um, depending on what we see about the money changing hands here, we could be looking at the most lucrative AI deal in history, uh, in terms of actual productive, you know, to the point of this podcast, right? You productive use of AI utilizing ai.

Um, if Apple is indeed paying Google billion plus dollars to, to use Gemini, this would be one of the, one of the more lucrative or maybe the most lucrative investments. I do wonder as well, um, what this means in terms of the private cloud compute, which they announced back at, uh, ww DC in 2024 as well. It was a great idea, which was to the point that both of you have made that Apple has developed quite a lot of silicon, uh, compute horsepower, and they are going to seamlessly extend that as a, a private enclave to a version of their apple silicon running in a private cloud and enable, um, basically cloud bursting of AI processing as needed.

Um, do we want to assume that Gemini is actually running on Apple's private cloud compute, or do we wanna assume that Apple has just sort of waved their hands, and maybe this is running on Google's infrastructure, uh, because of course they've done tremendous things with their TPU hardware and so on. Um, I, I guess there's no hint yet about how that's running. Is that true?

I think you would be foolish to conclude that, um, they're not going to avail themselves of every aspect and avenue that they can to effectively, meaning cost effectively serve these Gemini models to their, their somewhat sizable user base. Yeah. Yeah.

Because it could just be the model running on Apple, but Yeah. Yeah. And, and clearly I think that, I think we can assume that it's gonna run natively on Apple local devices, right?

Absolutely. I mean, it's not gonna be completely in the cloud, right? Yeah, no, that's what I'm saying is it's gonna be all of the above.

It's, it's depending on what you're doing, I would imagine you will see a graceful handoff of functionality, um, depending on what you're trying to do. If you're just trying to replicate what Google has with their circle to search capability on, on their phones, that's gonna run locally, why wouldn't it? Um, but maybe the search side of that transaction, not just the extracting the image, but the search of that image is something that you could very well and probably should hand off to a backend service somewhere.

Yeah, and I guess it also just shows from a kind of higher level, um, more abstract point of view in, in ai, the, you know, the, the importance of inference and, you know, inference is, is a, is a business, and this is inference obviously, because, you know, Google's gonna handle the, the training of the, of the frontier models, and then Apple gonna build upon that. Um, but every time, yeah, we're using it, that's off the inference, and that's what they're, um, essentially, um, yeah, not, not what they're paying for because they're paying for the actual model, but you know what I mean, it's kind of, yeah. That, that's how, that's how this, uh, the value is gonna get realized for the, for the users.

So we're doing a lot of, um, sort of guessing here about the specific deal with Apple and Google. I wonder, uh, if you all can maybe take a step back, what does this mean for the market in 2026 and beyond for ai? Does it look like, uh, this is, uh, what the future is going to be for enterprises trying to deploy AI applications?

Are they gonna put themselves in Apple's place and do a bakeoff and pick a model? Um, to an extent, yeah. I mean, I always caution that, uh, in most cases, um, the model is not the application.

And so, you know, you're not, um, you, you, they're not gonna be necessarily, um, talking directly to the model or be layers of abstraction. Um, but, uh, but yeah, you are gonna end up paying for, um, you're gonna end up paying for this. And this is where, um, you know, the kind of, you know, some of these other sort of trends we expect to, to see, um, the new metrics of inference time compute, and that, and that, that kind of thing was gonna become really, really important.

Um, because obviously it depends what you're doing. Obviously, if you're trying to create videos, um, you know, the, the expense of that compared to, you know, responding in text to some sort of, you know, text prompt is' just vastly different. So, so yeah, I think, I think, I don't think you'll see necessarily major enterprises directly, you know, saying we want to license Gemini.

I think, you know, there, there's more that there'll be, they'll be buying into a stack, um, or at least Google hopes they will be. Um, and, um, and that will, that will be, uh, a key part of it. And that's kind of, you know, the Gemini enterprise stuff is, is, is like that.

It's, um, obviously Google's gonna keep extremely tight control of, of the, of the models. Um, and, and then, then that contrasts with the open source approach of, you know, there's lots of, there's, you know, thousands of open source models out there. So, so organizations do have a choice of how they go about this.

Yeah. And if they can, you know, basically build a, a sort of control plane for their applications that abstracts the exact model underneath, you know, to basically, uh, route the request to the most appropriate model being the model that's best able to answer the question or to carry out the task and to do so with the right, you know, requirement for latency, for privacy, uh, for security, for, um, concurrency, for instance. All of these little things that, that make up the decisions that drive technology investments around building software are, are very much at play, in play here.

And that's why, you know, what we've, and we've said this on this podcast and, and elsewhere, that, you know, models are less of just a chat response model and more of a platform that is built on a rich set of capabilities that are exposed through APIs and SDKs. And so if I, as an enterprise builder, uh, am going to use, uh, or if I'm, if I'm gonna run a task, I want to know, a, does this model do caching, for example, for just one example of many measures, you know, can it do prompt caching so that I don't have to keep saying the same prompt back and forth? Does it do speculative decoding to, to optimize, uh, what the transaction is happening inside the model itself?

Does it have chain, uh, uh, what am I trying to say? Uh, community of experts, what is that called? Sorry.

Um, MOE model mixture of experts. Yeah, sorry. So all of these, these things that, that aren't just a model, but are the surrounding, you know, uh, sort of infrastructure of the model are what are, what are gonna drive a lot of purchasing decisions in the enterprise?

Yeah. It does seem as, as was the case with enterprise software as well, that the platform is more important than the underlying, um, you know, uh, infrastructure components. And I think that that seems likely to continue.

But speaking of infrastructure components, um, one more thing, uh, that came up that was interesting. There's been a lot of talk about, um, 2026 marking a transition from the sort of heavy GPU supercomputer data center model or finance model at least to more, uh, focus on inferencing and distributed compute and lighter weight. And we've just been talking about that a little bit, but at the same time, there are still investments happening in data center power and energy.

Um, what are the thoughts, uh, that you have about this sort of, um, one way or the other, uh, direction of the industry, Nick? Yeah, I think that will, um, that, you know, the last thing you said there will continue in, in 2026. I think the, the, the issue for the data center industry, um, in this year is obviously gonna be, yeah, the, the, the power and cooling issue and the, and the energy is the major, you know, energy is the major bottleneck.

They understand this very well, the people, the organizations that build that acquire land and build data centers, but it's filtering up to the enterprise as, as this is, this is gonna be a challenge, um, for, for, for organizations overall. So as you kind of, you stuff more and more into a rack, the power, the power requirements for that rack goes, you know, from sort of 15 kilowatts to a hundred kilowatts to, to more and more and more. And, and obviously the idea is obviously you want to stuff, as, you know, cram as many of these into a building.

Um, and then of course, what that results in is a lot of heat, um, but a lot, a lot of power to, to, to, you know, run it and then a lot of heat generated from it. So, um, you know, I think this year we're gonna see, and we've seen a little bit of it in 2025, uh, data center build outs will get delayed due to the lack of, um, energy. Um, there's obviously an interesting arguments about the kinds of energy and, you know, the kind of in front of the meter, behind the meter, uh, renewables, non-renewable, the contrast between the US approach and the Chinese approach.

But always already in the last few years, we know of many smaller, um, you know, data center operators that were kind of gonna, where the power is rather than where the customers are. Um, and so, yeah, that, that's been happening, but I think it was, and, um, kind of some delays. And I also think, um, yeah, it makes liquid calling mandatory.

And so for those organizations, for the data centers that would, would not, not having that, they have to be retrofitted and for the organizations that that sell that equipment, um, you know, I would imagine it's gonna be, you know, it's gonna be a pretty good 20, 26, 27, 28, 29 onwards and onwards and onwards. And obviously a lot of these are the big server companies, um, but also the, you know, the companies like Google that, that build their own data centers or build don't necessarily build themselves, but they build the equipment, um, that that goes within them. Um, and Amazon, the same kind of thing.

And Microsoft the same, same way. So I think it's, it's this, um, you know, air cooling is sort of hitting a physics wall. Um, there's only so much air you can blow across a rack to keep it cool.

Um, so I think, you know, this is, this is something we're gonna be looking at pretty closely in, in, in 2026. Yeah. Unless you put it in space, then, then it's pretty cool.

That's true. Um, clearly, but yeah, man, I, I, I feel like, um, it's not just about the power, but also about the actual infrastructure itself. And by that I mean the sand that makes up things like RAM and things like NVME drives, uh, in particular that are gonna drive a lot of the economics of the data center.

And, uh, I heard earlier last week from, uh, a vendor that specializes in, in storage and object storage in particular, say that they're already looking at a, a two a, a magnitude of two times the weight, you know, for getting, you know, that those, those basically NVME drives for their customers. So if you're waiting two years to, to get drives, what are you gonna do? You're, you're gonna optimize the drives you have.

And, uh, this vendor, uh, part of their go to market is now going to be saying to their customers, you've already spent X amount of money on amount money, your hard drives in the data center. Let's do two things. First, let's use our management system that gives you, you know, an X increase, fold increase in the amount of data that you can actually store on this thing.

And second, let's offload some of the workflows that may normally have sat above that in something like a database, for example. Let's just push it down into the storage layer where it can be run more optimally. So you're using fewer watts, you require less cooling, and you are, you know, consolidating those workloads in a, in a very effective manner within that existing rack.

That's, that's pretty fascinating to me. It's not about let's stand up a new nuclear reactor in a new data center. It's maybe we should optimize what we have.

Yeah. Very pragmatic approach to the, uh, to the, to the problem, which I think, yeah, many of our, our listeners will, uh, understand and, and, uh, want to do. Certainly sand must flow.

Yeah, we, we've gotta be pragmatic because, um, shortages of RAM and storage are, uh, just everywhere right now. And so we've gotta be thinking about how we're gonna optimize the, the commodity we have. And of course, that could affect everything we've discussed here.

So, um, this is, uh, getting a little bit long this week. Thank you so much for, uh, weighing in here on these, uh, seemingly disconnected, but actually quite connected topics, uh, of what's happening in the industry from, uh, digital sovereignty to, uh, apple plus Google plus question mark to, uh, the, uh, push and pull between infrastructure and, um, raw materials and, uh, AI applications. Uh, before we go, uh, let me quickly check in with both of you on what your, uh, where your research is taking you this week and, um, what you're going to be doing.

Um, Brad, uh, what's new with you? Well, I'm, I'm actually, um, looking at, uh, a new forecast that, uh, we're finalizing this week, this week, so hopefully I'll have that online, um, within another 10 days or so. And, uh, like Nick has with his, um, survey he's working on, we're, you know, we're always, we're always active building out new research, uh, over here at futur.

So I invite everybody that, that listens to this podcast to jump over to, to our site, the, the future group com, because you'll be able to gain access to a, a lot of this research. We don't gate everything we do, we, we try to, you know, share our insights as broadly as we can. And for me, I guess I finished my, the survey.

Um, it's is going to the field, um, it's in the field, um, by now by the time you listen to this. And so we'll have the results of that, uh, in a few weeks time. And then we'll be publishing, um, reports on it, we'll do, we'll do podcasts on it, we'll publish some of the data, and I'll be doing some presentations to clients, um, as, as to what's going on.

And always, with all these surveys, you're always trying to have a, a mixture of longitudinal questions, which it kind of gives you show your patterns over time and then trying to keep up to date, uh, with what, what's going on, uh, in the enterprise. So that's, that's what I'll be working on. Excellent.

And, um, as for me, um, we're gonna be hosting AI Infrastructure Field Day next week. Um, tune in live, uh, Wednesday and Thursday and Friday for presentations on a lot of the infrastructure that we've just heard about. Uh, very much looking forward to that one.

And of course, we've got another AI Field Day shaping up for, uh, in, uh, Q2 that is already just, uh, bursting at the seams with companies, uh, joining us for that. So keep an eye on the Tech Field Day socials for that. Thank you for listening to utilizing AI today.

Uh, if you enjoyed this discussion, please subscribe on YouTube or your favorite podcast application and consider giving us a rating or a review. This podcast is brought to you by the analysts and experts at the Futurum Group, where Insight meets ai. For show notes and more episodes, head over to Text Strong ai, the utilizing AI YouTube channel or the text TV app.

Thanks for listening, and we'll catch you next week. More horsepower, more power. Get those horses moving faster.

Your ai, it's huge. It needs more power. Or does it, can you actually do great things for your business with less horsepower and do things more efficiently?

Join me on the Tech Field Day podcast as we drive this stage coach through the ideas of efficiency in AI infrastructure. Welcome To the Tech Field Day podcast, where we bring together a group of IT technical experts to discuss a single idea about key concepts in the industry. This podcast features a variety of perspectives from members of the Tech Field Day delegate, and is often recorded in association with one of our events, in this case, AI Infrastructure Field four Tech published on, on this episode, we'll be discussing that AI needs resource efficiency and that more horsepower isn't the only direction.

But before the discussion, let's meet who's on the panel today. Hi, I'm Pete Welcher. Uh, I've been around networking for a long time now.

Retired, still doing it, 'cause I think it's fun. Hey, I'm Gina Rosenthal. I am a product marketing manager.

I've been doing that for a while, been doing product marketing for AI since about 2017. And I'm Andy Banta. I've been hanging around the infrastructure industry for quite a while, um, mostly from the technical aspect of it.

And I'm Alist Cook. I'm the event lead for AI infrastructure Field day here at Tech Field Day. And one of the themes that we see is that AI needs to kind of grow up from this land grab and this idea that massive amounts of infrastructure, ridiculously powerful GPUs stacked up in, uh, rack after rack with hopefully a very strong flaw and liquid cooling to get to ridiculous numbers of, uh, kilowatts, hundreds of kilowatts of power being delivered through a single rack.

Um, I've long felt that there needs to be a much more efficient way of delivering ai. I was hoping there'd be some revolution where this more horsepower would cease to be the issue, because all of this horsepower and all of this compute power, all that energy has a cost. It's really expensive to generate these things.

It's really expensive to build these data centers. And as we've covered in the tech fields and news rundown, there are contracts for more data centers to be built to host AI than have ever existed in the past, exist now. So this is a huge build out for ai, uh, using vast amounts of resources.

And even if we set aside the difficulty of actually producing those, the cost to people producing them, is that even a sensible thing for us to be doing with the planet? Should we be generating AI by heating the entire planet up with these massive GPUs? Um, it's kind of a a point of of difficulty along here.

Um, Andy, you've been around this, this block a few times before you've seen massive infrastructure get built out and maybe get more efficient. Can you see that happening here with ai? Well, I, I think it has to, uh, one of, I, I've actually been, um, talking about the power consumption used for data centers and AI data centers in general for the past couple years.

But one of the things that I would be really interested in hearing from, from the AI infrastructure companies out there is how they're doing things more efficiently, not how they're doing things faster or more densely or, uh, or various different ways of adding more and more. But it's, you know, tell us something about some algorithms that you're using to make more efficient use of the hardware that you have available or the speeds that you have available, or talk to us about a way, uh, a technique that you come up with to do more with less and, and instead of just telling us how you can do more, that's really what I'm interested in hearing more of from infrastructure companies. I completely agree, and I've got a slightly, uh, different perspective on the whole market.

I think the companies that developed the very large models needed, um, vast funding because of the huge costs that go into those LLMs, and they couldn't afford to specialize, uh, because they really needed something that would have a vast potential market. Unfortunately, what I think has happened is that they've assumed that's the future. I'm not sure could, it's certainly got a role particularly for conversational, um, interactions, I guess to put it.

But if you are trying to model something, be it network troubleshooting or, um, uh, some biological process or something, maybe a smaller dedicated model that knows about, say, computer networks, um, might be more suited to the task and a whole lot more efficient to run. So that's kind of a question in the back of my head. I don't have an answer, I'm just kind watching the space, uh, with intense interest.

I agree, both of y'all. And Andy, I agree with the whole idea around power. When you think about, if you read any type of description of what, uh, the data centers are being built for, it's described in gigawatts.

So, um, it's not a back to the future reference, but it's actually how much power they need to run, um, the servers and to run the different components of the servers. They never talk about the data, they never talk about how much data that will serve us. They never talk about how much data will be created.

They never talk about, you know, the actual part of ai, it's the end result of it. They only talk about what it takes to run it. But when you do an infrastructure, when you do any kind of architecture, you're always looking at what is the workload that's gonna be run on it.

And it's not at top speed because not everybody can afford to run everything at gigawatt speed all the time. Not only that, but if you think about what is ai, this is common, I'm gonna always come back to this. What is ai?

AI is what we used to call right now. What what is available from AI is what we used to call high performance computing. So machine learning and deep learning, and you don't need gigawatts of power to run those types of workloads.

So why aren't we talking about this as a true architecture instance, right? Like, what needs to happen, how much data you need, what do you need for each section of the pipeline that we're gonna be doing to transform this data? Um, that, those are kind of different questions, but I think those are important.

So we kind of skip over the whole, we can design for this, it's a workload, this is what we know how to do, and if it's workload, we know how to make it run more efficiently. So I'm, Well, you touched on something that I've been focusing on, which is, uh, where's the data? And I think we're gonna see that during the ai, uh, infrastructure field day, uh, presentations, because if the data's not local or if you've got, uh, such a huge LLM that you need to run your, um, training in multiple sites than just accessing large amount of data is yet another barrier as well as latency and bandwidth and power, Right?

And I mean, the, the other piece of this that I, I really wanna try to address on some of our discussion here is that the, um, in addition to the data, the ability to actually create the data isn't being done efficiently. Uh, and Fabrica presented at one of these sessions a long time ago before they got consumed, and part of their presentation was talking about the fact that the model is only, uh, theoretically capable of using about 70% of the GPUs available to it, and in practice is only using about 30% of the available to it. And that is clearly showing that more is done.

What you need here, what you need is to actually do the good old fashioned software engineering to make these models more efficient, to make use of the resources that are available to them. And that is really part of the efficiency that I would like to hear people talk about. Uh, Gina made this point a little bit earlier, uh, when we were in the pre-discussion for this about why aren't people looking at solutions similar to VMware of, uh, of sharing resources among various different processes.

Uh, and you know, I I think Phoenix can offer a little bit more information on what she was thinking, but that's like an excellent question that we don't hear from Ethan VMware these days. I think VMware has a hard time getting the message out, but, um, they have a lot, they've had a lot of tools for a long time where you could do things like virtualize the GPU and split it into, I wanna say 16, but it might have been eight to deliver that, that that vir that, uh, power to different virtual machines. So the same thing we did with Oracle back in the day, remember nobody thought you could virtualize Oracle.

That, that it would be, you would cost so much because you needed those servers to be bare. You need to, uh, uh, put that Oracle OS right onto the server, and that's how you had to run it. But once you started looking at, no, these are workloads, you can virtualize all of it, we can virtualize every component of it from a VMware perspective, and it's virtualization in general.

I'm sure it's all of the virtual, um, virtualization hypervisors that can do that for you. But, but it, we're not thinking about it that way. We're, I think it's, it's, it's definitely the hype that's driving that.

Like you have to have ai, you get ai, you have to have this many servers and this many GPUs that that's the formula, the blocks, and that's kind of how they're being sold. I don't see anybody pushing back and saying, nah, man, we need, what we're needing to do is we're needing to take this historical data, do this with it, it's gonna cost this much for us to crunch through it, and this is what we actually need to install in our data center. But people were doing this before the days of ai, they were doing it when you went to super compute, you would see people doing cool things on bare metal.

And I know this because we were there with VMware saying, and you could also do this with virtualization. So it, it, it's just workloads. And I think that the architects and the engineers need to get back to, to that piece, which would also force people back into the what is the use case?

Why are we gonna to do ai? What's on this side? What's on the other side, and what do we expect to report back?

You know, what, what's this gonna gain the company if we, Yeah. And it's, uh, and VMware certainly did do virtualization of GPUs, and they, they did it both for, they did it both in software, and they also were able to split out the hardware functions of various different GPUs and spread those to various different VMs. And through some of their later technology, they were actually able to share those across the network.

Um, I, one of the other concepts that has been brought up that I really haven't seen a whole lot of traction on is there are CXL vendors out there who are attempting to say, harvest the memory from your, um, systems that you were shutting down and put it into a cxl rack that you can actually use on a new system. The idea is that DRAM doesn't go bad. DRAM can last for a very long time, and even if it's not the newest, most, most fast DRAM that's out there, you can certainly make use.

But again, so I mean the, in addition to efficiency, we also don't hear too many people talking about reuse or sharing of resources. We're gonna have to, right? Because what happens now when all of a sudden the prices of everything is going through the roof, you can't get rid of your memory.

You can't, you really have to hold onto it because if you don't hold onto it, you, you're not gonna buy any new unless you pay three times what you're already paid, what you paid for it yesterday. So we're in this memory crunch because they can't make it fast enough. The same way you can't get ahold of GPUs because they can't make them fast enough, can't manufacture them.

We're in a place where the chips are getting smaller, but getting, doubling in, in capacity size. So they haven't really, uh, perfected that, um, that, that in the, in, in the manufacturing plants to get those rolling and get those out to customers. So many of them are already pre-sold And they're pre-sold, so you can't even get them once you, once you get ready to go deploy this.

So you're gonna have to get creative and figure out how do you architect for this workload with what I got or what I can maybe find on eBay for a decent price. Right. Well, Jane hit on another, um, thing that I think is key, which is shorthand would be ROI, but what's the company benefit from what people are doing?

And that's kind of the challenge with everybody does ai, uh, which seems to be the, a meme in some companies, and it really is much more helpful to start thinking about ROI. But another factor in it is that maybe not everybody, just like not everybody is cut out to be a programmer. People have their job they wanna do, and they may have a idea that AI can help them, but right now, uh, how to prompt A LLM and, uh, get good results and sort of cycle that is, uh, something that's not for everybody.

And we're starting to see some mechanisms coming, uh, gait and some of the other things that help with automating that sometimes even in massive ways, which could be a problem. But, uh, when is this, when is AI sort of experimentation gonna be easier for, uh, Joe Schmo in marketing or whatever to use, uh, as opposed to somebody who's a computer science, uh, specialist And, and then Pete, is that the right thing to have happen? Because as we're talking about this as, as doing AI through resource efficiency, what are the tools that that abstract away the intelligence of the operator, right?

That, that take away the specialization of being able to work with the AI tool? Well, that puts more load on the ai. That means that we need more horsepower in that AI to handle the fact that we're asking somebody who, who is not specifically skilled or oriented towards understanding the tool to just throw simple language stuff at this tool.

This is how we're getting this idea of needing more and more horsepower. I think coming back to one of Gina's points was that having an AI tool that is specific to a task rather than a general purpose ai. So building into your application instead of using, uh, a full 7 billion or 20 billion parameter model, that is a general purpose model.

Having a model that's specifically set up to handle maybe, uh, analyzing the productivity of your, um, factory floor. Having an AI that is specifically trained for that function is gonna be more efficient than feeding the all of the information about your production floor into a general purpose model. And that's where we need to look at that resource efficiency versus simply building more horsepower.

There isn't a single solution here, there isn't a single answer to this. Sometimes in order to get value out of ai, we're gonna need to a very specialized AI that does a very specialized task. And I come back to some of the early discussions around this, that a lot of the time it was still the, uh, predictive ai, the old fashioned machine learning that, uh, was looking for trends in data and, and extrapolating those trends and data a lot of the time, that's what the production AI was.

And strapping a chat bot, large language model on top of a huge amount of resources wasn't necessarily the right way to solve this. Another of the themes that we saw across our AI field days, uh, as as they were maturing, was the idea that chat bots aren't the answer. They, the AI needs to be invisible inside the application doing whatever it's doing.

Uh, it needs to not be the, the primary interface. It needs to just be the thing that makes the thing better. To quote Bobby Allen say, AI should disappear and, and should be task specific.

Well, that's where maybe the LLM gets used in a slightly different, uh, form of age agentic, uh, use, where it recognizes what field a question, what type of question is being asked, or what type of inferencing is, is being requested. And then it calls on a dedicated model that, uh, for instance, like what Cisco's done with an in-house network trained, uh, model to actually answer the question. So it calls an AI agent rather than a software code agent.

That's classic program. Yeah. Yeah.

And I mean that's, I think that's part of where Alistair was going. I I think part of it, uh, the idea of the efficiency could be that you, you don't need to train something on the entire world of knowledge to, for a specific thing. And I mean, if, if you go in with the idea that we only need to train something for a specific piece of knowledge, then you, you probably can do it more efficiently.

And it, it's, uh, you know, you talked about Cisco, uh, a friend of mine who worked at a MD said that one on one of their early models of their, uh, virtual assistant that was based on an AI model that he, the first thing he did was ask them, what is Epic? Which is the name of, uh, the empty desktop? And, uh, the virtual assistant couldn't answer it for them.

One of the things that, you know, you brought up, I, I think it's unfair to say that Joe Schmo in marketing, this is like LLMs are the perfect tool for marketing, right? And I know people aren't giving the right prompts yet, but, um, we had a conversation about this this morning on the Textron game. What happens when, um, what happened with email, when email first came out, everybody couldn't use it.

I mean, I had to support post drops who were astrophysicists that really had a hard time with email. So, you know, because it was so new, it was something brand new that nobody knew how to do. Um, and they didn't, they lost, they didn't understand the rules.

But now you can't imagine having an organization that doesn't have a proper email system running one way or the other. And I think that's how it's gonna be with ai. AI is gonna be how we communicate with computers in the future, right?

This is how, and now going forward, and it's gonna get more and more, we'll, we'll figure out the latency problem, we'll figure out the power problem. But people now, like the few people that do are, I would say the very, very technical people don't understand how important those marketing people are in the organization and how they can pick the information out and distribute it back. Again, defending marketing, because I'm product marketers, so have to do it.

But, but I, I think, you know, the very technical people forget that that data that we're moving around and trying to transform things into, one of the way it gets transformed is by the, the knowledge and the point of view that the people that aren't technical that are gonna take a little longer to train them on the prompts and, and how to do things properly, they understand the way to pull the information out because they have the vocabulary, everything else that nobody else wants to learn about marketing. If you're not a marketer, you're a technologist, you don't care, you don't wanna learn that, but the marketers will know the right words to pull that out and pull the information out of the data. And that's really what a lot of those LLM, um, workflows are all about, is to help pull information outta data.

So I think some of this is just, we are in such early stages and we were bombarded absolutely bombarded with hype about what this is and how you don't get on board, your company's gonna be sunk and all of the rest of it, you know? So I think it's really important to remember that all of us are suffering from height overload and we're gonna get there because this is how we communicate with computers now. Yeah.

But this goes back to the original point of yes, I, I agree with, with exactly what you're saying, but I, I think, uh, some of the, some of what's being lost here is that, um, the efficiency that you actually would get, lemme back up that this is, this is making things more inefficient rather than, uh, more efficient. And I think one of the things that we, we lose sight of is actually paying attention to the genuine computer science necessary. And instead, we, we keep adding layers that are, uh, wasting cycles to get the same job done.

Thousand percent agree. Have you ever looked at stuff and it looks like front page like AI stuff looks like front page from Microsoft way back in the day. Yes.

Yes, I agree. I agree with you. So Will, will it optimize when it looks like, um, MySpace see the front page?

I think that'll be a little bit optimized, but that's like when they're just, that's probably like when they're just to the point where it's too much and then they're like, holy crap, we gotta pull this back and make this really work for people and we're gonna make it good. Now, One other challenge that I'm, I think is lurking here with the idea that everybody does AI is, uh, it takes me back probably several decades to, uh, ro it, uh, the prospect of people just learning spreadsheets and doing corporate financials on a spreadsheet and screwing up a formula with consequences. And that's actually good news from the AI perspective, is we've been here before, so people learning and making mistakes, uh, hopefully you can fence them and be aware that, that that's a potential problem.

But, um, I think assisting people with frameworks that, uh, minimize the mistakes is probably a useful thing to think about In terms of resource efficiency. Um, I, I really don't expect that an unskilled user should be handed here. Here's an AI that is a general purpose.

Ai, you can do anything you want with it, and we're not gonna tell you any limitations around that, right? That's, I don't think that's a good way of using ai. AI should be invisibly embedded into the thing you're already doing.

So as a, as a marketing person, Jen is perpetually writing documents, writing, um, presentations. And if you're going outside of that, using AI to generate some things and then come back into it, the workflow's wrong. Let's, let's say there's just another layer being added that adds no value.

The AI should, should actually be embedded, however frustrating. It's to have the, uh, Google sheet say, would you like me to generate everything for you? Right?

That's where it should be. Uh, the AI should be a feature that makes the application you are using makes the tool you are using function better. It shouldn't be a separate place that you have to go.

Um, some of the evidence that we're misusing the separate place is, is absolutely ask the ai, what is epic when you can go to Google and say, what is epic? And be taken to the documentation that describes exactly what Epic is. People are using, um, chat GPT as a replacement for search and finding that chat.

GPT isn't necessarily up to date on the latest news. It certainly doesn't know, uh, about the ice storm that's coming to the Eastern US uh, this week, or about the landslips that happened close to me here in New Zealand over the last couple of days due to the weather. Um, but I can find that information through Google search, right?

Using AI and particularly chat interfaces for the wrong things is gonna lead to poor results. And I think hiding them away, Don't you think that if you, I'll go back to my front page, you know, that's why front page was so awful because it had everything in one, one place. So don't you think that we're gonna end up having more bloat problem and more resource problem if, um, if we put everything in the place?

Well, data sovereignty is the other thing I'm sitting here thinking, because if you have, uh, people just asking chat GPT, uh, well, it's already happening that, uh, very private and sensitive information is getting exposed because it's sloshing out into the ai. And given some recent results where people have been able to reproduce entire books or major chunks of books by feeding, by iteratively prompting the LLM, uh, they could probably get at that secret data, uh, confidential data as well, which is not a good thing. Returning to, to one of the things that Gina said that, so the sort of blow to building AI into everything, I think you've got a really vital point how rebuilding, putting AI into every single application independently as we're seeing at the moment isn't efficient.

Is it? It's the same thing again and again and again, uh, but we're seeing some movement towards some standards about how you inject ai, uh, or specific knowledge in AI into an application technologies like MCP, um, standardizing the way you attach to an ai. So I don't think there's necessarily the need for that, that complete rebuild.

Uh, and there's other standards like the agent to agent standards because as Pete was saying, a lot of the stuff is, is actually gonna be done by agents rather than directly by human interaction. So I think there are developments in that space of not having to rebuild, redeploy, uh, do again, exactly the same thing we've done in the past. So I think there is hope for us that the AI delivered into a specific applications doesn't have to be a completely from scratch, are really an efficient reimplementation of some other ai.

Uh, of course that's probably what gonna, what we're gonna see first because that's how you get to a minimum viable product. Uh, I think Pete mentioned that, that idea of, uh, we built huge large language models and, and so that's the way we view all of AI needs to be this huge amount of resource for these huge large language models. Again, I I view that a little bit as a minimum viable product.

I really hope as maturity turns comes along, we'll see more efficiency, more reuse of technology, and more use of tools that are specialized to a single function and that do that function more efficiently. The thing is, we've gotta stop the pace of going for more and larger, which was coming back to our original premise, more and larger isn't necessarily the right way to go. Um, so yeah, I, I think I've rambled all over the place on, on a bunch of stuff we've talked about.

Uh, and I, I really do hope we do see more efficient use more tasks, specific use of large language model AI and, um, some of that maturity of working out how we're getting business value rather than just the, um, arms race to acquire as much resources we can. Because if we don't, somebody else will acquire that resource and, and we'll lose, I think in the end, if all you do is acquire a resource and never use it, you're gonna lose anyway. You may as well, uh, look at efficiently using those resources that you acquire.

Well, there's competition for the resources. There's also time to market. So if you're using massive resort costly, uh, resources and training a model, you have to build this Mongo data center and everything just to get rolling, that's a long lead time.

And so anything you can do to make it lighter weight, it affects, uh, improves your time to market, Right? I, and you know, I, I think one of the other places where we could potentially see some efficiency is, uh, being able to filter out AI generated content when we're actually building our models. Uh, it's one of the things that that's, we, I mean, we, we end up with, uh, hallucination amplification when we do that.

And it's, um, it also means that we're, we, we end up with AI that's seeded with a imaginary information to begin with. And that's, that was one of the terrifying things about Deep seek was, was when they built deep seek, one of the things they did was use a lot of AI generated content to teach the ai. And yeah, that does seem like fantasy land being used to build another fantasy land.

And that doesn't translate well to being useful in the real world generally. I think one of the other struggles is right, you know, of course the marketer I use, I use, um, Microsoft. And, um, one of the things that's gotten way better is their copilot, especially if your documents are on the platform, that's gotten way better.

But you talk about, um, uh, the bloat is there and the bloat is there because the systems underneath, uh, office doc, you know, online office are not all the same. They've come from mismatches of what they have and put up on, you know, the different, different sites and we use them, but you can see it when you try to start to run this whole pipeline, you know, writing and producing and editing and adding things and that kind of thing, because you can't do it all in one place. So I think that that's a problem too, is like your businesses are going into AI and they might be going out and buying whatever they can to be the most powerful and the fastest and blah, blah, blah, but they're reaching data that they're using data, they're organizational data that's on all sorts of stuff, maybe even on tape.

There's a lot, that whole piece they're saying. That's what I've heard is that whole piece of preparing the data to actually be used by a model and to be used to make some business battle that take value. That takes a lot of time and a lot of doing.

And you think about rehydrating things from tape and getting things into an object ready, if that's the way you built things, how, how that looks for everything to be consumable by the model. So it's, it's, we we're gonna have a lot more bloat before we can get to the trim down thing, even if we're trying to, so like, we've got this problem of not being able to get the components, this whole problem with getting the data ready, you know, and then making it, um, co for the business users to use. So it's, I think that that idea of bringing data together from different sources leads to a, a sort of disconnect where, as you say, it's not all the same format or naming of things.

And this is the semantic layer that Brad Shiman from Futureum Research has been writing about recently. I'll be presenting, uh, a little bit of his research at our infrastructure field day tomorrow. And so I think there's gonna be a continuing conversation at our infrastructure field day about how you build this infrastructure, how you make it valuable, and we're probably gonna come in with a lens of how you make it efficient.

Andy, Pete, and join Gina, uh, here in, in, uh, Santa Clara with us. And I have a, a collection more awesome delegates. Uh, so before we fill out hours and hours of this conversation, I think I'm gonna put a line under it here.

And thank you all for joining us today at the Tech Field Day podcast. So before we close out this podcast, where can people connect with each of you and maybe continue this conversation or maybe decide they wanna buy you a beer at your local, uh, bar? I can be found on LinkedIn.

Um, my initials PJ Welcher should find me. You can find me on LinkedIn too. And I am so excited that my podcast partner will be with me at Tech Field at AI Field Day, and you can find us at Tech Aunties.

Okay. And you can find me on LinkedIn as well, Andy Banta, uh, as well as on Blue Sky Social, and I, uh, my, my occasional, um, blog posting on andy banted substack com. And of course you could find me, Alistair Cook on LinkedIn and all of your favorite social media, as well as at AI Infrastructure Field Day this week.

Looking forward to having some great, uh, conversations with people there. So thank you so much for listening to this episode of the Tech Field Day podcast. And of course, if you enjoyed this discussion, subscribe on whatever your favorite podcast application is or on YouTube so you don't miss a single episode.

Uh, give us a nice review as well, and a, a high rating because you love coming and listening to the Tech Field Day podcast, particularly when we have such entertaining delegates joining us that will help other people to find the awesome content. This podcast is brought to you by Tick Field Day, the home of it experts from across the enterprise and a part of the RUM group For upcoming events and more episodes, head to tick field day com slash podcast or view us on Techstrong tv. Thanks for listening and of course, we'll as always see you next week on the Tick Field Day podcast.

Thank you so much for having me. And also thank you for the forum, and I'm just so appreciative of the group here. And I, I love the session yesterday where folks were talking about their experiences and the one thing when I met Ira was actually on a delegation and it was like 20 people.

And the thing that I noted about him was that he is like the king of real talk. So he tells the truth, he puts it out there like shamelessly unabashed truth. And so I really appreciate that.

And so we were just kind of drawn together from that. But, um, no, I did not train myself completely or become the quantum expert or anything like that. I tell you, there are some amazing people that work in this space, whether they're cryptographers, this Paul is something, isn't it?

I don't know whether I should, I'm just gonna go back and forth, I'll just walk back and forth. But it, there are just so many bright minds and amazing people in this space. And I tell you, I, um, actually was drawn to this space by a client who worked for P PayPal.

He is a sister there and he wanted me to talk about emerging tech, and he knows that I'm always in the gray space, always in the space. That's not quite commercial yet, but we've gotta figure out how to secure it. And I, I, it forced me into really digging in really deep looking at not only IBM's technologies, which were, you know, um, huge in quantum tech.

IBM's been at this for a long time, but also looking at the industry completely, um, understanding how other people view the industry, whether that's MIT, whether that's AWS, Google, all the other, um, publicly held, um, D wave ti ion Q companies that are in this space, but more importantly really around our use case. So I wanna ask how many people know, very much would say they know very much about quantum comp as it relates to quantum computing. Okay.

And can you say what you do with Supply chain security and product security? So we have been up uplifting and, uh, doing crypto agility on products. So that awesome quantum money.

Are you also the SBO lady? Yes. Oh, this is so great.

I I'm so excited to hear you're coming 'cause this is what we'll be talking about. And you also said, uh, Yeah, so I'm the, uh, the CSO for Prince Georges County. Oh, cool.

We lived close, we lived there for, for 13 years. We lived in, um, I know, but we, yeah, yeah, yeah. So I, I came from the intelligent community having worked at, you know, three letter eight, you know, FBI and NSA.

So, uh, has some background and exposure to, you know, all the quantum stuff back at Fort Mead. So, okay, Awesome. And I think I saw your hand.

Yeah, So, um, one of the senior managers actually at Google, uh oh cool. Willow quantum ai. So, great.

Was there anybody else that wanted to add, I know a little bit about p qc, not Necessarily. Okay. Speaking of p qc, how many people are involved in a p qc project at this point?

That's post quantum cryptographic migration. 1, 2, 3. Anybody else over here?

Okay. Alright, so we'll go ahead and get started when we leave here today. I'm hoping that everybody will just feel slightly more comfortable.

Quantum's one of those spaces that it's just so overwhelming. I could boil the ocean today, but I only have, you know, what, 45 minutes. So what I'm gonna do is we're gonna talk a bit, I'm gonna walk through these slides, but I'm gonna try to leave as much time for q and a afterward.

'cause I'm sure there'll be lots of questions, but you'll get a feel for the industry. You'll get a feel for what people are doing from a project standpoint and what some of the issues are. Like, you don't want your board coming to you going, what are we doing here?

And you don't have like an answer that happens a lot. It's starting to happen even more. So I sit in a lot of different worlds, um, which is actually a good thing, but it also can be a little bit overwhelming.

I, I was sort on the board for the Department of Defense that manages the third party supply chain, the CMMC program for the Cy Arabia. And that is a huge, um, that's, that's their attempt to really sort of reign in all of the supply chain in terms of a level of security around compartmented unclassified information. So you can't do a bid, you can't work on a project if you don't have the level of certification that's starting November 10th.

And it's hugely impactful because everybody's been attesting to it. I think. Um, Bob talked about third party security attestation and people have been attesting to having the right level of security for probably about since 2017.

And now it actually you have to prove it. There are folks that are coming and they're going to verify. So that's one of my boards.

The other board that I serve on is the, um, CTA, the, uh, consumer Technology Association. Those are the folks that run the CES show. And with that I'm sort of their cyber person on the board, but also I work with the legions of member companies who are in some cases, you know, sort of close to security and other cases, maybe not so much.

But these types of topics are important. We've invested in Quantum World Congress, I am on the investment committee there. And so I work with the team that brings like the 30 some countries together to talk about quantum tech.

And so, um, and we'll be doing more of that. The CES show is making its best attempt to sort of mainstream this 'cause it's been in the lab, it's been in research, it's been in academia quite a bit, but not really mainstream. And so also, um, I have a company, it's called Quantum Crunch because it's crunch time and we work with boards C-suite insurance and we help you have the difficult conversations, but also we establish the plans that make people at least, um, get started on this project.

But not with, you know, spending, you know, tons of money, which is what has been happening a lot. People have been spending a lot of money and they've been finding that their projects have not gone in a place where they've achieved a lot to show for it. And so we, um, and we've done over 37 engagements, um, where we are working with customers who say the first call is always, oh gosh, my CEO called.

He wants to, he wants a briefing on this. He wants to know what we're doing. Um, I I, there's so many third parties, we don't know where to start.

There's so much information out there. We're not, we need to get up to speed. And so, and then also, um, the other thing to mention there that's really important is that the third party aspect of this, I can't underplay or overstate rather enough.

And it does get underplayed. And the thing about security mostly is that this is not a project to go alone. There are a lot of other people, this is an enterprise risk management level project.

Okay? So everybody knows about quantum computing and how its other worldliness is in terms of quantum mechanics and physics. It's gonna change the world as we know it.

All the use cases, you know, drug discovery, like you won't have to wake up in the middle of the night and hear on TV that the drug you've been taking for the last 20 years is being recalled and they want to, um, want you to be part of a class action suit. Drug discovery will be a lot easier with quantum computing. It's really not going to replace classical computing at all.

It's gonna be great on the, um, computation end of things. Really doing computation that classical computing can't even approach. So you've got, you know, the, the use cases you hear about all the time are drug discovery, material sciences.

You hear a lot about supply chain optimization. And then in the financial realm you hear a lot about modeling and, um, financial modeling, financial analysis and like Monte Carlo simulation and ways to manage market disruptions and predict the future. And so it has enormous promise, but for all of us lucky people, we get to deal with the, those are all the carrots.

We get the spinach, we get the project that really takes the one nefarious use case. And that is quantum's outsize computing power. Being able to break modern day encryption is what everybody here should be concerned about.

And it's also not Y 2K. And I will tell you why. So when I say a sufficiently capable quantum computer, I don't mean like commercially available quantum computer, I don't mean that, um, you know, it's uh, something that can't be done in the lab more or less.

And that's important because, you know, we have cryptographers all over the world that have been working on crypto analysis to create the resistant algorithms that, that can manage the speed and capability of quantum computing. And NIST has been the folks that have convened that and they have worked for eight years on this. They've worked for eight years, but they've also done multiple rounds.

They've started out with like 82 submissions, then they cut it down to 69, they broke those and they hit 2029, then 27 in the last four in 2022. And they released those standards in August. And it was thought I was at the we White House during the meeting and we were, you know, it was like, oh my God, this is happening.

Like the whole world's gonna just focus on this now. And it kind of still was a lot of, um, a lot of nothing, like not a lot of people doing a lot and a lot of convening around the world still. 'cause this is very much a global project.

And I give this a lot of credit. They do a lot of things, but this project right here is hard and people didn't really pick up on, you know, really moving forward with projects right away. So This was one of the dilemmas.

So as you talk to organizations, this was happening around the same time. It's just a little bit after chat GPT was released and everybody had their, um, code red meetings about what are we gonna do about this? They have their AI councils where everybody's getting together and they're trying to figure out and manage that risk or manage at least what the fallout could potentially be.

And you had legal involved, you had a lot of people making lots of, um, rules and policies around the organization. I was very involved in a lot of those meetings in the s and p, um, 500 realm. And it was kind of chaotic.

And it took this discussion like off the map. It really did because people were really concerned about this. There was a lot of already sort of, um, you know, shadow AI happening in the background.

So it just became less of a priority. But this is what everybody says. They're like one more thing.

And at that time, I think teammate had done a study with CISOs, I think they, I don't wanna say it was like 150 CISOs. And they kept saying that, you know, CISOs will never be responsible for at least the bias part of ai. And, and CISOs were avoiding these meetings too.

They were like, I'm not going to that council meeting. Like that's the CDO, the chief data officer. That's those folks, when they figure it out, they can come tell us 'cause we are really not shifting that far left, right?

But this idea of being a cost center is really a critical part of what we do. It's the idea that really the people who have to make these decisions have to understand the risk. And they can't conflate it with quantum tech.

Quantum tech is hardware software. This issue is just a use case of quantum. And so that's the confusion.

So these folks, as more investment comes into this area, they start to sort of conflate the idea and they worry about the timelines and they just go, oh, I can like not worry about this for 10 years. I don't really have to think about it. And so the accountable people, the third parties, which I don't wanna steal any of your thunder because No, no, that's fine.

It's a whole, that's a, that's just a whole thing. But it's also sort of like an excuse that people lean on too to not get started. And the problem with this whole thing is that it's just very, it's just complex when you're talking about baseline components and you're, you're at the software level and you get down to the CBO M level, the federal government sent out a, um, spreadsheet to, to all the agencies saying, Hey, you gotta do your inventory as if it was just a list of items, you know?

And, uh, they were confused. There were no instructions. They submitted it to OMB and then they end up coming back with like kind of a hodgepodge of responses.

And they had to keep going out there and they had to engage. Um, you know, uh, I think it was, MITRE was probably one of the folks that went out and tried to salvage this whole thing. But that's the complexity.

There's, it's very hierarchical when you're talking about all the parameters and ev all the elements associated with actually having, you know, an inventory of your cryptographic assets. Ooh, this is, I'm getting dizzy. It's really rocking here.

You guys are really rocking. Y'all are all rocking. You rocking next.

So thank you. I'm trying to rock the, um, factorization. So I just wanna say this a a thing about this because this is what makes the, the, um, migration difficult is that the, when you're talking about quantum resistant algorithms, they're not gonna be drop in replacements.

It's gonna always involve some testing. It's gonna involve extensive testing and everybody already has a whole number of exceptions and, and systems that can't be touched. 'cause they can't be broken and they just sit out there forever.

And so this, when you look at this now, this is a slide, this is actually a slide that we used to talk about all the time because the key link, when you look at 2048 RSA, that is, it's just hard to imagine that anybody could break that. 7 billion CPU years to break that under quantum. The cost of factoring of just plain factorization eight hours or the timeframe like that.

If you can wrap your mind around that kind of power, that's what we're talking about. And what I find is that the, um, you know, when you look at Adversarially, like sort of who's doing what around this space, there are some of our adversaries are spending more money than us, at least from the federal standpoint. And there's a lot of expectation that from the federal standpoint, we would be like pushing out a lot of guidance around this and that, um, we'd have more of a handle on a timeframe.

The reality is that you can print, put 37 folks in a room and they will be experts in this space and they will all say something different. And so it creates, it, it lost people into a false sense of security, of thinking this is gonna be so far from now, I'm not even gonna think about it. So NIST released the first set of standards, which you guys are probably familiar with this 2 0 3, 2 0 4, 2 0 5.

Everybody's heard about this. Um, everybody's working on it. There is a lot of, there are a lot of hybrid schemas where you're using classical as well as quantum resistant algorithms, particularly with, um, the first one, um, crystals kyber, which is sort of your general purpose, um, algorithm.

And the other two are more for digital signatures. And so the industry's dilemma is that everybody has, um, you know, there's this sort of fragmentation, but also there's this reliance on the outdated cryptography, what I was talking about before, the idea that you have these applications that can't be broken. Some people think of, well this is, I need to piggyback this project onto something else.

It either needs to be part of my zero trust initiative, or I need to slip it in where we're doing ref factorization on, um, you know, our major applications. But a lot of people will say, oh, I'm gonna go off and I'm gonna do this little mobile app and we're gonna sandbox, we're gonna work on this. And the only problem with that is that when it all comes down, this is not going to be, that's not going to gonna help you a lot because it's those core apps, the really important apps that are the ones that should be focused on.

I actually was around for Y 2K and whenever somebody calls this Q2 K, it just drives me crazy. It just drives me absolutely crazy. I used to work at one of my first IT jobs was a company that, um, basically pulled together LPAR space for developers in order if anybody Yeah, yeah.

Oh my, it is like cloud before there was cloud, right? So the idea of like changing a date field and people actually just having to like, uh, you know, from the standpoint of do some minor development goes was way simpler. But also, I have to say I did that in 97 and when I look at that, we had like years and years and we also had like major PSAs going, like everybody, the whole world knew this change was happening, but it was a date certain.

And so people often ask, well, you know, this could happen at any time. Do we know where the adversaries are on this? No, we don't.

We have, we absolutely don't. But I think that that's because of the complexity of the project. That's one of the reasons to sort of kind of get started.

The other thing that people are concerned with is this idea of duplicated effort. So they wanna be aligned around the third parties that work with them that they can't really know because they're a SaaS application or they're the cloud or they're, you know, there's some black box element to it that just keeps them from really having to be concerned with them. But also the way they're sort of pointing them to like, Hey, look at my website.

You know, like, yeah, we're talking about this. It's another form of what Bob was talking about. The third party checklist.

So the confusion, fragmentation, this is sort of has been the state of the industry and I work a lot. My deep verticals are financial government. I've worked with government as government, has been a client for 28 years.

I've been in cyber for 23 years. I worked across 20, 26 countries, four continents and, um, both public private sector health. And I would say the first movers in this are really like telecom.

And they're also, um, financial services. I think they have some really good consortia. And you have to find your vertical, you gotta find your people and you have to, you know, like spend some time there because you don't wanna redo things that people have already thought of or already worked on.

Okay, so this chilled me to the core. So that person down there is Jay, Jay Gata and he's, um, now officially the head of IBM research. And he just took Dario Gill's place.

Dario Gill has been a fixture for a very long time. And he is now going to the Department of educ, not, not education, I'm sorry, there's no department of education. I'm in the energy, he's going to the Department of Energy.

God bless him. You know, so Jay is picking this up. So Jay was, you know, I I told you that I'm an investor and sponsor for Quantum World Congress.

I was sitting in the audience. And I have to say, when you go to as an, even as a, when I was an I bmer, when I would take clients to Yorktown Heights to the Quantum Center, you just get chills. It's like going to Willy Wonka's factory.

'cause it's just so unbelievable. And the people are like, so like entrenched and so confident. And I'll tell you what, and since I've been working PQC since like 20 21, 20 22, and every deadline and every um, milestone that they put in front of themselves, they've more than met.

They've actually, I would say there's a bit of a speed up that's been happening. And so while NIST is saying 2030, he actually put this slide up and he said, you know, we're gonna have a fault tolerant, um, quantum computer in 2029. So wow.

Big gasp from the audience, but something to consider. And also they, they're transitioning from quantum on devices to quantum computers. But there's much more, like you hear in the press all the time, people talk about qubits, but there's other parameters that need to be measured in terms of progress.

When you're looking across these kind companies, um, speed, scale and quality. And he goes through all these numbers and you're just like, oh my God, we're behind. It's just, it just gave me a little bit of a headache when I was sitting there listening to him.

But they are very credible is what I'm trying to say. And I don't say that as a I IBM er per se. It's just that they have the, um, I've seen the ecosystems of, you know, pretty much everybody that's sort of in this space and the progress that they've been making.

They had a very, very early start. So look at this list and look at a good 'cause. It's your third parties.

So everybody on this list has signed up for a collaborative research, um, agreement, r and d agreement with the federal government, the cra and a lot of people came to this in the beginning, like kicking and screaming. 'cause you cannot have a patent out of any of the work that you produce. And it's also like you're just sort of giving it over to the government.

So at first it was hard to get people to be part of this and they were having trouble with their organizations, then it suddenly became fashionable to be on it from an optic standpoint. But then you still have the 80 20 rule. You have the 80% of the people who are, the, the purpose of the crater was to work out this migration, to work out interoperability and performance.

'cause before I said that these, um, new algorithms are not drop in replacements, larger key length, um, larger blocks, uh, block cipher, block size, um, very dependent around performance and reliability. And so when you consider all of that, you need these folks. So they've been in the toolbox, sandbox working it for I would say maybe about three or three or four years.

And some of them have come away, walked away with beautiful relationships with each other and not much to show for it. But others have really done the hard work and have come up with, um, some great evidence of that and, and the ability for you to go out there and test the work that they're doing today. So one of the ones that I wanna show you that I really have been impressed with is, um, CloudFlare.

So anybody here experience, um, heart bleed in 2014, if you live through heart bleed, that's kind of, I always like to use the heart bleed because it's very similar to sort of, if we were to find out that things were compromised, it would feel like that. And if you remember with Harpley that that was just a vulnerability, but it was also, um, you know, a situation where we really didn't know where all our assets were. And that was the hard part was sort of like, we don't know where everything is.

This is gonna affect, you know, between 40 and 50% of the internet and we've gotta go out here and patch open SSL, you know, and it was like a previous version. It had been out there since 2012 and then they discovered it in 2014. It was just, it was mayhem and you drop everything.

And we don't need to be in that situation this time because we actually have some warning and we could actually get started. And so this shows the, um, post quantum encrypted share of their H-T-T-P-S request traffic that they're seeing. And they are using, um, quantum resistant algorithms out there today.

And they have, I saw them, I spent some time with them in London, uh, in June. And they were saying that they actually are, oh, am I in trouble on time? No.

Okay. Oh, I wanna leave time for questions 'cause I want to hear what you guys have to say. We do we need to know what they're saying?

No. No. Okay.

Probably a safety briefing. Safety briefing. Okay.

Okay. We had to stop to hear them. Please go.

Oh, anybody here play bingo. Okay. So these are, they're a good example.

Like I said, there's some folks that are working so earnestly and they will help you with your plan. No point in you doing work that they have done already. And so Mackel Mosca is one of my partners, his company Evolution Q, they've worked extensively in the QKD arena, which is, uh, quantum key distribution, another form or another strategy around, um, protecting against, uh, quantum threats.

And he has this theorem that basically says that for all of the, the long held data that you may have, plus all of the time it would take to actually migrate it against the threat of the, um, timeline, kind of tells you where your organization is from a, um, procrastination standpoint. And so we use that the augmented theorem quite a bit when we're having these conversations with people at the board level, it's very hard for them to have anything other than a risk conversation that impacts finances. If it's not about finances, just get the hell out.

You know, like there's not, that really is the kind of the place where you have to come to. You have to really understand the core systems that, and you have to have some agreement. You'd be surprised organizations that don't have data catalogs, that have not prioritized systems or data, and that there's not agreement amongst the people at the very top about what systems are like, not only operationally important, but the ones that are like, Hey, if this thing goes down, we, you know, we, we have no revenue.

We, we, we're not even, we, we can't produce anything. And so elevating this project to enterprise risk management, if you have a risk manager in your organization, they should be like all over it, but in a, um, educated way. And that's what we do.

We do help people sort of sort through their own risk profile and sort through their own set of priorities. And, and I find that when we do our stakeholder groups, we do, um, exercises and we do simulations, but we also do these, um, stakeholder workshops. Oh my goodness.

The legal people know more than everybody about data flows, data mapping, like who's, you know, what's connected to what It's stunning sometimes, but they kind of live in that world. So that's the reasons for the urgency, I would say. One to point out is that the cost of waiting, so there's a real limited pool of people that are PKI engineers or people who are actually do the testing that understand it today.

I, I believe that there doesn't have to be a million buts in seats. I believe that there will be, um, repeatable processes that people will buy vertical start to understand and get, and they'll be able to do a lot on their own eventually. But if you go to most of these organizations that are like sometimes really big and they even have, um, you know, real focus on this, you're lucky to find five people that are, that sit in this space that understand this or that even can like sort out an the RFP or the RFI or write up their set of requirements or work with, um, consulting the big four that do all of your, um, audits, like both the internal audit and the external audit.

Oh boy man, I I, you have to talk to me. I have some stories for you. I'll just, I will, I'll leave it at that.

I'm on camera so I can't like say what I would really wanna say, but it's, it's malpractice at a level that, you know, nobody should be learning on your back. 'cause everybody's learning. And that's the thing.

Like you have to, there has to be some acknowledgement that this is like unprecedented. Nothing like this has happened before. So, So anyway, so one other little element that is very important here is that you have the RSA show every year, and it is the premier place that everybody kind of goes to, to like sort of fear the updates and stuff, and it's kind of strangely silent on this topic.

So, and I would, I would just, I'll just stop there, but like, catch me on the break and I'll tell you. So this was 2023 and one of our folks was up there and it was like it caused a, uh, press firestorm after Ady Shair, who's the s and RSA sort of said, oh, this con computer thing, you don't have to worry about this. That's 50 years from now.

Like, I don't even know why we're talking about it. Which he's important. He's the s in the RSA and Idy, you know, says that.

And the media just goes out and, you know, it's very similar to what, um, uh, Jensen Wong did it. Our, um, CTA, we had a CCF show this year, and he just offhandedly sort of said Quantum's like 20 years away. And literally all the public companies, they're stock up like 10%.

And people were like, wait a minute. But he did kind of go on an apology tour and even has invested, I think in quantum newum, uh, invested in a, um, uh, quantum company, but there's not an admission that this is sort of the folk space or it isn't, so to speak. And so people are very influential when they say these things.

It has a chilling effect o over a group of folks who already have so much responsibility and so much to worry about that it's easy to keep putting it off. So go to 2024. So I'm up there on 2024 and I gotta say it was the most incredible experience.

I have so much respect for the cryptographers. I mean, these people are like super crazy brilliant. And to be on conference call with the rss NA like, you know, Lynn Edelman and, uh, ADI Shair and Ron Ve and also other, the other people, um, Craig Gentry is like, was like mathematician of the year that, that year.

And, um, t Rabin like they are, and they work a lot on, uh, fully homomorphic encryption and other things that can protect us from these kinds of threats. But it was pheno, it was a phenomenal experience. But I say this to say that my primary role, I'm obviously not a cryptographer.

I have great respect and reverence for them, but my primary role was to say, when Adi Shamir says it's 50 years from now to like, you know, impress upon everybody that it's sooner than we think. And I think I accomplished that. And this was applauding.

They were like, oh my goodness, you said it. So let's go to the next one. In 2025, I sat in the audience next to Ron Reve and he's the R in the RSA and we were talking about this and I said, well, what do you think is gonna happen this year?

You think Heidi's gonna say that Quantum's 50 years away? And Ron is, um, you know, an an amazing, amazing person. He said, you know, I talked to him last night.

He goes, I don't think he's gonna say that. But then I guess like near the end of it, he says, well, you know, we, I don't know what's going been going on with Quantum. We've had no submissions to anybody.

Nobody's talking about quantum. And the fact of the matter is that I actually had a number of companies that I work for. 'cause the other thing that I, I work with companies who are early stage in the discovery, uh, crypto discovery realm and a lot of other smaller quantum companies.

And they had actually, I know for a fact that at least I knew about 16 submissions that just didn't make it in. And so is this a nefarious plotter plan? I no, I don't think so.

But I think that when you've had encryption that has been long held, and I mean, God bless these guys, this stuff is held since the seventies. I mean, it's pretty amazing. And they're pretty amazing.

And you know, it's hard to wrap your mind around the, the numbers that I showed. You know, when you compare classical and quantum and also the use cases, there aren't a lot of use cases because the use cases are very private. People pay like 20 million to have these quantum instances and, and to do, um, you know, the level of innovation and work on them.

And with that investment, they're top secret. So I know that was always a challenge at IBM, we really couldn't talk a lot about use cases or where things were in terms of actual, you know, go to market. So he, so yes, so we're, I guess all these companies are gonna all try again and try to be on the agenda for 2026.

We'll see what happens. Okay. So the indu industry is a bit of an echo chamber.

If you go to any of these conferences, it's like we're all talking to ourselves. We're not talking to security operations. People's very in the weeds in terms of, you know, really like pure cryptography.

It's all a lot of a math discussion and we know we've gotta change that. And we're trying to do more streamlining of that. But also I would say that, um, the, there's a little bit of, of, um, geopolitical concern in terms of the way that organ organizations and consortia are really sort of experiencing and or exhibiting some signs of sovereignty where they're like, well, we're gonna pick up the ball.

We're gonna go here and we're gonna pick up the ball. We're gonna go there. And what is, what we need as an industry is to coalesce and we need more harmonization so that everybody's not running.

'cause we all gotta interoperate. Right. Debbie, on that one, where is the create a roadmap for a company?

Ciso, CIO? It is, uh, it's coming up. Okay.

So I was just getting ready to start on that one. And that's a great, great question. Thanks for asking that, because there are things that you can do that's the, 'cause this is the thing we can like boil the ocean on the conundrum of what's not right.

You know, we are admiring the problem, but yeah. What, what can people do? So this, this idea of, you know, where to look for cryptographic assets, it's everywhere.

And that's what makes it complex. I think it was the CISO at at and t described it as looking for, um, I think he described it as looking for all of the, um, nails in his house at that were at right angles or something. Like it was just, I was like, yeah, it's probably quite a bit like that.

Also common misunderstandings, all of the language and the terminology, like people use a lot of different terminology. I like to say quantum resistant. I don't necessarily think anything in security is safe.

So I don't, I never liked the quantum safe, um, terminology, but it's used a lot. I would also say that, you know, people need to look at this as an encryption upgrade when they, when you're in cyber and you hear anything about quantum think in terms of, oh, this is an encryption upgrade and they're gonna do the same thing they did to me with ai. It's gonna be all on my, all on me to get it done.

But the reality is that there is a lot of cross-function when you start getting into these basic components and elements that you cannot get it done without the developers. You can't get it done without the system owners. And you mostly, mostly, mostly can't get it done without stake holders and ownership at the top of the organization that's giving everybody the same KPI, everybody's gonna give it the same attention.

Everybody's gonna show their progress. If you don't have that buy-in at the top, it's a, I I've seen it in government. Well, let's give it to a, you know, level 14 person and he's gonna run around trying to elicit, you know, responses from everybody and not have the sponsorship that's needed.

'cause it's that it's that critical, it's that important. So project maturity. So you'll see people saying all the time, this is what the project looks like, and I just want to really reiterate the preparation part.

Like, if you don't know enough about your data, you'll spend a lot of money. You'll just spend a lot of money and you'll spend a lot of time. And so I was, I had a, um, workshop where I actually did have, um, legal counsel and communications and some other people in with the soc people and encryption services.

Um, the incident response team, it was, there was a whole, you know, cross section of folks. And in that meeting it was amazing to get them. And this was like a two day workshop actually in the, in the workshop.

It, it was interesting to see how the encryption guys were on their heels because it's, they're very siloed. Like they were, you know, they do, they, they give notifications, some of them on, um, you know, ca expirations kind of where they live or the thumbs up or thumbs down on new applications or they, you know, they're, they're not, nobody is looking at this from an enter at an enterprise level and broadly it's very siloed. But what was really amazing was how, uh, in this one organization, there was not any recognition that there was actually, um, any sort of data catalog.

Like people were finding out in real time that what, oh, good, I don't have to go try to prioritize what's important first. And so these types of convenings are really important, but without the board and the C-suite, like saying this is important and without them really understanding it, like it's not just for cyber people. They gotta get the other people.

I think I have, I'm gonna run outta time. So this is a, um, sample, I won't say the company, but it's a sample of some of the reporting. People say, go out there and get your scan done.

You know, get your, discover your inventory and all you'll discover is that everything's red because everything's red, you know, and, and everything is a lot more, um, the scope is huge. And what happens is that people get that first set of information on discovery and they put their head down. They don't take that to a board or their boss or anybody.

It looks like everybody's failing. And that's not the case. Everybody's in the same boat, but it's also not in context.

And I work with a few companies who are working to solve that problem, the contextual piece, so that when you do have discovery, you're not looking at, you know, um, trying to decipher and match and do so much of this manually in terms of finding all your assets. And so there's a guy, Dr. Mark Teran, who has a company and it's called, um, Cusack, I think it is called.

Anyway, he talks about this in our global trust call. And he did such a good job of it that I wanna, I want to, I know he's on YouTube, and so I'll make sure that we get links to people to sort of look at what he has to say. 'cause he also addresses this idea of AI and quantum convergence, which is a whole other topic.

It would take hours to talk about that. And it's, there's some important things to be said about that 'cause and also supercomputing. So some people think that they'll converge and supercomputing will outrun quantum.

And it, it's just, it, there's a lot that keeps people in the pocket holding next. And then when you look at the tool categories, a lot of people are considering that tools will save the day. I'm actually a big proponent for tools.

Like we can, we're never gonna get this done if we're all just looking at and doing this with, you know, consultant help you. Really, the tools are coming, but they're a patchwork quilt of tools and they are coming together in a way that's going to be great when it actually happens because we need this more than anything. But these are the different categories.

And some people are building extensions, like with EDR, they're doing that quite a bit. And there's been a few tools that I have recommended to folks that were as good as the, you know, they were state of the art for where they were, but this is this space watch this space. It's constantly, um, advancing.

Next. And then this, I love this slide. So we just did a big, uh, briefing for 137 banks that are like Russell 3000 banks about what they need to do.

And one of the big complaints, like 80% of their members say, nobody's paying attention to us. We've gone to the vendors. They don't tell us what their PQC plans are.

And we take a picture of this. 'cause you can go to their sites and get, um, their most recent plans or where they collect the data about how things are advancing. AWS has a lot of, um, uh, information that developers can use today.

People can make a plan that says, Hey, we can, we worry about this legacy piece, but everything that we work on from this day forward, you're going to at least test out the new algorithms on, you know, like a policy like that at least has you with a plan for going forward. And so this is what we believe. So we believe in, you know, the priority being like your core asset, like your really big thing.

Spend some time on that. So the financial and privacy piece boards care about that. They really do.

Like, they don't wanna be caught flatfooted on this. The visibility thing. We believe it's a hygiene situation that should be, you know, it should be like the, the way we move going forward.

It just never has been historically that we have enterprise visibility tools are coming for that. Um, we also believe that as you're doing this testing, you have to, if you've developed a sidecar for testing, you've gotta be able to have the capability to roll back. So hybrid schemes are what people are working on today.

And then we, um, also the third party thing, just do it before the crypto inventories. And then, um, this last here, one right here, establishing cross-function KPIs is like the way, like it can't be you alone. You have to have help.

And I've run outta time. I think that's the last one. That's the last one that everybody appreciate your attention.

Hopefully, hopefully I haven't made you more apprehensive. Just know that this, you're not alone or behind. Everybody's in kind of in the same boat.

Thanks so much. Hi everyone. Welcome back here to Tech Drunk tv.

I'm really excited to introduce you to our next guest. His name is Israel Mazen. Israel is the co-founder and CEO of a company called Mexico.

We're going to hear all about it, but first, let's hear all about Israel. Israel, welcome to Tech Drunk tv. Thank you for having me here.

It's a pleasure. Israel, I mentioned you're the co-founder and CEO of Mexico, but you know, that's relatively recent. Talk to us kind of about your arc, about, you know, how you got here today.

Yeah. So thank you. I have, uh, me and my team, uh, more than 30 years experience in building software companies, especially in cyber.

Uh mm-hmm. We had companies that were public in the nasdaq. We acquired companies, we did exits, and uh, actually we came together to, uh, four years ago to, uh, build this, uh, to build Mexico.

So, uh, and we have, as I said, a long time experience in building worldwide software companies. And this is, uh, why we came back together to the game, uh, uh, to build another, uh, large independent company. This is our goal in Mexico.

Absolutely. So you got the gang back together? Yes.

Yes. All the, we have all the founders that we were many years together, uh, we gather back together. Yes.

Excellent. Um, you know what, as long as you mention them, maybe they're watching this and they'd like to hear their name mentioned. Who's, who's on the founding team here with you?

Okay, so, uh, with me we have Ellie Macia. That with me is the CTO and the brain. Mm-hmm.

Uh, of all the products. Mm-hmm. I have Gidon Hasam that is also the founder, co-founder and COO and, uh, customer success, head of customer success.

Uh, amazing that my brother also COO long, long time with us in this, uh, journey. And then, uh, a few other that join us that working with that 30 years Dov gal, the CFO also working with us, uh, in the last 30 years, including, uh, public companies, private companies, uh, z that is the head of r and d, all of them working with me in the last 30 years. Beautiful.

You know, Israel, like you, I've, I've also been an entrepreneur 30 plus years. It's a long time, right? And, um, I'm mostly in cyber as well, right?

Until I, about 10, 12 years, well, almost 13 years ago now, I stopped all that. I started writing and speaking and somehow wound up from a blog became What's Today, text Drunk. But I also, I surround myself with people that I've worked with, 2025, you know, 30 years.

I, I know how that is. But you can't just say, Hey, let's get together. We'll have, you know, we'll have a couple of drinks and we'll think about what we want to do.

There's gotta be a, a problem that calls to you that says, Hey, this is, this is a problem. And it's not just a problem We've seen ourselves. It's a problem I know other people are having and let's, let's build something that solves this problem.

It's not enough just to build a big company or a successful company that at the heart of it, at its core, there's gotta be some problem that you're passionate about trying to solve for everyone. What, what is that passion? What is that challenge?

At Mexico? This is true. When we, uh, gather again to do this company, we saw that there are a lot of scams that, uh, uh, uh, around the world that all of us can fall, fell to these scams that actually try to steal your credential, try to scam you in a fake sites, fake social media, social engineering, and actually try to, uh, steal your identity.

And then do a lot of stuff like stealing money, stealing goods, stealing, uh, frequent flyers. It depends. But the main thing that we saw that happening significantly also to our family members and friends that trying to actually, uh, get our credential still our identity.

Uh, I think that everyone, I believe that you see it, everyone see it happening every, every minute, every second, all over the world. So this was something that we saw and we said how it's, uh, not solved, uh, really in, in good way. So after we understand it, we talked with more than a hundred potential clients to see if they think that they solved it, why they didn't solve it.

And this is what we understand that the current solution that they had before we developed and started to sell our product didn't really solve their issues. They start maybe part of it, maybe give them some insight, but didn't solve the problem. So this is why we gathered again and we said, we must actually solve this problem.

You know, Israel, they say God works in funny ways. Just last week, just last week, friend of mine reaches out and tells me that they got a, a, a funny Instagram message from someone from me. I said, I didn't send you anything.

Well, let me see what you got. Sure enough, they got a, a follow request from a new, a new Instagram account. Shimmel Allen, no picture, nothing Shimel that.

Allen, I look it up. Sure enough, he is follow or they are following 85 of my friends, they're following. And you know how it works, right?

I follow you. You say, oh, that's Shimel. I'll follow him back.

Yeah. 12 of my friends or connections followed him, followed this account back. So I, I reached out to those.

The nice thing is on Instagram, I could see who, you know, who followed who. So I reached out to the ones who followed and I said, Hey, I see you're following this account, Shimel Allen, I want you to know it's not me and you should block and report it, but let me know, did they reach out to you? What's the scam?

Because this has gotta be a scam, right? As you said, yes. And sure enough, about half of them had gotten messages once they were following about clicking a link for the Bill and Melinda Gates Foundation.

Another one was about a book or something. They all different sort of scams, right? With, with the, with the idea is that these people should click links.

So they were using me as a trusted source to, to, to really get people who are connected to me. Now, I wrote to Instagram, you know, I went on Instagram. I said, someone's imitating me.

They're imitating me. That's a scam. You know, you get the automated message back.

I'm very sorry, this doesn't violate our community standards. There's nothing you could do. They gimme a suicide prevention hotline.

If I'm that upset about it. I wrote an article last week about it. These companies don't care Meta and, and, and, and Google and what have you.

They're not, you know, they talk a good game, but they don't help you when something like this happens. How do you help when something like this happens? You describe exactly what's happened to many organizations in banks or airlines or retail or hospitality, that you get fake, uh, social media advertisement or profile or fake link to a site that look like the their site.

You. And, but it's not, it's a fake. And actually what they're doing, they get your, uh, actually identity and your credentials, and then they can do a lot of stuff.

Transfer money, uh, buy a vacation, uh, with your credentials and many other stuff. Uh, steal your credit card number. So this is the man you describe exactly what's happening in many type of vectors, like fake sites, fake social media, fake profile.

What we are doing and what we developed, that it's a completely different on the other solution that were in the market. That we develop a solution that can in real time where to our customers base, uh, that it's, uh, banks, large banks, or retail airlines. Um, so it's give them in real time when it's happen and they send to you as the, their user.

Some like fake link like this or fake, uh, social media or whatever we know to detect it in real time. And even when attackers start to prepare an attack before we attack you, we have a way to, to start detect it. And we put it as an event, an additional event until we know that it will be an attack for sure.

So we can actually alert to our customers, these users can fail to these scams can be thousand users, 10,000 users, a hundred thousand users or more users. And then we can, uh, also, uh, protect them by sending fake, uh, credentials to the hacker or they, or, uh, stop their, uh, devices or whatever, uh, that they cannot access, uh, to this account. So, and many other protection capabilities.

So our main, uh, difference or advantage that we know to detect it, our solution is, uh, preemptive and in real time, and we can detect all of this, uh, that happening to our customers in real time. And, uh, protect the users. We can, I think, prevent more than 70 or 80% of account takeover that happening to the users of the organizations.

Israel, you mentioned you founded Mexico four about four years ago. Yes. But, uh, recently you guys announced a series a, uh, round of financing, a healthy series, a $37 million.

Usually you see a series a, a year, two years, you know, into the company. I'm gonna imagine because of the, the history of you and your team, you kind of bootstrapped itself, funded for, for a couple years before you went out to raise money here. But anyway, congratulations on the raise.

Tell us a little bit about who, what, why, why now. Okay, so yes, we raised 37. Our goal was to raise 25 million, but we had very high demand for the round.

So we, uh, increased it to 37. And, uh, we did it now because we are going, uh, three times zero over year in, in our a RR. And we want, we will plan to continue growth three times year over year.

And, uh, it was the right timing because we plan for the next few years the same growth. So we need to put the infrastructure, uh, expand our sales team, support team, uh, customer success team. Also, we are planning to developing more and more products on the same platform.

So we expend also expanding our, uh, development and QA team. And so this is, uh, it was the right time for us to raise another round. Uh, and also the investors that invested first account investors, couple ventures and venture guide that was in the seed round, invested in this round and the new investors, one of them is our customer, customer of US National Bank of Canada, that they saw the strengths of the product and the company.

So they invested the, they actually one of the leader in this round. And then we have, uh, PAG PS group that it's, uh, the family office of Steve pca. It's a guy that, and the, uh, is the board of, uh, he is in the board of, uh, uh, garner.

He's a senior advisory board of Bain Capital. He was the owner of both of Celtics till August. Very well known that can help us a lot already help us to bring customers and potential partners.

And the other one, it's a very wealthy family with many, many businesses in Latin America, in Dominican Republic, uh, the Leon family. And, uh, they can help us. They already introduce us to their businesses and, and potential clients for us.

So it's bring us not just money, but potential for many more customers. Strategic investors. Yes, exactly.

Strategic investors. Yes. Yeah, that's excellent.

Excellent, excellent. Um, you know, look, today everybody is with AI and AI and AI and more ai. How is AI changing the game when it comes to digital impersonation, account takeover, fraud, et cetera?

Yes, it's very, uh, very good. Uh, question because what's happening today that with using ai, using AI kits, uh, phishing as a service came as a service. What's happening that everyone also almost can do is to these scams.

You don't need to be sophisticated hecker. And also you can do a lot of them in very short time because you create them through ai. So it means that there are many more attacks.

I can tell you that the estimation that in till, uh, by 27, the aggregate amount of losses from this type of scam as the account take over re Ford will be over $300 billion over the world. It's a huge losses. So the, the, it's changed completely the, because everyone can do today these scams and very fast with ai.

So what we are doing also, and we develop also solution and we continue to develop it, that will know to detect all of these AI scams that coming from, uh, these fish kits or these scam kits or AI generated by AI to detect it immediately and then protect against them. But it's make you much easier to do these scams and much faster. Absolutely.

Absolutely. Um, so I, I wanna make clear, right, your MCO sells enterprise customers, right? Yes.

Organizations who are susceptible to this. There's a huge problem, like me as a cons, as a single person, as a consumer, any plans may be to offer something to help people, you know, just regular people, not organizations. Yeah.

So for now, we are more B2B uh, player, and, uh, we are selling especially to organizations, but I can tell you that there's more and more organizations will use solutions like us and our solution. So you as a customer will be protected. So let's assume that most of the organization will have something like this, then you'll be protected because, uh, you cannot actually, when you access, you get protected.

For now, we are planning to stay in B2B, uh, play. Yes. Now you guys can do a great job, right?

Have a great product. At the end of the day though, some of the responsibility has to be on the social media, the email providers, the exes, the, the, the Facebook meta, whatever, uh, you know, apple, Google, all of these companies that, that, you know, they, they just don't seem, I mean, they talk a good game, but they, the, the actions don't match the words. Sometimes in, in taking this as a very serious problem, you're in a great spot because you represent, you'll be having all these companies that you are, they're customers of yours.

You'll maybe have some leverage. What, what can we do at that, on that side of the equation? Yes.

I think to try to make it better. Yeah. I think first regulators are more, more serious now about it.

So they, you'll seek, uh, countries in uk, Europe, now starting in the us Australia, that the regulators will force the organizations to protect their customers. And if not, they give them penalties and they need to give the money back and penalties. So we see it more and more that regulators are, start being more serious and care about this.

Uh, and this is what I think will happen because I think that, uh, to protect the real users, the end users, uh, uh, or regulation, uh, regulators will force the, uh, organizations to protect their users. And this what will mean that they must, uh, uh, deploy solutions that can protect their users. I think they'll happen more and more, uh, in the next few years.

I will tell you in the case with me last week, I don't think they counted on most of my friends being cybersecurity people. So when they reached out to my friends, my friends saw it was a scam, and they all, you know, did what they had to, they gave 'em a good time about it. But if not, you know, that was me.

I, I feel bad for other people. Yeah. Israel, for companies out here watching this saying, you know what, this is a problem.

We'd like to maybe check it out. What, what's the on ramp? What's the best way to engage with me, Mexico?

Okay. So, uh, we actually, uh, selling directly, so they can of course contact us through our website. com, all the contacts, you know, they can contact us and they'll get answer immediately.

Uh, we have people around the world that's supporting them. We have channels, uh, that's selling us like Deloitte in emea, in Latin America, local channels in Italy, Spain, Latin America. So, uh, when, if someone want to approach us, we, they can approach us through our contact information on the site, and then someone immediately will be in touch with them.

com. Yes. com.

M-E-M-C-Y-C-O. Very good. Israel, first of all, congratulations on the oversubscribed series, a more importantly, thank you.

Thank you. You're welcome. More importantly, look, I could tell from personal, tell you from personal experience, this is a real problem.

It's a problem we need to solve. It's a problem we gotta take seriously. So I, I wish you lots of success, you and the team in solving this problem.

Thank you so much. I, I agree with you. And, uh, I believe that more and more customers will use our solution, and then the customers, you end user will like you and other will be much more protected and do safer, you know, transactions in the internet.

Absolutely. All right, Israel, ma Israel Mazen, co-founder, CEO of Mexico here on Textron tv. We're gonna take a break.

We'll be back with more in a little bit. I am Mitch Ashley of the Futurum Group. I'm Scott Ban with Solutional.

We're here today to give you an overview of the Nokia data center Fabric reliability study, a survey that addresses some key issues in modern data center networking. We ran a futurum research survey of a hundred IT infrastructure leaders from large enterprise IT organizations with a goal of understanding how data center network reliability is decided and measured both today and into the future. Mitch and I wanna cover three main takeaways in this video.

First, reliability is the number one decision criterion. Second, operational challenges, especially human error, still drive incidents. And third teams claim meaningful automation, AI ops, adoption.

And we wanna unpack that a little bit. Yeah. Three important messages.

Number one, though, reliability is not a nice to have. It anchors the design, the design, the operations, and ultimately in the business outcomes. Resilience is the end game.

Yeah. Not a huge surprise, right? That reliability was the top priority.

Um, there's some interesting supporting stats that go around that. Mitch, can you talk us through 'em? I think first of all, 86% of the respondents ranked reliability as a top decision criter.

So it wasn't just, just above the midpoint, it was well almost, you know, get, you don't get 86% in, in responses for very many questions. Uh, and the things that, that it was sat on top of were the ease of integration operations. We know those are also challenges.

So why does this matter? A single hour, hour downtime is a widely expected to hit service levels and a also revenue. So 47% foresaw a major service disruption risk, a 68 expected direct revenue loss.

So it's a big deal. 74% of organizations said they had greater than one incident of an outage in the past 12 months. So it's not a rare occurrence.

Um, when we see that many organizations saying they're having at least one out one outage a year, and that can be due to hardware failures, human error. Those are common top causes. So now regarding human error, let's talk a little bit about that.

We saw that amongst, um, multiple operational challenges, um, that drive those, uh, that drive the outages and incidents that we're seeing. What, um, more is underneath those statistics, Mitch? It's a significant factor.

5%, called it a frequent top cause, things like that. So it, it's certainly just more than a factor. It's an important aspect of when there is an outage, but it's also more than that.

Um, there are often other failures that come alongside with human error at some point in that process. That can be things like hardware or software failures. So how do we address this?

When we asked the respondents, 35% said that they emphasized strict process and training. 25% said focus on resilience and recovery. Recovery, and only 12% said they aim to eliminate errors via automation.

Meaning we know that errors are gonna happen, but we have to be able to handle those, respond to those we wanna resilient architecture, implementation, and also as well as the implementation or the automation that we're doing. So, you know, teams are struggling to meet the evolving needs of the business because we know those are under constant change and also limit the scope or, or run extra planning cycles. Those are things that, that they're struggling with.

Oftentimes they'll even postpone important tasks due to confidence levels, uh, when they're not sure if that's something they're ready to implement or if this is the right timing to do that. Last but not least, of course, skills always come up, but it's a significant gap. 54% said that that was skill gap was an issue, and several in incited cited that state versus desired monitoring limits were a factor as well.

So on the implementation and use of automation in AI ops and the actual adoption, um, versus interest in automation and AI ops adoption, what did you find in, uh, in that bucket of responses? Well, they, they said that here's what they're using today. Uh, a 67% said that they're using automated monitoring.

50%, actually 58% said they're using infrastructure's code. I've particularly found that interesting. And of course that's, uh, you know, followed by things like ticketing, auto failure over, but ai, ML based incident prediction was pretty significant at 54 4%.

So I think this says that we're investing in A IML as part of the, the solution set, but also I think we know that, you know, tooling does not necessarily equal positive outcomes. Only 36% reported dedicated AIOps tooling as of now. And many are advanced practices are still in the maturing stages.

So, you know, adoption is both planned and underway, but separating tooling use from operational reality gains is still key. Yeah, that separation, uh, and, you know, that fine, fine grain understanding of are we just interested in automation and AIOps versus we're really, you know, going full force. We're gonna see that journey continuing I think for years with many enterprises really just getting started in earnest.

Definitely tracks. I agree with you So much that you covered in, uh, in this survey. We're only touching the tops of the trees here.

Where can people go to get the full report and, and plow through this and understand the fuller picture? com and download the report from there. There's a section for analyst reports and the latest analysis that we've done.

We'll also include a link with the video to make it easy to go right to the report. It's free, download it, you've got, you'll in seconds, you'll be looking at some really compelling and interesting information. Definitely agree, Mitch, thanks for the pointer and for the readout.

You bet, Scott. Thank you. ai Leadership Insight series.

I'm your host, Mike Bazaar. Today we're with Dr. Arun Subramanian, who's the CEO for Articulate, and we're having a little chat about how AI will go into various vertical industry segments as we kind of develop more domain specific languages.

Dr, welcome to the show. Mike. Thank you so much for having me.

Yeah. So what will it take to accomplish this? 'cause I think if I look at most of the solutions that people are using today, their general purpose, they're horizontally oriented, and yet in each vertical industry segment there's different nomenclature, different terms, different workflows.

So how do we kind of flip this on its head as it were, and make this more vertical industry friendly? So actually, if you think about it, like the models have gotten significantly better, right? So the general purpose models, uh, today I would say have gone from being a middle school student to a very competent high school students, sometimes even, uh, a competent, uh, uh, college graduate.

However, the nuances of individual domains, as you said, are completely lost. And, um, like, what I mean by that is you can take, um, um, say even a high school student, give them a textbook on something very complicated. Tell them that the answer to the question lies in the textbook and go find it.

As long as they know how to go look at the indices, they can go do the table of contents, figure out where mostly things are, they can go retrieve something for you, maybe even the right answer. But then when they do that, they don't know whether this is the right context. They certainly don't know whether it's the right answer and the person consuming it has no idea whether the depth of the information has been considered before giving you an answer.

And that's what we mean by domain specificity versus generalist giving you an answer, right? Mm-hmm. And, um, the, the flippant example I can give you is if you have a question about, uh, say brain surgery, you go ask a brain surgeon.

The brain surgeon looks at a textbook, analyzes, uh, the information you're giving them, and then gives you an answer. A high school student looks at the same textbook, gives you the same answer, which answer would you prefer? And hands down, it's a very obvious question, but when we go and ask a general purpose model that is designed to understand pretty much everything from a five-year-old asking to give you a, a nursery rhyme to a, uh, financial analyst going and doing deep financial analysis of a particular investment thesis, the answers are more or less plausible, but they're anything but specific and most of the time anything but accurate.

So that delta is what we are trying to to navigate, right? I'll give you another example that may sound very und. So take an example of a case where you wanna understand, um, a table, but then not just understand the table, but you want as a financial analyst, or say you are a safety engineer in a plant somewhere, and you wanna know that if this system looks at a table, it reproduces the table precisely every single time you look at the table.

Right Now, there's no computing system on the planet that will give you 100% reproducibility. 999% repeatable, you needed to have tested your algorithms, your systems at least 10 million times across a variety of different tables. That's what we mean by going from a general purpose system to a system that works in an enterprise in a responsible way.

Mm-hmm. How does that kind of get worked through? And I'm asking this question because a lot of the answers that you get from AI are probabilistic.

Yes. And you know, that generally means they're right. Some percentage of the time that's less than a hundred.

Yes. And a lot of the things in a vertical industry workflow or deterministic, right? Yes.

They're supposed to be done the same way every time exactly the same way. And precisely, um, how do I marry these two things together to come up with something that is, you know, where one plus one equals five or better? Yes.

So actually it is not really as much of, uh, a conflict as, uh, it seems on the outset, right? So for example, when you ask a language model to make a prediction, it's really only making a prediction of a probability. There is no certainty there at all.

And each time you ask it, depending on what, uh, is going on in the system, is going to give you a slightly different answer. However, you almost never put just a model to get you to an outcome. In fact, it's always a system that gets you an outcome.

Even today when somebody uses, uh, Chad GPT or, uh, say cloud or Gemini three, they're not really just hitting a model unless you're hitting a model API, if you're using the application, you're hitting a system. So you ask a question, the system interprets your question before it hands it to a model. And after the model gives it an output, there is a series of deterministic tools that interpret the outcome before you ever see an answer.

So it is a combination of deterministic tools with probabilistic models that make the overall system deterministic. And many times the final answer is something that is retried many times over. Mm-hmm.

And the user may never actually see the retry in between. Right. And are we essentially also using multiple AI agents to check the work of each of the AI agents or the models or whatever it may be, so that we're not overly dependent upon the single guess of one?

Right. That's absolutely essential because, uh, like, so we've, uh, developed a system called Model Mesh, which is now evolved into Asian mesh. And this was developed more than three years ago from a concept standpoint, at a time when it was not at all obvious to people that you needed more than one model to get you to an outcome.

In fact, the the overall narrative was it is a model war, one model or two models are going to win. And we went into this knowing that you need a combination of different models to work together, add runtime to get you to an outcome. And second, even at runtime, your judge cannot be the, the same family as your player.

That's what is the main differentiator. Because even if you say multiple agents, one is generating an answer, another agent is actually critiquing it, the agents cannot come from the same lineage of models. If they do, you're more or less going to go to the average.

Mm-hmm. You at least need the judge to have an independence beyond just saying an instruction set. Right.

And that one, we've seen it multiple times over. And in fact there is a, the recent paper from Stanford that quantified it, they called it the semantic collapse, which is once you get to about 10,000 documents, if all you're doing is vector search, almost all documents look similar to all other documents irrespective to what question you're asking. Hmm.

That's why most of these kinds of question answer systems work really well. When you have a few hundred documents, you go to a few thousand, they start getting fuzzy. And then beyond a few thousand documents, they're completely fuzzy.

And that's the reason why, Um, as you kind of think about this for a minute, we've had domain specific languages for a long time, but they were, shall we say, arcane and only a few people could, uh, master them. And now it seems like we're gonna put a natural language interface in front of them, which then makes them more accessible. Yeah.

So does that mean we'll see an explosion in r and d around domain specific languages? 'cause I think a lot of people didn't go build those because they were like, well, the only a handful of people could use it, but now maybe everybody can use 'em. So actually it is the, the other way around, right?

So where before you needed a domain specific language to describe a set of actions and activities related to the domain and only the advanced practitioners of that domain would even use that. Today, pretty much the interface is natural language. And the translation layer is what is missing.

Because say you ask an aerospace engineer about, um, what is lift compared to a manufacturing engineer? What is lift? It's the same English word means two completely different things.

And the context might also be relevant in terms of what you're talking to them. But knowing that context and knowing the domain makes a difference between giving you the right answer versus giving you the absolute wrong answer. Right.

And to your question of do I need to learn, uh, the domain at all? Do I need to know the nuances of the domain? That's where the unlock is.

Because today a finance person who doesn't understand the industry walks into a new company, they will have to learn the lingo of the company and the industry before they can actually like function properly. Tomorrow. When you have these domain specific models properly deployed, they can ask the same question, the answer will be contextualized to wherever they are.

That's really what is important. Mm-hmm. So what is your best advice to the leaders of various vertical industry segments, whether it's manufacturing or finance, about how to infuse AI into their workflows?

I think everybody's kind of having the same issues. They're like, we love the idea and the concept and we see the potential, but when it comes to the, uh, the actual execution, everybody seems to struggle. Yes.

So I'm, I would say more than advice, I would say I would have three recommendations, right? So first and foremost is we are well past the question of is AI useful? If you are still doubting, if AI is useful, you'll really have to look at people who have actually made it useful for themselves.

Now you can ask the question, I can see them finding use, I can't see it myself. Right? That's where the gap is.

And that's the gap that you are, um, highlighting. The second point to that one is if you accept the fact that we do live in a different world, and you don't have to take it from, uh, a, it's a self-serving answer for me, because I'm running an AI company, you can take it from the fact that we live that daily. We can't operate if we don't find use from ai.

That's number one. The second one is the general purpose. AI systems that are out there are necessary, but not sufficient to get you to an outcome when you're running a deep industry.

What I mean by that is, take a time when email didn't exist, take a time when intranet in a company didn't exist the first time a few companies started implementing those, those were actually significant differentiators in terms of how they operated. They got more productivity than their competitors. But very quickly, everybody caught up.

Today, nobody would say having email is a differentiator, but everybody would say, not having email is a blocker. That's where we are going with general purpose ai. So not having AI in your company in a safe way that everybody can use would significantly be a blocker.

Mm-hmm. Okay. Excuse me.

But the third one is you just having AI is not gonna make a differentiation. What is going to be different is what do you do with your own know-how, with your own data to improve your own operations that only you can do? And that's where domain specificity comes in.

We call that hyper-personalization. How do we take AI systems that are either domain specific or general purpose, make it specific to you and take advantage of that. Now, without that, everybody kinda sort of looks the same.

Like today, if somebody sends you a one page document that looks well written, not that much of a differentiator, unfortunately, unless you can very clearly see the writing is original. And unfortunately, we are very quickly going to a point where original writing is very rare. Hmm.

So as you kind of put all this together though, um, who is gonna take the lead on these projects? 'cause to your point, everybody and his brother is gonna have the same kind of capability Yes. On the tooling.

Yes. So who should be at the forefront of these initiatives? So mainly business leaders who actually have to show business outcomes, right?

So this is not a, I have a technology, let me try to find a use for it. This is about saying, look, I have a business goal today. I cannot meet that business goal.

Most of that business goal is about increasing your revenue, increasing your margins, whatever that might be. Because you have to separate things into bottom line and top line. And I look at, uh, how leaders have to implement this as bottom line has a ceiling.

So you can only get so much productivity outta the system. Top line actually is significantly improving the business. You cannot just do one, you'll have to do both.

And it is about where the businesses at that point in time, in terms of what you do first. But not doing either will be a significant disadvantage. Mm-hmm.

Now, I'll give you some examples, right? Take a simple case of, okay, if you are not somewhere like giving your, uh, people the ability to automatically summarize meetings automatically, like identify action items automatically send that into your systems today. That's a significant disadvantage because not only that you are having somebody in the, the meeting figuring out how to take notes and then figure out where the the action items come from.

But also the way to track it manually is always gonna be slower than what these systems can do today. Today it's trivial. Whether you use Zoom, whether you use teams, anything, it's nearly trivial to be able to use it.

But the number of companies who don't use it still surprise me in a safe way. Right? That's table stakes.

What is that one thing that you see organizations doing today? It just makes you shake your head a little bit and say, folks, we need to just be a little bit smarter than that. So if it's one thing, it's still a significant portion of, uh, enterprises, especially operating like we are still in 2024.

What I mean by that is they're assuming that the world is very similar to what it was in 2024 and 2025. Um, I can tell you, working in an environment where every two days, what we thought was our significant differentiator becomes stable stakes. It, it feels like every day is something new.

But even in very traditional industries, like we operate in manufacturing, we operate in energy, we operate in oil and gas in aerospace, but traditionally, a change like this would've taken 20 years to go through. We are seeing customers change in a matter of weeks. And they're not changing because it's a fad.

They're changing because they can get to an outcome that they can measure that is fundamentally different, right? And the number of, uh, enterprises that we even show saying, look, this is what is happening out there, and we can actually show you evidence. They look at that from a distance and go, but that's not me.

Uh, I, I won't get affected by that. I will continue to operate the way I'm operating. The, the tide for that is going to change significantly in 2026 just because the tools are out there.

Folks, you heard in here, change is coming. It's already here. The only issue is figuring out how to operationalize this whole new set of technology in a way that gives you something that looks like a competitive advantage.

But don't assume that whatever you created is gonna be there for long. 'cause everybody else is gonna do the same thing really quickly. Hey, doctor, thanks for being on the show.

Thanks a lot, Mike. Thank you so much for having me. All right.

And thank you all for watching the latest episode of the Techstrong AI Leadership Insight series. You can find this episode and others on our website. We invite you to check all those out.

Until then, we'll see you next time. One of our predictions for AI in 2026 was that sovereign AI would become increasingly important. And this was emphasized by AWS that their European Sovereign Cloud launch in Potsdam This week.

We expect to see an increasing focus on digital sovereignty in the coming year, and many are looking at using specially developed AI models. But news came out last week of Apple and, uh, Google tying up for next generation series based on the Gemini model. So perhaps this signals a trend toward leveraging external foundational models.

We'll also consider the push and pull between infrastructure like data center and power versus distributed inferencing on this episode of utilizing AI featuring Nick Patients and Brad Shiman of the Future and Group, Welcome to utilizing ai, the podcast focused on practical applications of artificial intelligence from the Futurum group. Every Wednesday, we explore news and use cases of the ways in which AI is transforming enterprise IT and the industries it serves. I'm your host Stephen FoST, president of the Tech Field, a business unit here at the Futurum Group.

And joining me today, we have two of our fantastic Futurum analysts. Let's meet who's on the panel today, Nick? Hi.

Yeah, I'm Nick Patience on the AI platforms practice lead at futurum. Um, and I, I look at everything that's to do, to do with enterprise ai and, uh, hi everyone. Brad Shiman.

I also work at futurum, where I am also a practice lead for a different research practice around data intelligence, analytics, and infrastructure. We kicked off this year of the utilizing AI podcast with a look at some of the predictions for 2026, and we don't wanna revisit those quite yet, but I think that it's worth looking into some in more detail now that, uh, things have started and we're starting to attend events. So, Nick, lemme turn it over to you to talk a little bit about what you've learned, uh, this week, uh, at at events, as well as, uh, what you're thinking in terms of sovereign ai.

Sure. Yeah. Thanks David.

So one of the, one of my predictions, um, for the year and a sort of key part of our research agenda for AI platforms, there was, um, I think seven things in all, um, was about sovereign ai. And, and so I think it's, it's, it's something that was kind of a, I guess a niche geopolitical, um, idea and sort of peculiar in some ways to continental Europe. Uh, and now as I'm sure many of our viewers and listeners know, it's, it's expanded, uh, in importance to more or less every country, uh, in the world.

And there's two way, there's, there's the kind of national sovereign ai, uh, issues, um, but which are, which are very important, but, but probably, um, slightly too grandiose in thinking for, for kind of enterprise ai, um, users. But down at the kind of enterprise level, I think it's, it's, it's, uh, it's gonna be a key, um, a key topic. And, and this, um, and I was recently at this, uh, AWS launch of this European Sovereign Cloud.

Uh, the launch was in Germany. Uh, the cloud is in, uh, Germany, and it's an entirely separate cloud, um, from, from AWS. It's not a, it's not a region, um, such as it is entirely separate cloud that will be, um, uh, staffed by European Union citizens.

Um, and essentially, uh, although it's owned by the parent company, obviously that's, that's unavoidable, uh, is run as a separate, um, unit essentially. And, you know, the, the idea is that it can give, uh, companies and organizations, um, you know, absolute control over their data, um, who has access to it, who doesn't have access to it, where it, where it resides, um, and so on and so forth. And I think, so I think it's, and then sort of specialized security and, and, and all sorts of other things.

So I think it's exem exemplifies one of the things we're gonna be looking at this year. And I guess the, the, the interesting thing I think from it is that that's a hyperscaler doing that. And when you think of sovereignty, um, you know, an American hyperscaler is, is both in a very, you know, an interesting but challenging position.

Um, so there's also a lot of, you, there's a lot of organizations, uh, and not just in Europe. This is not just a Germany thing anymore. This is to say this is a, um, to more or less a global thing, but there's lots of organizations that are looking at, um, cloud and thinking, you know, maybe I want to, you know, have, have more control by having more on premises.

And so there is repatriation going on, and there's obviously a lot of data that, um, that never, that never actually, obviously left on-prem. And so it's kind of, it's a little bit of a dance being carried out here by, by the hyperscalers. And they're getting certifications from various countries to say that they meet security requirements and compliance requirements, um, while also trying to grab a bit of that business and say, well, you don't need to have, you don't need to kind of go to a server, a server and storage and networking companies and get stuff and, and deploy it in your data centers and go to software companies.

You can do it all with us. And so I think it's gonna be, um, a dominant theme in 2026 and 20 and beyond. Um, and it, yeah, I just thought the AWS approach was really interesting and they roped out, you know, Matt Garman was there, the CEO, um, various politicians.

It was a pretty, pretty big affair. And they, they put a lot of effort into it. And so it kind of shows you how important, um, a Ws thinks it's, and, and, uh, I'm sure the others will, uh, the other hyperscale as well as well.

Yeah. And I, I would add to that, Nick, that's, um, it's, it's not just about the infrastructure itself, but also about the software that runs on it. And we're seeing, as you mentioned, uh, and rightfully so, that this is a global concern.

And it's not just about legislative compliance, it's, it's about autonomy and being able to anticipate, you know, the fragility and, uh, you know, chaos that sometimes seems to, to make daily headlines for companies right now. And, um, so you're seeing in Europe and other, uh, regions that have some coordination amongst the nations, uh, efforts to, to basically free themselves from, uh, any sort of, uh, obligations they may have to external parties. And one of the, um, the big sort of unseen bits of fallout from the tariff wars that are going on right now, uh, regards software and, um, what sort of obligations a country may have to US-based firms doing business with the United States.

And so if for instance, you know, you are running Microsoft Office or Google, um, office workspace, sorry if they keep naming it different things, but, uh, anyway, if you're running that you, you sort of have obligations in terms of how you manage data and what data that company can see for your employees. And so we are starting to see companies not just locked down or take control, I should say, of the infrastructure, but uh, of the further up the stack for the software that they're running on that infrastructure. Yeah, totally.

And then what a key part of what, um, AWS are doing, the European sovereign Cloud level is also all the, all the access and identity management stack is also within, within this cloud. And so it's not because there's, you really have to look at any other way. It's not just about the data and, and where it resides.

It's about, you know, who can have access to, um, you know, the, the systems, the software and the data. Um, and so it's, uh, it's, and the, and the metadata, I mean, it's down at that level of, of granularity of, of kind of control, um, that they're, that they're talking about. I wonder if I could ask a question about it.

Um, what about the models themselves? Is there any, um, thought of creating, um, regional, uh, sovereign models that, uh, are restricted in terms of, uh, training data or, uh, fine tuning, uh, that would avoid sort of, um, regional biases, certainly training or embrace regional biases? Sorry, yeah.

Or, yeah, I think certainly in terms of using European, yeah, in training data, obviously there's the link, the list, there's the language aspect to all this obviously. Um, but yeah, there are, there are definitely, um, yeah, there are definitely, you know, specific, um, yeah, specific models. Um, there already has been, you know, there's been, you know, specific ones in, in Greek and stuff, some from Singapore and, and Japan and, and all over.

But, uh, I think it kind of speaks to that general, um, um, you know, interest in, in, in sovereignty, um, all, all over. As I said, it's, I think for, for, for kind of, for organizations, you kind of think about it as a sort of strategic autonomy because obviously they're not, you know, no matter how big you are as a, as a, as an organization, um, you know, you're not, you're not, you know, you're not the nation. And so there's kind of two different ways of looking at it.

There's, there's the, there's the, uh, national way of looking at it, which is, you know, we're gonna control the supply chain. We've been, do everything ourselves. Um, which is somewhat illusion, you know, a bit of an illusion, um, even for the US and China because obviously both interdependent on each other, um, in different ways.

So obviously China, um, is kind of more advanced in the production of energy and has taken open source models that the approach from the US obviously, um, you know, designs, you know, the world's most advanced chips, but they're made in Taiwan. So there's kind the idea that be completely independent, um, is I think an illusion, but from an enterprise point of view is this kind of strategic autonomy kind of lens. I, I like to think about, um, how, how, how it should be looked through.

And, you know, if you think about models themselves as, as, you know, a representation of patterns, you know, in the training data that, um, that model itself is literally, you know, a knowledge base, a database, a, you know, map of the institutional knowledge of, uh, an organization or a community or a, something even larger. And, you know, it's, you know, very much about being able to represent the, the way that, uh, not just the language itself as, as Nick you mentioned, but also the nuances of the way that people that speak that language natively think about themselves, think about the rest of the world, you know, from their context. And I, I thought, you know, this, when I heard this, what I'm gonna describe in a second that they, this person was out of their mind, uh, back in 2022, uh, a CEO of a certain company that, that focused on, um, you know, diffusion models and creating images, uh, said everybody, every country, every city, every person will have their own model that will represent them.

And you know, when you think about the data science that goes into training and fine tuning a model, especially at a frontier scale, it's a brute force effort that is very demanding in time and money, uh, and GPUs. Uh, and the fact that we're starting to see companies, and I think we talked about it on this podcast, uh, a little bit ago with AWS and what they were doing with their, um, Azure Forge capability to, to customize models. We're starting to see even the frontier model makers, uh, give tools to the enterprise to take control of that representation of their own company in a mu in a much friendlier, uh, more accessible manner than having to hire a bunch of data scientists.

You know, it's interesting, Brad, that that brings to mind an announcement that we heard, uh, last week or the week before about, um, well, it wasn't really much of an announcement. Essentially, um, apple and Google have jointly, uh, let it be known to Jim Cramer of all people that, uh, apple is gonna use Gemini to build their next generation Siri. Now it remains to be seen what exactly this means.

And, uh, certainly we'll be watching for that at WWDC this year. But, uh, now, almost two years ago at ww DC Apple, uh, previewed a very expansive vision of what they would do. It sounds like Apple then spent a year plus, uh, sort of spinning their wheels trying to develop their own model before finally throwing in the towel and deciding to use the pretty incredible Gemini from Google instead after, um, sounds like a bake off.

Um, given all of this, what does this say in terms of using in-house models, uh, versus Brad? Yeah, it's, it's funny, the timing is, is kind of interesting in that, um, as we were just discussing with AWS that, um, you know, frontier model makers are making it easier for companies to take ownership of a foundational model and fine tune it for use internally. And this partnership, uh, is very much, you know, an extension of that idea because yes, uh, apple has a longstanding relationship with OpenAI, which is what they announced, you know, with what Stephen, you, you, uh, mentioned with their early, we're gonna remake Siri idea.

And, um, you know, that was basically a handoff to open ai, uh, to, to, you know, sort of obfuscates the identity or identifiable information about the user and then pass that on to, to, to open ai. And with what we know about Google Gemini, and its open source rel near relatives, I guess you call them in the Gemma family of models, uh, that you can literally just take the, what's great about the Gemini family and distill it down to a smaller model that still maintains a lot of the capability of the larger model, and then fine tune that thing to meet very specific use cases, not just running in the cloud, but running on the device itself. And we all know that Apple's been spending quite a bit on its system, on a chip design to, to be able to, to run AI effectively on, on their own platforms.

And if you look at their mlx, uh, compiler ex, what do you call it, sorry, in execution environment for their AI endpoints, it's pretty impressive what the, the size of model that it can run on some, you know, on a laptop, for example. So I think it's at a good time, it's a good timing for Apple to, to make this, uh, joint announcement, not a press release, not not anything of any, any note, uh, in our usual circles, because we call it an official leap. Yeah, right.

It's, it was called literally an, an official joint statement, my goodness. A, any anyway, you know, if, if, if Apple is serious about preserving user privacy, um, then this is a good partnership to execute on that, to go beyond what they had, you know, like, like you said, Steven tried internally and perhaps didn't succeed with, or ostensibly didn't succeed with. I think it's, yeah, it's interesting because they, they were, so, apple was so far ahead when they bought the original technology from SRI didn't, they Stanford, uh, research Institute, wait, I can't, I can't remember when it was the nineties or 2000 or something, I dunno, a long time ago.

Um, and then, you know, it just sort of, you know, sounded quite impressive when there were no alternatives and then suddenly when the world turned us didn't, um, to be completely blunt, I mean, I'm a I'm an Apple user, you know, love, love, love the kind of vertically integrated stack and everything and everything like that. Um, but it's, um, I don't use it, um, sir, because it wasn't, it was no particularly functional for me, but, so this should make quite a lot of difference. But as you say, the privacy aspect, that was, that was the pitch, wasn't it?

That was Apple's pitch. This is why we're different. And it's, I wonder, I think you're right from a technical point of view, this, this will enable 'em to maintain that.

I wonder how important that still is for users, um, given the amount of, you remember the kind of early days of, of, of generative AI when like, do not put things, do not upload things into that. Um, well, you know, guess what? I think everybody's doing a lot of that.

Um, so I, I, I do wonder how important that is. But, but, but also I think, or even, even if it's possible, right? Yeah.

It, yes. Is there not, did we see, last week there was some legislation put forward in the European Union to, uh, basically, um, scrape data before it's ever encrypted in apps like WhatsApp or, or Signal, alright. Yeah, yeah, yeah.

And I think the, the other that's in other places too, yeah. Mm-hmm. The sensitivity or the lack of publicity they want they were seeking has probably got more to do with the, um, the judgment against the Google, um, from last year, 20 24, 20 25, wasn't it?

Which as far as I'm, I read, you know, that bar's Google from entering maintaining exclusive agreements that last more than one year. So this presumably is not an exclusive agreement. So if Apple were to go back to AI or philanthropic and, and, you know, they could sign another deal.

Um, so which is, which is interesting. I mean, that's the, uh, the judge didn't say, you know, specifically this company with this company and, you know, and these kind of deal just said any, any, any exclusive deal that last more than one year. Um, so, so I think it's, um, you know, that's probably got something to do with, um, some of the, the lack of, uh, shouting this from the, uh, from the rooftop plus the money involved, which I know they didn't announce, but there's numbers out there.

Well, right. What is the money situation? Because we, we do know that Apple, um, pays Google, or sorry, Google pays Apple quite a bit of money, uh, to make their search engine the default engine in Safari.

Mm-hmm. And that was part, that was the reason for the, um, part of the reason for the antitrust, wasn't it? Exactly.

And then now rumor is Apple is gonna pay Google what, a billion a year, um, for this. So, um, I'm sure they're completely separate. Yes, I'm sure they're totally separate transaction, but the Gemini just shows reinforces the importance of Gemini, didn't it?

The little old, um, DeepMind folks in, uh, here, in here in London, um, you know, who are behind all that. You know, it's, uh, now actually, uh, just turning it into absolute real money, um, along with Google Cloud's performance last year, and, um, and, and what we think it'll be doing this year as well. It's, uh, it's a real business, these, uh, these AI models.

Yeah. And I think that's actually the most important thing. In fact, um, depending on what we see about the money changing hands here, we could be looking at the most lucrative AI deal in history, uh, in terms of actual productive, you know, to the point of this podcast, right?

You productive use of AI utilizing ai. Um, if Apple is indeed paying Google billion plus dollars to, to use Gemini, this would be one of the, one of the more lucrative or maybe the most lucrative investments. I do wonder as well, um, what this means in terms of the private cloud compute, which they announced back at, uh, ww DC in 2024 as well.

It was a great idea, which was to the point that both of you have made that Apple has developed quite a lot of silicon, uh, compute horsepower, and they are going to seamlessly extend that as a private enclave to a version of their apple silicon running in a private cloud and enable, um, basically cloud bursting of AI processing as needed. Um, do we want to assume that Gemini is actually running on Apple's private cloud compute, or do we wanna assume that Apple has just sort of waved their hands, and maybe this is running on Google's infrastructure, uh, because of course they've done tremendous things with their TPU hardware and so on. Um, I, I guess there's no hint yet about how that's running.

Is that true? I think it would foolish to conclude that, um, they're not going to avail themselves of every aspect and avenue that they can to effectively, meaning cost effectively serve these Gemini models to their, their somewhat sizable user base. Yeah.

Yeah. Because it could just be the model running on Apple, but Yeah. Yeah, and I mean, clearly I think that, I think we can assume that it's gonna run natively on Apple local devices, right?

Absolutely. I mean, it's not, not gonna be completely in the cloud, right? Yeah, no, that's what I'm saying is it's gonna be all of the above.

It's, it's depending on what you're doing, I would imagine you will see a graceful handoff of functionality, um, depending on what you're trying to do. If you're just trying to replicate what Google has with their circle to search capability on, on their phones, that's gonna run locally, why wouldn't it? Um, but maybe the search side of that transaction, not just the extracting the image, but the search of that image is something that you could very well and probably should hand off to a backend service somewhere.

Yeah, and I guess it also just shows from a kind of higher level, um, more abstract point of view in, in ai, the, you know, the, the importance of inference and, you know, inference is, is a, is a business, and this is inference obviously, because, you know, Google's gonna handle the, the training of the, the frontier models, and then Apple gonna build upon that. Um, but every time, you know, we're using it, that's off the inference, and that's what they're, um, essentially, um, yeah, not, not what they're paying for, for the actual, you know, what's, how that's how this value gonna realized for users, users. So we're doing a lot of, um, sort of guessing here about the specific deal with Apple and Google.

I wonder, uh, if you all can maybe take a step back, what does this mean for the market in 2026 and beyond for ai? Does it look like, uh, this is, uh, what the future is going to be for enterprises trying to deploy AI applications? Are they gonna put themselves in Apple's place and do a bake off and pick a model?

Um, to an extent, yeah. I mean, I always caution that, uh, in most cases, um, the model is not the application. And so, you know, you're not, um, you're, you're, they're not gonna be necessarily, um, talking directly to the model.

It'll be layers of abstraction. Um, but, uh, but yeah, you're gonna end up paying for, um, you're gonna end up paying for this. And this is where, um, you know, the kind of, you know, some of these other sort of trends we expect to, to, to see, um, you know, the new metrics of inference time compute, and that, and that, that kind of thing is gonna become really, really important.

Um, because obviously it depends what you're doing. Obviously, if you're trying to create videos, um, you know, the, the expense of that compared to, you know, responding in text to some sort of, you know, text prompt is, is just vastly different. So, so yeah, I think, I think, I don't think you'll see necessarily major enterprises directly, you know, saying we want to license Gemini.

I think, you know, th there's more that there'll be, they'll be buying into a stack, um, or at least Google hopes they'll be, um, and, um, and that will, that will be, uh, a key part of it. I mean, that's kind of the know the Gemini enterprise stuff is, is is like that. It's, um, obviously Google's gonna keep extremely tight control of, of the, of the models.

Um, and, and then, and that contrasts with the open source approach of, you know, there's lots of, there's, you know, thousands of open source models out there. So, so organizations do have a choice of how they go about this. Yeah.

And if they can, you know, basically build a, a sort of control plane for their applications that abstracts the exact model underneath, you know, to basically, uh, route the request to the most appropriate model being the model that's best able to answer the question or to carry out the task and to do so with the right, you know, requirement for latency, for privacy, uh, for security, for, um, concurrency, for instance. All of these little things that, that make up the decisions that drive technology investments around building software are, are very much at play, in play here. And that's why, you know, what we've, and we've said this on this podcast and, and elsewhere that, you know, models are less of just a chat response model and more of a platform that is built on a rich set of capabilities that are exposed through APIs and SDKs.

And so if I, as an enterprise builder, uh, am gonna use, uh, or if I'm, if I'm gonna run a task, I wanna know a, does this model do caching, for example, for just one example of many measures, you know, can it do prompt caching so that I don't have to keep saying the same prompt back and forth? Does it do speculative decoding to, to optimize, uh, what the transaction is happening inside the model itself? Does it have chain, uh, uh, what am I trying to say?

Uh, community of experts. So what is that called, sorry? Um, MOE model mixture of experts.

Yeah, sorry. So all of these, these things that, that aren't just a model, but are the surrounding, you know, uh, sort of infrastructure of the model are what are, what are gonna drive a lot of purchasing decisions in the enterprise? Yeah, it does seem as, as was the case with enterprise software as well, that the platform is more important than the underlying, um, you know, uh, infrastructure components.

And I think that that seems likely to continue. But speaking of infrastructure components, um, one more thing, uh, that came up that was interesting. There's been a lot of talk about, um, 2026 marking a transition from the sort of heavy GPU supercomputer data center model or finance model at least to more, uh, focus on inferencing and distributed compute and lighter weight.

And we've just been talking about that at little bit, but at the same time, they're still investments happening in data center power and energy. Um, what are the thoughts, uh, that you have about this sort of, um, one way or the other, uh, direction of the industry, Nick? Yeah, I think that will, um, that, yeah, the last thing you said there will continue in, in 2026.

I think the, the, the issue for the data center industry, um, in this year is obviously gonna be, yeah, the, the, the power and cooling issue and the, and the energy is the major, you know, energy is a major bottleneck. They understand this very well. The people, the organizations that build that acquire land and build data centers, but it's filtering up to the enterprise as, as this is, this is gonna be a challenge for, for, for organizations overall.

So as you kind of, you stuff more and more into a rack, the power cons, the power requirements for that rack goes, you know, from sort of 15 kilowatts to a hundred kilowatts to, to more and more and more. And, and obviously the idea is obviously you want to stuff, as, you know, cram as many of these into a building. Um, and then of course, what that results in is a lot of heat, um, but a lot, a lot of power to, to, to, you know, run it and then a lot of heat generated from it.

So, um, you know, I think this year we're gonna see, and we've seen a little bit of it in 2025, uh, data center build outs will get delayed due to the lack of, um, energy. Um, there's obviously an interesting arguments about the kinds of energy and, you know, the kind of in front of the meter, behind the meter, uh, renewables, non-renewable, the contrast between the US approach and the Chinese approach. But always already in the last few years, we know of many smaller, um, you know, data center operators that were kind of gonna, where the power is rather than where the customers are.

Um, and so yeah, that, that's been happening, but I think it was result on and, um, kind of some delays. And I also think, um, yeah, it makes liquid calling mandatory. And so for those, those organizations, for the data centers that would, would not, not having that they have to be retrofitted and for the organizations that that sell that equipment, um, you know, I would imagine it's gonna be, you know, it's gonna be a pretty good 20, 26, 27, 28, 29 onwards and onwards and onwards.

And obviously a lot of these are the big server companies, um, but also the, you know, the companies like Google that, that build their own data centers or build don't necessarily build themselves, but they build the equipment, um, that that that goes within them. Um, and Amazon, the same kind of thing. And Microsoft the same, same way.

So I think it's, it's this, um, you know, air cooling is sort of hitting a physics wall. Um, there's only so much air you can blow across a rack. Keep it cool.

Um, so I think, you know, this is, this is something we're gonna be looking at pretty closely in, in, in 2026. Yeah. Unless you put it in space, then, then it's pretty cool.

That's true. Um, clearly, but yeah, man, I, I, I feel like, um, it's not just about the power, but also about the actual infrastructure itself. And by that I mean the sand that makes up things like RAM and things like NVME drives, uh, in particular that are gonna drive a lot of the economics of the data center.

And, uh, I heard earlier last week from, uh, a vendor that specializes in, in storage and object storage in particular, say that they're already looking at a, a two a, a magnitude of two times the wait, you know, for getting, you know, that those, those basically NVME drives for their customers. So if you're waiting two years to, to get drives, what are you gonna do? You're, you're gonna optimize the drives you have.

And, uh, this vendor, uh, part of their go-to market is now going to be saying to their customers, you've already spent X amount of money on your hard drives in the data center. Let's do two things. First, let's use our management system that gives you, you know, an X increase, fold increase in the amount of data that you can actually store on this thing.

And second, let's offload some of the workflows that may normally have sat above that in something like a database, for example. Let's just push it down into the storage layer where it can be run more optimally. So you're using fewer watts, you require less cooling, and you are, you know, consolidating those workloads in a, in a very effective manner within that existing rack.

That's, that's pretty fascinating to me. It's not about let's stand up a new nuclear reactor in a new data center. It's maybe we should optimize what we have.

Yeah. Very pragmatic approach to the, uh, to the, to the problem, which I think, yeah, many of our, our listeners will, uh, understand and, and, uh, wanna do, well, certainly sand must flow. Yeah, we, we've gotta be pragmatic because, um, shortages of RAM and storage are, uh, just everywhere right now.

And so we've gotta be thinking about how we're gonna optimize the, the commodity we have. And of course, that could affect everything we've discussed here. So, um, this is, uh, getting a little bit long this week.

Thank you so much for, uh, weighing in here on these, uh, seemingly disconnected, but actually quite connected topics, uh, of what's happening in the industry from, uh, digital sovereignty to, uh, apple plus Google plus question mark to, uh, the, uh, push and pull between infrastructure and, um, raw materials and, uh, AI applications. Uh, before we go, uh, let me quickly check in with both of you on what your, uh, where your research is taking you this week and, um, what you're going to be doing. Um, Brad, what's new with you?

Well, I'm, I'm actually, um, looking at, uh, a new forecast that, uh, we're finalizing this week, so hopefully I'll have that online, um, within another 10 days or so. And, uh, like Nick has with his, um, survey he is working on, we're, you know, we're always, we're always active building out new research, uh, over here at rum. So I invite everybody that, that listens to this podcast to jump over to, to our site, the, the futurum group com, because you'll be able to gain access to a, a lot of this research.

We don't gate everything we do, we, we try to, you know, share our insights as broadly as we can. And for me, I guess I finished my, the survey. Um, it's, it's going to the field, um, it's in the field, um, by now by the time you listen to this.

And so we will have the results of that, uh, in a few weeks time. And then we'll be publishing, um, reports on it, we'll do, we'll do podcasts on it, we'll publish some of the data, and I'll be doing some presentations to clients, um, as, as to what's going on. And always with all these surveys, you're always trying to have a, a mixture of longitudinal questions, which kind of gives you, show you patterns over time and then trying to keep up to date, uh, with what, what's going on, uh, in the enterprise.

So that's, that's what I'll be working on. Excellent. And, um, as for me, um, we're gonna be hosting AI infrastructure Field day next week, um, tune in live, uh, Wednesday and Thursday and Friday for presentations on a lot of the infrastructure that we've just heard about.

Uh, very much looking forward to that one. And of course, we've got another AI Field day shaping up for, uh, in, uh, Q2 that is already just, uh, bursting at the seams with companies, uh, joining us for that. So keep an Eye on the Tech Field Day socials for that.

Thank you for listening to utilizing AI today. Uh, if you enjoyed this discussion, please subscribe on YouTube or your favorite podcast application and consider giving us a rating or a review. This podcast is brought to you by the analysts and experts at the Futurum Group, where Insight meets ai.

For show notes and more episodes, head over to Text Strong, do ai, the utilizing AI YouTube channel or the text TV app. Thanks for listening, and we'll catch you next week. More horsepower, more power.

Get those horses moving faster. Your ai, it's huge. It needs more power.

Or does it, can you actually do great things for your business with less horsepower and do things more efficiently? Join me on the Tech Field Day podcast as we drive this through the ideas of efficiency in AI infrastructure. Welcome to the Tech Field Day podcast, where we bring together a group of IT technical experts to discuss a single idea about key concepts in the industry.

This podcast features a variety of perspectives from members of the tech field they delegate, and it's often recorded in association with one of our events. In this case, AI infrastructure Field. Day four Tech Field Day is part of the Future Group, and this podcast is also published, uh, on our sister company site tech tv.

On this episode, we'll be discussing that AI needs resource efficiency and that more horsepower isn't the only direction. But before the discussion, let's meet who's on the panel today. Hi, I'm Pete Welcher.

Uh, I've been around networking for a long time now. Retired, still doing it because I think it's fun. Hey, I'm Gina Rosenthal.

I'm a product marketing manager. I've been doing that for a while, been doing product marketing for AI since about 2017. Yeah, and I'm Andy Banta.

I've been hanging around the infrastructure industry for quite a while, um, mostly from the technical aspect of it. And I'm Alistair Cook. I'm the event lead for AI infrastructure Field Day here at Tick Field Day.

And one of the things that we see is that AI needs to kind of grow up from this land grab and this idea that massive amounts of infrastructure, ridiculously powerful GPUs stacked up in, uh, rack after rack with hopefully a very strong flaw and liquid cooling to get to ridiculous numbers of, uh, kilowatts, hundreds of kilowatts of power being delivered through a single rack. Um, I've long felt that there needs to be a much more efficient way of delivering ai. I was hoping there'd be some revolution where this more horsepower would cease to be the issue because all of this horsepower and all of this compute power, all that energy has a cost.

It's really expensive to generate these things. It's really expensive to build these data centers. And as we've covered in the tech fields and news rundown, there are contracts for more data centers to be built to host AI than have ever existed in the past, exist now.

So this is a huge build out for ai, uh, using vast amounts of resources. And even if we set aside the difficulty of actually producing those, the cost to people producing them, is that even a sensible thing for us to be doing with the planet? Should we be generating AI by heating the entire planet up with these massive GPUs?

And it's kind of a a point of, of difficulty along here. Um, Andy, you've been around this, this block a few times before you've seen massive infrastructure get built out and maybe get more efficient. Can you see that happening here with ai?

Well, I, I think it has to. Uh, one of, I, I've actually been, um, talking about that power consumption used for data centers and AI data centers in general for the past couple years. But one of the things that I would be really interested in hearing from, from the AI infrastructure companies out there is how they're doing things more efficiently, not how they're doing things faster or more densely or, uh, or various different ways of adding more and more.

But it's, you know, tell us something about some algorithms that you're using to make more efficient use of the hardware that you have available or the speeds that you have available, or talk to us about a way, uh, a technique that you come up with to do more with less and, and instead of just telling us how you can do more, that's really what I'm interested in hearing more of from infrastructure companies. I completely agree, and I've got a slightly, uh, different perspective on the whole market. I think the companies that develop the very large models needed, um, vast funding because of the huge costs that go into those LLMs, and they couldn't afford to specialize, uh, because they really needed something that would have a vast potential market.

Unfortunately, what I think has happened is that they've assumed that's the future. I'm not sure it's certainly got a role particularly for conversational, um, interactions, I guess to put it. But if you are trying to model something, be it network troubleshooting or, um, uh, some biological process or something, maybe a smaller dedicated model that knows about, say, computer networks, um, network might be more suited to the task and a whole lot more efficient to run.

So that's kind of a question in the back of my head. I don't have an answer, I'm just kind of watching the space, uh, with intense interest. I agree with both of y'all.

And Andy, I agree with the whole idea around power. When you think about, if you read any type of description of what, uh, the data centers are being built for, it's described in gigawatts. So, um, it's not a back to the future reference, but it's actually how much power they need to run, um, these servers and to run the different components of the servers.

They never talk about the data. They never talk about how much data that will serve us. They never talk about how much data will be created.

They never talk about, you know, the actual part of ai, it's the end result of it. They only talk about what it takes to run it. But when you do an infrastructure, when you do any kind of architecture, you're always looking at what is the workload that's gonna be run on it.

And it's not at top speed because not everybody can afford to run everything at gigawatt speed all the time. Not only that, that, but if you think about what is ai, this is common, I'm gonna always come back to this. What is ai?

AI is what we used to call right now. What what is available from AI is what we used to call high performance computing. So machine learning and deep learning, and you don't need gigawatts of power to run those types of workloads.

So why aren't we talking about this as a true architecture instance, right? Like, what needs to happen, how much data you need, what do you need for each section of the pipeline that we're gonna be doing to transform this data? Um, that, those are kind of different questions, but I think those are important.

So we kind of skip over the whole, we can design for this, it's a workload, this is what we know how to do, and if it's a workload, we know how to make it run more efficiently. So I I'm with you. I I wanna hear more about that.

Well, you touched on something that I've been focusing on, which is, uh, where's the data? And I think we're gonna see that during the ai, uh, infrastructure field day, uh, presentations, because if the data's not local or if you've got, uh, such a huge LLM that you need to run your, um, training in multiple sites, then just accessing large amount of data is yet another barrier as well as latency and bandwidth and power. Right?

And I mean, the, the other piece of this that I, I really wanna try to address on some of our discussion here is that the, um, in addition to data, the ability to actually create the data isn't being done efficiently. Uh, and Fabrica presented at one of these sessions a long time ago before they got consumed, and part of their presentation was talking about the fact that the LAMA model is only, uh, theoretically capable of using about 70% of the GPUs available to it, and in practice is only using about 30% of the GPUs available to it. And that is clearly showing that more is done.

What you need here, what you need is to actually do the good old fashioned software engineering to make these models more efficient, to make use of resources that are available to them. And that is really part of the efficiency that I would like to hear people talk about. Uh, Gina made this point a little bit earlier, uh, when we were in the pre discussion for this about why aren't people looking at solutions similar to VMware of, uh, of sharing resources among various different processes.

Uh, and you know, I I think Phoenix can offer a little bit more information on what she was thinking, but, uh, that's like an excellent question that we don't hear from Ethan VMware these days. I Think VMware has a hard time getting the message out, but, um, they have a lot, they've had a lot of tools for a long time where you could do things like virtualize the GPU and split it into, I wanna say 16, been that vir that, uh, power to different virtual machines. So the same thing we did with Oracle back in the day, remember, nobody thought you could virtualize Oracle.

That, that it would be, you would cost so much because you needed those servers to be bare. You needed to, uh, uh, put that Oracle OS right onto the server, and that's how you had to run it. But once you started looking at, no, these are workloads, you can virtualize all of it, we can virtualize every component of it from a VMware perspective, and it's virtualization in general.

I'm sure it's all of the virtual, um, virtualization hypervisors that can do that for you. But, but it, we're not thinking about it that way. I think it's, it's, it's definitely the hype that's driving that.

Like you have to have ai, you get ai, you have to have this many servers and this many GPUs. That's the formula, the blocks, and that's kind of how they're being sold. I don't see anybody pushing back and saying, nah, man, we need, what we're needing to do is we're needing to take this historical data, do this with it, it's gonna cost this much for us to crunch through it, and this is what we actually need to install in our data center.

But people were doing this before the days of ai, they were doing it when you went to super compute, you would see people doing cool things on bare metal. And I know this because we were there with VMware saying, and you could also do this with virtualization. So it, it, it's just workloads.

And I think that the architects and the engineers need to get back to, to that piece, which would also force people back into the what is the use case? Why are we going to do ai? What's on this side?

What's on the other side, and what do we expect to report back? You know, what, what's this gonna gain the company if we, Yeah. And, uh, it's, um, and VMware certainly did do virtualization of GPUs, and they, they did it both.

They did it both in software, and they also were able to split out the hardware functions of various GPUs and spread those to various different VMs. And through some of their later technology, they were actually able to share those across the network. Um, I, one of the other concepts that has been brought up that I really haven't seen a whole lot of traction on is there are CXL vendors out there who are attempting to say, harvest the memory from your, um, systems that you were shutting down and put it into a c XL rack that you can actually use on a new system.

The idea is that DRAM doesn't go bad. DRAM can last for a very long time, and even if it's not the newest, most, most fast DRAM that's out there, you can certainly make use it again. So, I mean, the, in addition to efficiency, we also don't hear too many people talking about reuse or sharing of resources.

We're gonna have to, right? Because what happens now when all of a sudden the prices of everything is going through the roof, you can't get rid of your memory. You can, you really have to hold onto it because if you don't hold onto it, you, you're not gonna buy any new unless you pay three times what you're already paid, what you paid for it yesterday.

So we're in this memory crunch because they can't make it fast enough. The same way you can't get ahold of GPUs because they can't make them fast enough, can't manufacture them. We're in a place where the chips are getting smaller, but getting, doubling in, in capacity size.

So they haven't really, um, perfected that, um, that, that in the, in, in the manufacturing plants to get those rolling and get those out to customers. So many of them are already pre-sold And they're pre-sold, so you can't even get them, you once you get ready to go deploy this. So you're gonna have to get creative and figure out how do you architect for this workload with what I got or what I can maybe find on eBay for a decent price.

Right. Well, Jane hit on another, um, thing that I think is key, which is shorthand would be ROI, but what's the company benefit from what people are doing? And that's kind of the challenge with everybody does ai, uh, which seems to be the, a meme in some co companies, and it really is much more helpful to start thinking about ROI.

But another factor in it is that maybe not everybody, just like not everybody is cut out to be a programmer. People have their job they wanna do, and they may have an idea that AI can help them, but right now, uh, how to prompt A LLM and, uh, get good results and sort of cycle that is, uh, something that's not for everybody. And we're starting to see some mechanisms coming, uh, gait and some of the other things that help with automating that sometimes even in massive ways, which could be a problem.

But, uh, when is this, when is AI sort of experimentation gonna be easier for, uh, Joe Schmo and marketing or whatever to use, uh, as opposed to somebody who's a computer science, uh, specialist And, and then Pete, is that the right thing to have happen? Because as we're talking about this as, as doing AI through resource efficiency, what are the tools that that abstract away the intelligence of the operator, right? That the, the takeaway, the specialization of being able to work with the AI tool, well, that puts more load on the ai.

That means that we need more horsepower in that AI to handle the fact that we're asking somebody who, who is not specifically skilled or oriented towards understanding the tool to just throw simple language stuff at this tool. This is how we're getting this idea of needing more and more horsepower. I think coming back to one of Gina's points was that having an AI tool that is specific to a task rather than a general purpose ai.

So building into your application instead of using, uh, a full 7 billion or 20 billion parameter model, that is a general purpose model. Having a model that's specifically set up to handle maybe, uh, analyzing the productivity of your, um, factory floor. Having an AI that is specifically trained for that function is gonna be more efficient than feeding the all of the information about your production floor into a general purpose model.

And that's where we need to look at that resource efficiency versus simply building more horsepower. There isn't a single solution here, there isn't a single answer to this. Sometimes in order to get value out of ai, we're gonna need a very specialized AI that does a very specialized task.

And I come back to some of the early discussions around this, that a lot of the time it was still the, uh, predictive ai, the old fashioned machine learning that, uh, was looking for trends and data and, and, and extrapolating those trends and data a lot of the time, that's what the production AI was. And strapping a chat bot, large language model on top of a huge amount of resources wasn't necessarily the right way to solve this. Another of the things that we saw across our AI field days, uh, as as they were maturing, was the idea that chat bots aren't the answer.

The, the AI needs to be invisible inside the application doing whatever it's doing. Uh, it needs to not be the, the primary interface. It needs to just be the thing that makes the thing better to quote Bobby Allen.

So AI should disappear and, and should be task specific. Well, that's where maybe the LLM gets used in a slightly different, uh, form of agentic, uh, use, where it recognizes what field a question, what type of question is being asked, or what type of inferencing is, is being requested. And then it calls on a dedicated model that, uh, for instance, like what Cisco's done with an in-house network trained, uh, model to actually answer the question.

So it calls an AI agent rather than a software code agent. That's classic program. Yeah.

Yeah. And I mean, that's, I think that's part of where Alistair was going. I, I think part of it, uh, the idea of the efficiency could be that you, you don't need to train something on the entire world of knowledge to, for a specific thing.

And I mean, if, if you go in with the idea that we only need to train something for a specific piece of knowledge, then you, you probably can do it more efficiently. And it, it's, uh, you know, you talked about Cisco, a, a friend of mine who worked at AMD said that one on one of their early models of their, uh, virtual assistant that was based on AI model, that he, the first thing he did was asked them, what is Epic? Which is the name of, uh, virtual A, couldn't answer it for them.

One of the things that, you know, you brought up, I, I think it's unfair to say that Joe Schmo in marketing, like m the for marketing, right? And game, what happens when, um, what happened with email, when email first came out, everybody couldn't use it. I mean, I had to support post drops who were astrophysicists that really had a hard time with email.

So, you know, because it was so new, it was something brand new that nobody knew how to do. Um, and they didn't, they lost, they didn't understand the rules. But now you can't imagine having an organization that doesn't have a proper email system running one way or the other.

And I think that's how it's gonna be with ai. AI is gonna be how we communicate with computers in the future, right? This is how, and now and going forward, and it's gonna get more and more, we'll, we'll figure out the latency problem, we'll figure out the power problem, but people dunno how to use it now because nobody knows how to use it now.

Like the few people that do are, I would say the very, very technical people don't understand how important those marketing people are in the organization and how they can pick the information out and distribute it back. Again, I'm defending marketing because I'm product marketers, so have to do it. But, but I, I think you know, that, that very technical people forget that that data that we're moving around and trying to transform things into, one of the way it gets transformed is by the, the knowledge and the point of view that the people that aren't technical that are gonna take a little longer to train them on the prompts and, and how to do things properly, they understand the way to pull the information out because they have the vocabulary, everything else that nobody else wants to learn about marketing.

If you're not a marketer, you're a technologist, you don't care, you don't wanna learn that, but the marketers will know the right words to pull that out and pull the information outta the data. And that's really what a lot of those LLM, um, workflows are all about, is to help pull information out data. So I think some of this is just, we are in such early stages and we were bombarded absolutely bombarded with hype about what this is and how you don't get on board, your company is gonna be sunk and all the rest of it, you know?

So I think it's really important to remember that all of us are suffering from hype overload and we're gonna get there because this is how we communicate with computers now. Yeah, but this goes back to the original point. Yes, I agree exactly what you're saying, but I, I think, uh, some, some what's being lost here is that, um, the efficiency that you actually would get back up this is, this is making things more inefficient rather than, uh, more efficient.

And I think one of the things that we, we lose sight of is actually paying attention to the genuine computer science necessary. And instead, we, we keep adding layers that are, uh, wasting cycles to get the same job done. A thousand percent agree.

Have you ever looked at stuff and it looks like front page like AI stuff looks like front page from Microsoft way back in the day? Yes, it's, yes. Oh, I agree.

I agree with you. So will, will it optimize when it looks like, um, MySpace see the front page? I think that'll be a little bit optimized.

But that's like when they're just, that's probably like when they're just to the point where it's too much, too much and then they're like, holy crap, we gotta pull this back and make this really work for people and we've gotta make it good. Now, One other challenge that I'm, I think is lurking here with the idea that everybody does AI is, uh, it takes me back probably several decades to, uh, ro it, uh, the prospect of people just learning spreadsheets and doing corporate financials on a spreadsheet and screwing up a formula with consequences. And that's actually good news from the AI perspective is we've been here before, so people learning and making mistakes, uh, hopefully you can fence 'em and be aware that that's a potential problem.

But, um, I think assisting people with frameworks that, uh, minimize the mistakes is probably a useful thing to think about In terms of resource efficiency. Um, I, I really don't expect that an unskilled user should be handed here. Here's an AI that is a general purpose ai, and you can do anything you want with it.

And we're not gonna tell you any limitations around that, right? That's, I don't think that's a good way of using ai. AI should be invisibly embedded into the thing you are already doing.

So as a, as a marketing person, Jenner is per perpetually writing documents, writing, um, presentations. And if you're going outside of that, using AI to generate some things and then come back into it, the workflows rock, let's just say there's just another layer being added that adds no value. The AI should, should actually be embedded, however frustrating.

It's to have the, uh, Google sheet say, would you like me to generate everything for you? Right? That's where it should be.

Uh, the AI should be a feature that makes the application you are using makes the tool you are using function better. It shouldn't be a separate place that you have to go. Um, some of the evidence that we're misusing the separate place is, is absolutely ask the ai what is epic when you can go to Google and say, what is epic?

And be taken to the documentation that describes exactly what Epic is people are using. Um, chat, GPT is a replacement for search and finding that chat GPT isn't necessarily up to date on the latest news. It certainly doesn't know, uh, about the ice storm that's coming to the eastern US uh, this week or about the landslips that happened close to me here in New Zealand over the last couple of days due to the weather.

Um, but I can find that information through Google search, right? Using AI and particularly chat interfaces for the wrong is gonna lead to poor results. And I think hiding them away, Don't you think that if you, I'll go back to my front page, you know, that's why front page was so awful because it had everything in one, one place.

So don't you think that we're gonna end up having more bloat problem and more resource problem if, um, if we put everything in the place? Well, data sovereignty is the other thing I'm sitting here thinking, because if you have, uh, people just asking chat GPT, uh, well, it's already happening that, uh, very private and sensitive information, it's getting exposed because it's sloshing out into the ai. And given some recent results where people have been able to reproduce entire books or major chunks of books by feeding, by iteratively prompting the LLM, uh, they could probably get at that secret data, uh, confidential data as well, which is not a good thing.

Returning to, to one of the things that Gina said that, so the sort of bloat to building AI into everything, I think you've got a really vital point how rebuilding, putting AI into every single application independently as we're seeing at the moment isn't efficient. Is it? It's the same thing again and again and again, uh, but we are seeing some movement towards some standards about how you inject ai, uh, or specific knowledge in AI into an application technologies like MCP, um, standardizing the way you attach to an ai.

So I don't think there's necessarily the need for that, that complete rebuild. Uh, and there's other standards like the agent to agent standards because as Pete was saying, a lot of the stuff is, is actually gonna be done by agents rather than directly by human interaction. So I think there are developments in that space of not having to rebuild, redeploy, uh, do again, exactly the same thing we've done in the past.

So I think there is hope for us that the AI delivered into a specific applications doesn't have to be a completely from scratch or really inefficient re-implementation of some other ai. Uh, of course that's probably what we're gonna see first because that's how you get to a minimum viable product. Uh, I think Pete mentioned that, that idea of, uh, we built huge large language models and, and so that's the way we view all of AI needs to be this huge amount of resource for these huge large language models.

Again, I I view that a little bit as a minimum viable product. I really hope as maturity turns comes along, we'll see more efficiency, more reuse of technology, and more use of tools that are specialized to a single function and that do that function more efficiently. The thing is, we've gotta stop the pace of going for more and larger, which was coming back to our original premise, more and larger isn't necessarily the right way to go.

Um, so yeah, I, I think I've rambled all over the place on, on a bunch of stuff we've talked about. Uh, and I, I really do hope we do see more efficient use more tasks, specific use of large language model AI and, um, some of that maturity of working out how we're getting business value rather than just the, um, arms race to acquire as much resources we can. Because if we don't, somebody else will acquire that resource and, and we'll lose, I think in the end, if all you do is acquire resource and never use it, you're gonna lose anyway.

You may as well, uh, look at efficiently using those resources that you acquire. Well, there's competition for the resources. There's also time to market.

So if you're using massive resort costly, uh, resources and training a model, you have to build this Mongo data center and everything just to get rolling, that's a long lead time. And so anything you can do to make it lighter weight, it affects, uh, improves your time to market, Right. I, and you know, I, I think one of the other places where we could potentially see some efficiency is, uh, being able to filter out AI generated content when we're actually building our models.

It's one of the things, it's, we, I mean, we, we end up with, uh, hallucination amplification when we do that. And it's, um, it also means that we're, we, we end up with AI that's seeded with a imaginary information to begin with. And that's, that was one of the terrifying things about Deep Seek was, was when they built deep seek, one of the things they did was use a lot of AI generated content to teach the ai.

And yeah, that does seem like fantasy land being used to build another fantasy land. And that doesn't translate well to being useful in the real world generally. I think one of the other struggles is right, you know, of course the marketer I use, I use, um, Microsoft.

And, um, one of the things that's gotten way better is their copilot, especially if your documents are on the platform, that's gotten way better. But you talk about, um, the BLO is there and the bloat is there because the system's underneath of Office Doc, you know, online office are not all the same. They've come from mismatches of what they have and put up on, you know, the different, different sites and we use them, but you can see it when you try to start to run this whole pipeline, you know, writing and producing and editing and adding things and that kind of thing, because you can't do it all in one place.

So I think that that's a problem too, is like your businesses are going into AI and they might be going out and buying whatever they can to be the most powerful and the fastest and blah, blah, blah, but they're reaching data that they're using data, their organizational data that's on all sorts of stuff, maybe even on tape. There's a lot, that whole piece they're saying. That's what I've heard is that whole piece of preparing the data to actually be used by a model and to be used to make some business battle that take value.

That takes a lot of time and a lot of doing. And you think about rehydrating things from tape and getting things into an object ready, if that's the way you built things, how, how that looks for everything to be consumable by the model. So it's, it's, we we're gonna have a lot more bloat before we can get to the trim down thing, even if we're trying to, so like, we've got this problem of not being able to get the components, this whole problem with getting the data ready, you know, and then making it, um, palatable for the business users to use.

So it's a, there's a lot to be done. Did you just say garbage in, garbage out? Yeah, that's what I said.

I think that that idea of bringing data together from different sources leads to a, a sort of disconnect where, as you say, it's not all the same format or naming of things. And this is the semantic layer that Brad Shiman from Futureum Research has been writing about recently. I'll be presenting a a little bit of his research at our infrastructure field day tomorrow.

And, uh, I think there's gonna be a continuing conversation at our infrastructure field there about how you build this infrastructure, how you make it valuable. And we're probably gonna come in with a lens of how you make it efficient. Andy, Pete, and join Gina, uh, here in, in, uh, Santa Clara with us.

And I have a, a collection more awesome delegates. Uh, so before we fill out hours and hours of this conversation, I think I'm gonna put a line under it here. And thank you all for joining us today at the Tech Field Day podcast.

So before we close out this podcast, where can people connect with each of you and maybe continue this conversation or maybe decide they wanna buy you a beer at your local, uh, Bar? I can be found on LinkedIn, um, my initials PJ Welcher, you'd find me, You can find me on LinkedIn too. And I am so excited that my podcast partner will be with me at Tech Field at AI Field Day, and you can find us at tech aunties com.

Okay. And you can find me on LinkedIn as well, Andy Banta, uh, as well as on Blue Sky Social, and I, uh, my, my occasional, um, blog posting on andy banta com. And of course you could find me, Alistair Cook on LinkedIn and all of your favorite social media, as well as at AI Infrastructure Field Day this week.

Looking forward to having some great, uh, conversations with people there. So thank you so much for listening to this episode of The Tech Field Day podcast. And of course, if you enjoyed this discussion, subscribe on what if your favorite podcast application is on YouTube so you don't miss a single episode.

Uh, give us a nice review as well on a, a high rating because you love coming and listening to the Tick Field Day podcast, particularly when we have such entertaining delegates joining us that will help other people to find the awesome contact. This podcast is brought to you by Tick Field Day, the home of it experts from across the enterprise and a part of the Future Group for upcoming events and more episodes, head to tick field day com slash podcast or us on tech tv. Thanks for listening and of course we'll see you next week on the podcast.

Thank you so much for having me. And also thank you for the forum, and I'm just so appreciative of the group here. And I, I love the session yesterday where folks were talking about their experiences and the one thing when I met Ira was actually on a delegation and it was like 20 people.

And the thing that I noted about him was that he is like the king of real talk. So he tells the truth, he puts it out there like shamelessly unabashed truth. And so I really appreciated that.

And so we were just kind of drawn together from that. But, um, no, I did not train myself completely or become the quantum expert or anything like that. I tell you, there are some amazing people that work in this space, whether they're cryptographers, this Paul is something, isn't it?

I don't know whether I should, I'm just gonna go back and forth, I'll just walk back and forth. But it, there are just so many bright minds and amazing people in this space. And I tell you, I, um, actually was drawn to this space by a client who worked for P PayPal.

He is a sister there and he wanted me to talk about emerging tech, and he knows that I'm always in the gray space, always in the space. That's not quite commercial yet, but we've gotta figure out how to secure it. And I, I, it forced me into really digging in really deep looking at not only IBM's technologies, which were, you know, um, huge in quantum tech.

IBM's been at this for a long time, but also looking at the industry completely, um, understanding how other people view the industry, whether that's MIT, whether that's AWS Google, whether the other, um, publicly held, um, dwa ti ion Q companies that are in this space, but more importantly really around our use case. So I wanna ask how many people know, very much would say they know very much about quantum comp as it relates to quantum computing. Okay.

And can you say what you do with Supply chain security and product security? So we have been up uplifting and, uh, doing crypto agility on products, so that Awesome. We can, Are you also the SBO lady?

Yes. Oh, this is so great. I I'm so excited to hear you're coming 'cause this is what we'll be talking about.

And you also said, Uh, yeah, so I'm the, uh, the C of Prince George's County. Oh, cool. We lived, we lived there for, for 13 years.

We lived in, um, I know, but we, yeah, yeah, yeah. So I, I Came from the intelligent community having worked at, you know, three letter eight, you know, FBI and NA. So, uh, has some background and exposure to, you know, all the quantum stuff back for me, so.

Okay. Awesome. And I think I saw your hand.

Yeah, So, um, one of the senior managers actually at Google, uh, cool Willow quantum ai, So great. Was there anybody else that wanted to add, I know a little bit about p qc, not Necessarily. Okay.

Speaking of PQC, how many people are involved in a PQC project at this point? That's post quantum cryptographic migration. 1, 2, 3.

Anybody else over here? Okay. Alright, so we'll go ahead and get started when we leave here today.

I'm hoping that everybody will just feel slightly more comfortable. Quantum's one of those spaces that it's just so overwhelming. I could boil the ocean today, but I only have, you know, what, 45 minutes.

So what I'm gonna do is we're gonna talk a bit, I'm gonna walk through these slides, but I'm gonna try to leave as much time for q and a afterward 'cause I'm sure there'll be lots of questions, but you'll get a feel for the industry. You'll get a feel for what people are doing from a project standpoint and what some of the issues are. Like, you don't want your board coming to you going, what are we doing here?

And you don't have like an answer that happens a lot. It's starting to happen even more. So I sit in a lot of different worlds, um, which is actually a good thing, but it also can be a little bit overwhelming.

I, I serve on the board for the Department of Defense that manages the third party supply chain, the CMMC program for the cyra. And that is a huge, um, that's, that's their attempt to really sort of reign in all of the supply chain in terms of a level of security around compartmented unclassified information. So you can't do a bid, you can't work on a project if you don't have the level of certification that's starting November 10th.

And it's hugely impactful because everybody's been attesting to it. I think. Um, Bob talked about third party security attestation and people have been attesting to having the right level of security for probably about since 2017.

And now it actually, you have to prove it. There are folks that are coming and they're going to verify. So that's one of my boards.

The other board that I serve on is the, um, CTA, the uh, consumer Technology Association. Those are the folks that run the CES show. And with that I'm sort of their cyber person on the board, but also I work with the legions of member companies who are in some cases, you know, sort of close to security and other cases, maybe not so much.

But these types of topics are important. We've invested in Quantum World Congress, I am on the investment committee there. And so I work with the team that brings like the 30 some countries together to talk about quantum tech.

And so, um, and we'll be doing more of that. The CES show is making its best attempt to sort of mainstream this 'cause it's been in the lab, it's been in research, it's been in academia quite a bit, but not really mainstream. And so also, um, I have company, it's called Quantum Crunch because it's crunch time and we work with boards c-suite insurance and we help you have the difficult conversations, but also we establish the plans that make people at least, um, get started on this project.

But not with, you know, spending, you know, tons of money, which is what has been happening a lot. People have been spending a lot of money and they've been finding that their projects have not gone in a place where they've achieved a lot to show for it. And so we, um, and we've done over 37 engagements, um, where we are working with customers who say the first call is always, oh gosh, my CEO called, he wants to, he wants a briefing on this.

He wants to know what we're doing. Um, I I, there's so many third parties, we don't know where to start. There's so much information out there.

We're not, we need to get up to speed. And so, and then also, um, the other thing to mention there that's really important is that the third party aspect of this, I can't underplay or overstate rather enough and it does get underplayed. And the thing about security mostly, mostly is that this is not a project to go alone.

There are a lot of other people, this is an enterprise risk management level project, okay? So everybody knows about quantum computing and how it's other worldliness is in terms of quantum mechanics and physics. It's gonna change the world as we know it.

All the use cases, you know, drug discovery, like you won't have to wake up in the middle of the night and hear on TV that the drug you've been taking for the last 20 years is being recalled and they want to, um, want you to be part of a class action suit. Drug discovery will be a lot easier with quantum computing. It's really not going to replace classical computing at all.

It's gonna be great on the, um, computation end of things, really doing computation that classical computing can't even approach. So you've got, you know, the, the use cases you hear about all the time are drug discovery, material sciences. You hear a lot about supply chain optimization.

And then in the financial realm you hear a lot about modeling and um, financial modeling, financial analysis and like Monte Carlo simulation and ways to manage market disruptions and predict the future. And so it has enormous promise, but for all of us lucky people, we get to deal with the, those are all the carrots. We get the spinach, we get the project that really takes the one nefarious use case.

And that is quantum's outsized computing power. Being able to break modern day encryption is what everybody here should be concerned about. And it's also not Y 2K.

And I will tell you why. So when I say a sufficiently capable quantum computer, I don't mean like commercially available quantum computer, I don't mean that, um, you know, it's uh, something that can't be done in the lab more or less. And that's important because, you know, we have cryptographers all over the world that have been working on crypto analysis to create the resistant algorithms that, that can manage the speed and capability of quantum computing.

And NIST has been the folks that have convened that and they have worked for eight years on this. They've worked for eight years, but they've also done multiple rounds. They started out with like 82 submissions, then they cut it down to 69, they broke those and they hit 2029, then 27 in the last four in 2022.

And they released those standards in August. And it was thought I was at the White House during the meeting and we were, you know, it was like, oh my god, this is happening. Like the whole world's gonna just focus on this now.

And it kind of still was a lot of, um, a lot of nothing, like not a lot of people doing a lot and a lot of convening around the world still. 'cause this is very much a global project and I give this a lot of credit. They do a lot of things, but this project right here is hard and people didn't really pick up on, you know, really moving forward with projects right away.

So This was one of the dilemmas. So as you talk to organizations, this was happening around the same time. It's just a little bit after chat GPT was released and everybody had their, um, code red meetings about what are we gonna do about this?

They have their AI councils where everybody's getting together and they're trying to figure out and manage that risk or manage at least what the fallout could potentially be. And you had legal involved, you had a lot of people making lots of, um, rules and policies around the organization. I was very involved in a lot of those meetings in the s and p um, 500 realm.

And it was kind of chaotic and it took this discussion like off the map. It really did because people were really concerned about this. There was a lot of already sort of, um, you know, shadow AI happening in the background.

So it just became less of a priority. But this is what everybody says. They're like one more thing.

And at that time, I think teammate had done a study with CISOs, I think they, I don't wanna say it was like 150 CISOs. And they kept saying that, you know, CISOs will never be responsible for at least the bias part of ai. And, and CISOs were avoiding these meetings too.

They were like, I'm not going to that council meeting. Like that's the CDO, the chief data officer. That's those folks, when they figure it out, they can come tell us 'cause we are really not shifting that far left, right?

But this idea of being a cost center is really a critical part of what we do. It's the idea that really the people who have to make these decisions have to understand the risk. And they can't conflate it with quantum tech.

Quantum tech is hardware software. This issue is just a use case of quantum. And so that's the confusion.

So these folks, as more investment comes into this area, they start to sort of conflate the idea and they worry about the timelines and they just go, oh, I can like not worry about this for 10 years. I don't really have to think about it. And so the accountable people, the third parties, which I don't wanna steal any of your thunder because No, no, that's fine.

It's a whole, that's a, that's just a whole thing, but it's also sort of like an excuse that people lean on too to not get started. And the problem with this whole thing is that it's just very, it's just complex when you're talking about baseline components and you're, you're at the software level and you get down to the CBO M level, the federal government sent out a, um, spreadsheet to, to all the agencies saying, Hey, you gotta do your inventory as if it was just a list of items, you know? And, uh, they were confused.

They were on instructions, they submitted it to OMB, and then they end up coming back with like kind of a hodgepodge of responses. And they had to keep going out there and they had to engage. Um, you know, uh, I think it was Mitre was probably one of the folks that went out and tried to salvage this whole thing.

But that's the complexity. There's, it's very hierarchical when you're talking about all the parameters and ev all the elements associated with actually having, you know, an inventory of your cryptographic assets. Ooh, this is, I'm getting dizzy.

It's really rocking here. You guys are really rocking. Y'all are all rocking.

You rocking next. So thank you. I'm trying to rock the, um, factorization.

So I just wanna say this a a thing about this because this is what makes the, the, um, migration difficult is that the, when you're talking about quantum resistant algorithms, they're not gonna be drop-in replacements. It's gonna always involve some testing. It's gonna involve extensive testing and everybody already has a whole number of exceptions and, and systems that can't be touched 'cause it can't be broken and they just sit out there forever.

And so this, when you look at this now, this is a slide, this is actually a slide that we used to talk about all the time because the key link, when you look at 2048 RSA, that is, it's just hard to imagine that anybody could break that. 7 billion CPU years to break that under quantum. The cost of factoring of just plain factorization eight hours or the timeframe like that.

If you can wrap your mind around that kind of power, that's what we're talking about. And what I find is that the, um, you know, when you look at Adversarially, like sort of who's doing what around this space, there are some of our adversaries are spending more money than us, at least from the federal standpoint. And there's a lot of expectation that from the federal standpoint, we would be like pushing out a lot of guidance around this and that, um, we'd have more of a handle on a timeframe.

The reality is that you can print, put 37 folks in a room and they will be experts in this space and they will all say something different. And so it creates, it, it allows people into a false sense of security, of thinking this is gonna be so far from now, I'm not even gonna think about it. So NIST released the first set of standards, which you guys are probably familiar with this 2 0 3, 2 0 4, 2 0 5.

Everybody's heard about this. Um, everybody's working on it. There is a lot of, there are a lot of hybrid schemas where you're using classical as well as quantum resistant algorithms, particularly with, um, the first one, um, crystals kyber, which is sort of your general purpose, um, algorithm.

And the other two are more for digital signatures. And so the industry's dilemma is that everybody has, um, you know, there's this sort of fragmentation but also there's this reliance on the outdated cryptography, what I was talking about before, the idea that you have these applications that can't be broken. Some people think of, well this is, I need to piggyback this project onto something else.

It either needs to be part of my zero trust initiative or I need to slip it in where we're doing ref factorization on um, you know, our major applications. But a lot of people will say, oh, I'm gonna go off and I'm gonna do this little mobile app. We're gonna sandbox, we're gonna work on this.

And the only problem with that is that when it all comes down, this is not going to be, that's not going to gonna help you a lot because it's those core apps, the really important apps that are the ones that should be focused on. I actually was around for Y 2K and whenever somebody calls this Q2 K, it just drives me crazy. It just drives me absolutely crazy.

It used to work at one of my first IT jobs was a company that, um, basically pulled together LPAR space for developers in order if anybody Yeah, yeah. Oh my, it is like cloud before there was cloud, right? So the idea of like changing a date field and people actually just having to like, uh, you know, from the standpoint of do some minor development goes was way simpler.

But also, I have to say I did that in 97. Now when I look at that, we had like years and years and we also had like major PSAs going, like everybody, the whole world knew this change was happening, but it was a date certain. And so people often ask, well you know, this could happen at any time.

Do we know where the adversaries are on this? No, we don't. We we absolutely don't.

But I think that that's because of the complexity of the project. That's one of the reasons to sort of kind of get started. The other thing that people are concerned with is this idea of duplicated effort.

So they wanna be aligned around the third parties that work with them that they can't really know because they're a SaaS application or they're the cloud or they're, you know, there's some black box element to it that just keeps them from really having to be concerned with them, but also the way they're sort of pointing them to like, Hey, look at my website. You know, like, yeah, we're talking about this. It's another form of what Bob was talking about.

The third party checklist. So the confusion, fragmentation, this is sort of, has been a state of the industry and I work a lot. My deep verticals are financial government.

I've worked with government as government, has been a client for 28 years. I've been in cyber for 23 years. I worked across 25, 26 countries, four continents and, um, both public private sector health.

And I would say the first movers in this are really like telecom. And they're also, um, financial services. I think they have some really good consortia.

And you have to find your vertical, you gotta find your people and you have to, you know, like spend some time there because you don't wanna redo things that people have already thought of or already worked on. Okay, so this chilled me to the core. So that person down there is Jay, Jay ga better and he's um, now officially the head of IBM research.

And he just took Dario Gill's place. Dario Gill has been a fixture for a very long time and he is now going to the Department of Education, not, not education, I'm sorry, there's no Department of Education. I'm in the energy, he's going to the Department of Energy.

God bless him. You know, so Jay is picking this up. So Jay was, you know, I I told you that I'm an investor and sponsor for Quantum World Congress.

I was sitting in the audience and I have to say, when you go to as an, even as a, when I was an I, IBM er, when I would take clients to Yorktown Heights to the Quantum Center, you just get chills. It's like going to Willy Wonka's factory. 'cause it's just so unbelievable.

And the people are like so like entrenched and so confident. And I'll tell you what, and since I've been working PQC since like 20 21, 20 22, and every deadline and every um, milestone that they put in front of themselves, they've more than met. They've actually, I would say there's a bit of a speed up that's been happening.

And so while NIST is saying 2030, he actually put this slide up and he said, you know, we're gonna have a fault tolerant, um, quantum computer in 2029. So wow. Big gasp from the audience, but something to consider.

And also they, they're transitioning from quantum on devices to quantum computers, but there's much more, like you hear in the press all the time, people talk about qubits, but there's other parameters that need to be measured in terms of progress. When you're looking across these kind companies, um, speed, scale and quality. And he goes through all these numbers and you're just like, oh my God, we're behind.

It's just, it just gave me a little bit of a headache when I was sitting there listening to him. But they are very credible is what I'm trying to say. And I don't say that as a I IBM er per se.

It's just that they have the, um, I've seen the ecosystems of, you know, pretty much everybody that's sort of in this space and the progress that they've been making. They had a very, very early start. So look at this list and look at a good 'cause it's your third parties.

So everybody on this list has signed up for a collaborative research, um, agreement or r and d agreement with the federal government, ACRA. And a lot of people came to this in the beginning, like kicking and screaming. 'cause you cannot have a patent out of any of the work that you produce.

And it's also like you're just sort of giving it over to the government. So at first it was hard to get people to be part of this and they were having trouble with their organizations, then it suddenly became fashionable to be on it from an optic standpoint. But then you still have the 80 20 rule.

You have the 80% of the people who are, the, the purpose of the Cradle was to work out this migration, to work out interoperability and performance. 'cause before I said that these, um, new algorithms are not drop in replacements, larger key length, um, larger blocks, uh, block cipher, block size, um, very dependent around performance and reliability. And so when you consider all of that, you need these folks.

So they've been in the toolbox, sandbox working it for I would say maybe about three or three or four years. And some of them have come away, walked away with beautiful relationships with each other and not much to show for it. But others have really done the hard work and have come up with, um, some great evidence of that and, and the ability for you to go out there and test the work that they're doing today.

So one of the ones that I wanna show you that I really have been impressed with is, um, flare. So anybody here experience, um, heart Bleed in 2014 if you live some Heart Bleed, that's kind of, I always like to use the heart bleed because it's very similar to sort of, if we were to find out that things were compromised, it would feel like that. And if you remember with Heart Bleed, that that was just a vulnerability, but it was also, um, you know, a situation where we really didn't know where all our assets were.

And that was the hard part was sort of like, we don't know where everything is. This is gonna affect, you know, between 40 and 50% of the internet and we gotta go out here and patch open SSL, you know, and it was like a previous version that had been out there since 2012. And then they discovered it in 2014.

It was just, it was mayhem and you drop everything. And we don't need to be in that situation this time because we actually have some warning and we could actually get started. And so this shows the, um, post quantum encrypted share of their H-T-T-P-S request traffic that they're seeing.

And they are using, um, quantum resistant algorithms out there today. And they have, I saw them, I spent some time with them in London, uh, in June. And they were saying that they actually are, oh, am I in trouble on time?

No. No. Okay.

Oh, I want to leave time for questions 'cause I want hear what you guys have to say. We do we need to know what they're saying? No.

No. Okay. Unless Probably a safety Briefing.

Safety briefing. Okay. Okay.

Admiral Roger. Then we had to stop to hear them. Please go.

Oh, anybody here play bingo? Okay. So these are, they're a good example.

Like I said, there's some folks that are working so earnestly and they will help you with your plan. No point in you doing work that they have done already. And so, Mikel Mosca is one of my partners, his company Evolution Q, they've worked extensively in the QKD arena, which is, uh, quantum key distribution, another form or another strategy around, um, protecting against, uh, quantum threats.

And he has this theorem that basically says that for all of the, the long held data that you may have, plus all of the time it would take to actually migrate it against the threat of the, um, timeline, kind of tells you where your organization is from a, um, procrastination standpoint. So we use that the augmented theorem quite a bit. When we're having these conversations with people at the board level, it's very hard for them to have anything other than a risk conversation that impacts finances.

If it's not about finances, just get the hell out. You know, like there's not, that, that really is the kind of the place where you have to come to. You have to really understand the core systems that, and you have to have some agreement.

You'd be surprised organizations that don't have data catalogs or have not prioritized systems or data, and that there's not agreement amongst the people at the very top about what systems are like, not only operationally important, but the ones that are like, Hey, if this thing goes down, we, you know, we, we have no revenue. We, we, we're not even, we, we can't produce anything. And so elevating this project to enterprise risk management, if you have a risk manager in your organization, they should be like all over it, but in a, um, educated way.

And that's what we do. We do help people to sort of sort through their own risk profile and sort through their own set of priorities. And, and I find that when we do our stakeholder groups, we do, um, exercises and we do simulations, but we also do these, um, stakeholder workshops.

Oh my goodness. The legal people know more than everybody about data flows, data mapping, like who's, you know, what's connected to what It's stunning sometimes, but they kind of live in that world. So that's the reasons for the urgency, I would say.

One to point out is that the cost of waiting, so there's a real limited pool of people that are PKI engineers or people who are actually do the testing that understand it today. I, I believe that there doesn't have to be a million buts in seats. I believe that there will be, um, repeatable processes that people will buy vertical start to understand and get, and they'll be able to do a lot on their own eventually.

But if you go to most of these organizations that are like sometimes really big and they even have, um, you know, real focus on this, you're lucky to find five people that are, that sit in this space that understand this or that even can like sort out an the RFP or the RFI or write up their set of requirements or work with, um, consulting the big four that do all of your, um, audits, like both the internal audit and the external audit. Oh boy, man, I have, I, you'll have to talk to me. I have some stories for you.

I'll just, I will, I'll leave you at that. I'm on camera so I can't like say what I would really wanna say, but it's, it's malpractice at a level that, you know, nobody should be learning on your back. 'cause everybody's learning.

And that's the thing. Like you have to, there has to be some acknowledgement that this is like unprecedented. Nothing like this has happened before.

So, so anyway, so one other little element that is very important here is that you have the RSA show every year and it is the premier place that everybody kind of goes to, to like sort of hear the updates and stuff and it's kind of strangely silent on this topic. So, and I would, I would just, I'll just stop there, but like, catch me on the break and I'll tell you. So this was 2023 and one of our folks was up there and it was like it caused a, uh, press firestorm after Adi Shair, who's the s and RSA sort of said, oh, this con computer thing, you don't have to worry about this.

This is 50 years from now. Like, I don't even know what we're talking about it. Which he's important.

He's the s in the RSA and Adi, you know, says that. And the media just goes out and, you know, it's very similar to what um, uh, Jensen Wong did at our, um, CTA. We had, uh, the CF show this year.

And he just offhandedly sort of said Quantum's like 20 years away and literally all the public companies, they're stocked up like 10%. And people were like, wait a minute. But he did kind of go on an apology tour and even has invested, I think in quantum invested in a, um, a quantum company.

But there's not an admission that this is sort of the folks' space or it isn't so to speak. And so people are very influential when they say these things. It has a chilling effect over a group of folks who already have so much responsibility and so much to worry about that it's easy to keep putting it off.

So go to 2024. So I'm up there on 2024 and I gotta say it was the most incredible experience. I have so much respect for the cryptographers.

I mean, these people are like super crazy brilliant. And to be on conference call with the RSNA, like, you know, Lynn Delman and uh, ADI Shair and Ron Ve and also other, the other people, um, Craig Gentry is like, was like mathematician of the year that, that year. And um, t ben, like they are, and they work a lot on, uh, fully homomorphic encryption and other things that can protect us from these kinds of threats.

But it was pheno, it was a phenomenal experience. But I say this to say that my primary role, I'm obviously not a cryptographer. I have great respect and reverence for them, but my primary role was to say, when Adi Shamir says it's 50 years from now to like, you know, impress upon everybody that it's sooner than we think.

And I think I accomplished that NIST was applauding. They were like, oh my goodness, you said it. So let's go to the next one.

In 2025, I sat in the audience next to Ron Reve and he's the r and the RSA and we were talking about this and I said, well, what do you think's gonna happen this year, you think i's gonna say that Quantum's 50 years away? And Ron is, um, you know, an an amazing, amazing person. He said, you know, I talked to him last night.

He goes, I don't think he's gonna say that. But then I guess like near the end of it he says, well, you know, we, I don't know what's going been going on with Quantum. We've had no submissions to anybody.

Nobody's talking about quantum. And the fact of the matter is that I actually had a number of companies that I work for. 'cause the other thing that I, I work with companies who are early stage in the discovery, uh, crypto discovery realm and a lot of other smaller quantum companies.

And they had actually, I know for a fact that at least I knew about 16 submissions that just didn't make it in. And so is this a nefarious plotter plan? I no, I don't think so.

But I think that when you've had encryption that has been long held, and I mean, God bless these guys, this stuff is held since the seventies. I mean, it's pretty amazing and they're pretty amazing. And you know, it's hard to wrap your mind around the, the numbers that I showed.

You know, when you compare classical and quantum and also the use cases, there aren't a lot of use cases because the use cases are very private. People pay like 20 million to have these quantum instances and, and to do, um, you know, the level of innovation and work on them and with that investment, their top secret. So I know that was always a challenge that IBM we really couldn't talk a lot about use cases or where things were in terms of actual, you know, go to market.

So he, so yes, so we're, I guess all these companies are gonna all try again and try to be on the agenda for 2026. We'll see what happens. Okay.

So the indu industry is a bit of an echo chamber. If you go to any of these conferences, it's like we're all talking to ourselves. We're not talking to security operations people.

It's very in the weeds in terms of, you know, really like pure cryptography. It's all a lot of a math discussion and we know we've gotta change that. And we're trying to do more streamlining of that.

But also I would say that, um, the, there's a little bit of, of, um, geopolitical concern in terms of the way that organ organizations and consortia are really sort of experiencing and or exhibiting some signs of sovereignty where they're like, well, we're gonna pick up the ball. We're gonna go here and we're gonna pick up the ball. We're gonna go there.

And what is, what we need as an industry is to coalesce and we need more harmonization so that everybody's not running. 'cause we all gotta interoperate. Right.

Debbie, on that one, where is the create a roadmap for a company? Ciso, CIO. It's a, it's coming up.

Okay, so I was just getting ready to start on that one. And that's a great, great question. Thanks for asking that because there are things that you can do that's the, 'cause this is the thing we can like boil the ocean on the conundrum of what's not right.

You know, we are admiring the problem, but yeah, what, what can people do? So this, this idea of, you know, where to look for cryptographic assets, it's everywhere. And that's what makes it complex.

I think it was the CISO at at and t described it as looking for, um, I think he described it as looking for all of the, um, nails in his house at that were at right angles or something. Like it was just, I was like, yeah, it's probably quite a bit like that. Also common misunderstandings, all of the language and the terminology, like people use a lot of different terminology.

I like to say quantum resistant. I don't necessarily think anything is security is safe. So I don't, I never liked the quantum safe, um, terminology, but it's used a lot.

I would also say that, you know, people need to look at this as an encryption upgrade when they, when you're in cyber and you hear anything about quantum think in terms of, oh, this is an encryption upgrade and they're gonna do the same thing they did to me with ai. It's gonna be all on my, all on me to get it done. But the reality is that there is a lot of cross-function when you start getting into these basic components and elements that you cannot get it done without the developers.

You can't get it done without the system owners. And you mostly, mostly, mostly can't get it done without stake holders and ownership at the top of the organization that's giving everybody the same KPI, everybody's gonna give it the same attention. Everybody's gonna show their progress if you don't have the buy-in at the top.

It's a, I I've seen it in government. Let's give it to, uh, you know, level 14 person and he's gonna run around trying to elicit, you know, responses from everybody and not have the sponsorship that's needed. 'cause it's that it's that critical, it's that important.

So project maturity. So you'll see people saying all the time, this is what the project looks like and I just want to really reiterate the preparation part. Like, if you don't know enough about your data, you'll spend a lot of money.

You'll just spend a lot of money and you'll spend a lot of time. And so I was, I had a, um, workshop where I actually did have, um, legal counsel and communications and some other people in with the soc people and encryption services. Um, the incident response team there was, there was a whole, you know, cross section of folks.

And in that meeting it was amazing to get them. And this was like a two day workshop actually in the, in the workshop. It, it was interesting to how the encryption guys were on their heels because it's, they're very siloed.

Like they were, you know, they do, they, they give notifications, some of them on, um, you know, ca expirations kind of where they live or the thumbs up or thumbs down on new applications or they, you know, they're, they're not, nobody is looking at this from an enter at an enterprise level and broadly it's very siloed. But what was really amazing was how, uh, in this one organization there was not any recognition that there was actually, um, any sort of data catalog. Like people were finding out in real time that what, oh good, I don't have to go try to prioritize what's important first.

And so these types of convenings are really important, but without the board and the C-suite like saying this is important and without them really understanding it, like it's not just for cyber people. They gotta get the other people. I think I have, I'm gonna run outta time.

So this is a, um, sample, I won't say the company, but it's a sample of some of the reporting. People say, go out there and get your scan done. You know, get your, discover your inventory and all you'll discover is that everything's read because everything's read, you know, and, and everything is a lot more, um, the scope is huge.

And what happens is that people get that first set of information on discovery and they put their head down. They don't take that to a board or their boss or anybody. It looks like everybody's failing.

And that's not the case. Everybody's in the same boat, but it's also not in context. And I work with a few companies who are working to solve that problem, the contextual piece so that when you do have discovery, you're not looking at, you know, um, trying to decipher and match and do so much of this manually in terms of finding all your assets.

So there's a guy, Dr. Mark Te, who has a company and it's called, um, Cusack, I think it is called. Anyway, he talks about this in our global trust call.

And he did such a good job of it that I wanna, I want to, I know he's on YouTube and so I'll make sure that we get links to people to sort of look at what he has to say. 'cause he also addresses this idea of AI and quantum convergence, which is a whole other topic. It would take hours to talk about that.

And it's, there's some important things to be said about that 'cause and also supercomputing. So some people think that they'll converge and super computing will outrun quantum and it, it's just, it, there's a lot that keeps people in the pocket holding next. And then when you look at the tool categories, a lot of people are considering that tools will save the day.

I'm actually a big proponent for tools. Like we can, we're never gonna get this done if we're all just looking at and doing this with, you know, consultant help you. Really, the tools are coming, but they're a patchwork quilt of tools and they are coming together in a way that's going to be great when it actually happens because we need this more than anything.

But these are the different categories. And some people are building extensions like with EDR, they're doing that quite a bit. And there's been a few tools that I have recommended to folks that were as good as the, you know, they were state of the art for where they were, but this is this space watch this space.

It's constantly, um, advancing. Next. And then this, I love this slide.

So we just did a big, uh, briefing for 137 banks that are like Russell 3000 banks about what they need to do. And one of the big complaints, like 80% of their members say, nobody's paying attention to us. We've gone to the vendors.

They don't tell us what their PQC plans are. And we take a picture of this. 'cause you can go to their sites and get, um, their most recent plans or where they collect the data about how things are advancing.

AWS has a lot of, um, uh, information that developers can use today. People can make a plan that says, Hey, we, we worry about this legacy piece, but everything that we work on from this day forward, you're going to at least test out the new algorithms on, you know, like a policy like that at least has you with a plan for going forward. And so this is what we believe.

So we believe in, you know, the priority being like your core asset, like your really big thing. Spend some time on that. So the financial and privacy piece boards care about that.

They really do. Like, they don't wanna be caught flatfooted on this. The visibility thing.

We believe it's a hygiene situation that should be, you know, it should be like the, the way we move going forward. It just never has been historically that we have enterprise visibility tools are coming for that. Um, we also believe that as you're doing this testing, you have to, if you've developed a sidecar for testing, you've gotta be able to have the capability to roll back.

So hybrid schemes are what people are working on today. And then we, um, also the third party thing, just do it before the crypto inventories. And then, um, this last here, one right here, establishing cross-function KPIs is like the way, like it can't be you alone.

You have to have help. And I've run outta time think that's the last one. That's the last one.

Thanks everybody. Appreciate your attention. Hopefully, hopefully I haven't made you more apprehensive.

Just know that this, you're not alone or behind. Everybody's in kind of in the same boat. Thanks so much.