Techstrong TV September 3, 2025

I don't want no effing data center in my backyard. You are watching Textron Gang. Hey everyone.

Happy Wednesday. Well, you know, when you get those long three day holidays, it takes a Tuesday to get back into things. And by today, hopefully you're back on your normal, even keel as we head into post Labor Day.

You know, it's, for me, it was always schools back in town. Summer's over. Let's get busy and let's get busy here on Textron Gang.

Today. We've got a great show and some great gang members. Let me introduce you to them.

We have Kate Scarsella. Kate welcome. I haven't seen this gentleman in a while, my friend, guy, courier guy.

Good to see you. Chris Blas, of course, joining us as always, and fresh off his Mexican vacation. They let him back in.

Mitch Ashley. Hey, Mitch, how are you? Senior York.

Of course, we're joined by the Dean of all Yankee fans, Mike Ard and of editor editors too. But hey, Mike, A great high. It's be a tough September, man.

I'm looking at the schedule and I'm like going, oh boy, It's a tough September schedule, but they're playing well. Mm-hmm. We'll see.

We, we will talk offline about it more, but, uh, Ben Rice, Go. Blue Jays. Yep.

All right. So Mike, let, let's jump into things today. You know, it, it, it seems like we're building data centers, or we wanna build data centers, or we're pledging to build data centers, or we're hallucinating about data centers, but we're building data centers the size of Manhattan and just, you know, encompassing thermonuclear, uh, devices and hydroelectric plants and everything else.

So what's the deal? Well, let's see how that all turns out, because the folks at Data Center Watch are reporting that, well, local groups of which there is something now of 142 of them have managed to get, uh, $18 billion worth of these projects canceled. And 46 billion are further delayed.

So they might be more cancellations coming down the pike, but it sure looks like to me, the local rebellion against the AI is starting to form. So, are we gonna see more of this, Alan, and where are we gonna put these data centers? So, look, this is just such a classic, classic NIMBY case, right?

We all want these data centers, all these billionaires and sovereign funds and politicians falling over themselves with checks for hundreds of billions of dollars, at least pledges of hundreds of billions of dollars. And every, every city wants to be the new silicon data center capital of the world, whether it's west, south, southwest Texas, or, or, or Phoenix or Ohio, or, you know, the, the, the great heartland. And every, everybody wants these data centers until they're in your backyard.

And that's such a typical thing. Yeah, no, no, no. It's a great thing.

Great thing. I want all those jobs. We want all that money.

It's going to ruin my view. I don't want it. There's gonna be a nuclear, uh, power plant running it.

Forget about it. What can we do? And they go through the usual playbook.

Have we done the environmental impact? Have we, you know, looked at all the other things? Ha have we, you know, do we have lead in our pipes?

I mean, they'll go, you know, anyway, and, and, and they're valid. I, I get it. Who wants an ugly data center right outside, you know, some Monmouth data center right outside your house.

Right. You know that you don't want that as your view. My fear is, though, that under the current administration, you know, no one cares about the spotted owls and the little darter fish or whatever they're called.

They're just gonna drill baby Jill. And, and it may very well be the same thing here. Mm-hmm.

Chris. Yep. I, i, there may be some of that too, but, you know, the, you know, look, I've been playing with critical infrastructure for 35 or so years, right?

And this is a whole world of told you so, coming to roost, you know, 'cause mostly what I I see is, you know, the, the, the legitimate pushback, you know, and yeah, I agree. Environmentally, we need to account for the cost. And in critical infrastructure of what everybody thinks about is the grid and the power grid.

The North American power grid is an amazing beast. It is a single creature of, you know, titanic proportions. It's been engineered over the last a hundred years into what it is and what it isn't is capable of, of supporting the kind of systems, the power usage profiles that are being proposed for a lot of these data centers.

We talked about this last week with, uh, Dan O'Brien, I think. But, uh, you know, they're using evaporative cooling, which is wonderful and cheap, as long as water is cheap and you can throw it away and turn it into mist, you know, to cool down your data center, which historically it's been for decades and decades where we've seen these things. You know, they're universities and, you know, office parks.

Um, not this time. You actually have to account for all the costs. You know?

And this is where, you know, ideologically all of the parties that I belong to, come to home to roos, because this is about writing it down, owning it. And if I'm on a data center and I do, then I gotta figure out what the actual cost is, you know, not what the actual dollar is given. Whatever the current, you know, economic climate is, the whole thing's, the supply chain soup to nuts.

And these power sy you know, you can't just pop up power grids. These take decades and you can't on, let me give you a bit of hope at the end, that with the kind dollars we're talking about, yes, you can, you know, we all agree, you know, that, uh, well, I agree. And I know on this, on this show, we tend to agree that small nuke, you know, kind of things we're talking about, that's lovely.

Uh, battery is lovely. You know, there's lots of ways to do things. But at the heart of this story, I think is evaporative cooling unrealistic, you know, uh, applications of what's supposed to be cheap technology that just happens to cost the water we all drink, right?

That's not good math. That's not good capitalism. Well, I, you know, let me just say this.

Is, these data centers are, are on a scale unbe before, you know, nothing we've seen before. And frankly, all the water in the world isn't gonna cool. 'em, we need, we need next gen cooling solutions, right?

When we talk about liquid, cool. They're not talking about water. They're, you know, they're sort of more like antifreeze kind of things, but Well, it that needs more power.

And I'm sorry, I didn't really make my point. 'cause when we agreed last week, if I, I, I'm pretty sure it was that. Anyways, we, we all need, we need more power, right?

You need green. You want, We Need all we need of power. We Yes.

Which needs power, which is more power than just using water. Well, that's the, so it needs even more Power. The, that's the position I'm gonna take on this, which is, this is the thing that will get the power grid upgraded, right?

We all talk about our aging infrastructure. You can't, you can't live on the infrastructure that we have now. It will force it to be upgraded.

And so in that way, I'm all for it. I, I totally get not in my backyard and get off my lawn kinda thing. Um, you know, I don't wanna be living next to a data center.

Uh, but it's, it's a fact of life that's gonna happen. And we're, we're driving down that road, so let's do it. Right?

Let's really build out the infrastructure to support this, not just for today, but for tomorrow. And it's not the Worst. I'm not infrastructure that, uh, this is really a story.

I, I think, uh, you, you, you started down that path and one, which is, uh, we've seen this before. I mean, there's no context here. This is, so the, the, the publisher of this research is a firm called Market Data, uh, sorry, data Center Watch, which is owned by a, um, an AI consulting company and, and legit one.

And I'm not saying they're bending numbers or anything like that at all, but where's the context for this? Uh, I think that there's an agenda here to, which is pretty explicit when it comes to data center Watch to expose Nimbyism in data center built outs Yeah. And activism.

Yeah. You know, Gar, now that you mention it, I don't know if just Mike Ard didn't pull this up strictly to have us discussing it on the gang today. See, As opposed, He assigned the story.

So Mike, the posting the other day, a story, I, I just, it's why is it not, why is this not a story? This is far older than Nimbyism. This is like the oldest technological story ever.

And it's certainly, uh, the oldest and original one in it, which is, you wanna do all this fun software stuff, you have to do hardware build outs, including resources and all that stuff. And sometimes we're gonna get to an example later where the hardware is way ahead. In this case, the hardware is way behind.

And, uh, uh, you know, there's no supply. There's not, there's insufficient supply for all the demand to build data centers, insufficient supply for appropriate real estate in, partly in part because a lot of these, they, they want to put data centers where there's already, you know, uh, a good network connection. So they can't just put 'em up in the Arctic like they, they w would do otherwise.

There's also power issues. There's all this stuff, right? So, like Mitch said, it's gonna get built out.

I don't think we have, we're missing two contexts here. One is what are we capable of building in terms of power, uh, uh, delivery, uh, globally? We're just sitting here going, wow, we just don't have enough.

It's a disaster. Ah, this guy is falling. Or we're saying, shut up and get outta the way of innovation.

And, and, you know, let's steamroll over all the problems because of all the benefits that are gonna come. Um, so that, that's, you know, one side of how it's not a story. This is just supply and demand in, in action.

I would disagree on the sense that the level of scale here is phenomenal, and more communities are seeing this and experience in this, in the form of water issues. And they're starting to see electricity costs go up. And as Tip O'Neill once said, all politics is local.

And when you start messing around with the local stuff, you'd be surprised how much it will come around and kick you in the head. And, and frankly, yeah, but Just compare it to, sorry Kate, I'll let you That's okay. Definitely let you go.

Just compare it to building roads takes freaking forever. Nobody wants to bulldoze neighborhoods. They don't want a big fat road right next to them, but they have a congestion problem.

So where's the difference? I just don't see it. God, you, Your Bronx upbringing is shining through.

To your point, though, in here in New York, we were talking about rethinking about how the Cross Bronx Expressway works because of Robert Moses bulldozed entire neighborhoods in the fifties and sixties. Exactly. They determined that the entire destroy the Bronx, right?

So that, that whole thing might become a tunnel. Who knows? And, and the only thing that, you know, talking about it not being a story for this non-story, it's actually appearing in a lot of different, um, you know, podcasts and, and other, uh, frankly, other, other news shows where you have, you know, citizens, you know, gathering together and just saying, you know, no, to these data centers.

I continue to believe that we have not seen, you know, I, I can't help but remember over and over again, walking through these huge, you know, main, where all the mainframes were, and then all of a sudden there was like one mainframe, and, and you saw the, the, you know, the, the footprint of where all these other, you know, uh, mainframes were, I worry that we're rushing to build really, really large data centers without yet fully understanding this capability of where we're going. So that's one major concern. I think looking at the cost more than just the, the financial cost, is it really important, especially with resources being where they are today.

We don't have unlimited resources. We all know that. So we do have to rethink how we do things.

And, and I don't think that that's a, a bad thing. And people getting upset about, you know, not in my backyard is legitimate. And we have to think about, you know, or how do we make something more aesthetically suitable?

You know, do we put like a nice park around it or through it or, you know, sounds, Incorporate sounds like something outta the New York City zoning commission. You could build the biggest building in the world, just put a couple of benches outside in some trees. com, right?

And I remember on using that, people complaining that dot coms are gonna destroy everything. And I think I sort of said what guy said, and it's, it turned out to be true, right? 001% is more than we're we're gonna get with academic budgets and so forth.

Right? You know, and, and Mitch, what you said about the, about the, uh, uh, as an old school power grid person, if you want to get tens of billions of dollars into your grid, suddenly this is the way all those projects he thought were a good idea now, is to get approval and lay the, you know, copper Everywhere. But let, let, let's, let's just put this into the, there's a couple of different buckets y'all are hitting here.

You like, when I say y'all, Y'all, I do very much so Appropriate in this context conversation. Another Long Island accent. There's a couple of buckets here.

Number one is the issue of the power, regardless of where we build the data center, we gotta bring the power to it, or we build the power there in the case of, let's say a, you know, container sized nuclear facility or something. So you got power, you got the cooling. Kate, you bring up a point, and I'm not sure it got through to the audience, and, and the rest of us, which is we're building tomorrow's data centers for today's and yesterday's technology.

Who knows what tomorrow's technology requirements are going to be? Are we going to need a data center the size of Manhattan to do the kind of workload that we're anticipating gets done? It may be not the case, but anyway, so there's that aspect, there's the power aspect, and then there's the NIMBY aspect, right?

I remember Why don't forget the connectivity too. Well, I, I, yes. And, and connectivity fiber or however we're connecting, that's, that Is probably the most limiting factor because that's what's causing the 600 data centers in Virginia or whatever, Right?

Be because it's the latency. And I think it's also why we're seeing places that we wouldn't think of as data centers, data center central vie for this business, right? And it's gonna take all those things coming together for it.

And if those people are willing to put up you, you know, I, I live in an area where if you development is near big transmission lines that your, your property prices go down. Could you imagine? You know, who's gonna wanna live near these things?

Anyway, we're about outta time for this one. I, I would, I would just point out one last thing. There is a, you know, a a, an election cycle coming up in November.

It'll be interesting to see how many local officials get turned out over data center issues. We'll keep an eye on that. You're watching Textron Gang Discover Textron Group, the epicenter of tech innovation.

We are your go-to for reaching IT leaders and practitioners worldwide. Our secret impactful content that sparks awareness, engagement, and top quality leads with us. You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more.

Join our satisfied clients. Let's revolutionize your tech journey. Contact us today and tell your story to the world in the most powerful way with Textron Group.

Hey folks, we're back in. Something happened on the way to AI Nirvana, and it may be just the simple fact that a lot of this stuff is harder than anybody imagined. Uh, there's a new report out from OpenText talking about how 57% of folks now think that AI is a crucial project for them.

That's down from other studies that had it up in, around the 90%. And more interesting to me is at least 57% also said that it's just hard and they're not ready. And they got a lot of work to do and there's a lot of data management issues.

But Chris, did we kinda underestimate what's required here to succeed? Of course we did. And I, you know, for those of you who are watching this in Clipse, you know, Alan set this up perfectly at the end of, uh, the, the last segment because we're building the data centers, you know, on the whole data center issue.

We're building the data centers today for today and yesterday, but the infrastructure is tomorrow. And this is the same issue, right? You know, it's, it's, it's, you know, I'm, I'm talking to a monolithic ai, I don't know what I'm doing with it.

You know, it's, it's using 8 million joules of power, you know, to, you know, organize my calendar. And yes, it's, but, but the marks of a successful, uh, wave, you know, it's not just a blip is no matter how bad the, the downsides are, you know, we end up, you know, doing the kinds of things we were talking about in the last segment. We work it out.

So it's a mess. And I like this show, 'cause it gives me a sort of weekly reminder of, of where I've been on this path. And as we all know, like, you know, as of June this year, you know, the world has changed.

You know, it's changing month to month. The data center plans, I think they're in intrinsically wrong. I said that on this show.

You know, I think we're building the wrong sort of hardware 'cause we're thinking about it wrong. You know? And, and we're not even looking in the future a couple of years, you can run an AI on your laptop at home, right?

On, on a small, not, not at, you know, it's not, but it's a remarkable what it'll do 3, 4, 5, 10 years from now. This isn't all about data centers and big ai the way we're thinking about it. So there's so many ands to this, you know, answer to yes to your question that it's kind of hard to find a single place to stop.

Mm-hmm. You know, Mitch, I seem to, you know, I think it was, it's been a couple years now that we had some sort of, uh, hackathon that we were all running and you were at. And I remember looking at that and how amazed I was at how all these developers and people that I knew were really smart were like wrestling with the fundamentals of data management.

And I was like, it seems like, um, maybe we'll have a renaissance in data management discipline and practices. But it feels like, you know, a lot of people just don't have the grasp of data engineering and what's required. Well, the, uh, you know, data is the new oil, right?

But oil is sticky and messy. It's not always fun to work with. And, and that's the challenge is, is we don't really, uh, curate our data, you know, uh, comprehensively.

So we can just dump it into a rag or into some kind of a database that, that, uh, AI can use. I think the other part of it, Mike, is, and then how are we gonna use it? How are we gonna protect it?

How is AI going to process it? Uh, what kind of applications are we using? You know, I, I go back to, uh, you know, the, the idea that that AI is and will change everything.

Well, that's a formula for, well, it's gonna be hard. It's not gonna be easy. 'cause when things change, you change how you work, you change how you think, how you solve problems.

And while it may automate things and let you do more things, do, do, uh, more work for the output for the same amount of work that doesn't come for free. So we're in this, we're in this learning period, and it isn't the first time we've rushed headlong into the cloud or into blockchain, or into fill in your favorite technology that, that maybe worked out in the end, but wasn't, didn't, didn't deliver on every promise in the first six months of our, uh, our wild appetites to adopt it. So data is something that I think is a, a discipline.

Now I see job postings for more data scientists, more data curators, data analysts, people need that skill. So maybe we're doing a little less code gener writing, so we need more data people. And in the article it says, um, Barry says that, um, con context is king.

Mm-hmm. And it just makes me think with, with data as something I've always spoken about, like, we have data swamps, and just like those swamps in Florida and the south man, if you get in them, they'll kill you, right? And so the, the data is, is one big data swamp that's, you know, gonna kill us if we don't, you know, really start to, to address, you know, context and, and you know, I'm, I'm all for, you know, this whole deep bloating, right?

I, I, how, how many times do I have to say we, I think we just have too much data and in so doing, we, we are not able to manage it. And I don't care how much, you know, we put towards it. I, I think we have to, to look at data differently instead of doing the same old thing that's just not working.

You know, there's, there's so much of this, you know, in the last segment, I, I got to pull out the, uh, you know, old OT person saying, you know, told you so, right? You don't put the money into the grid, you know, you're gonna get an some ants someday. Now we have, ans it'll cost you more Kate, exactly what you just said.

You know, this is data science and information technology people for the last 50, 70 years saying, you know, what we should be doing. And, and the practicalities of business are, we don't. So we get swamps of data.

And I, as I look, as we look at this, you know, given the tools, we have given AI as it is and these semantic systems that, you know, give us, you know, a lot of speed and like, Mitch, to your point, they speed up a lot. You now, you might have more work to do, but you're doing more work. Um, and, and what you, what we have to do is the same sort of thing that we do internally in our heads or as cultures, is not save everything.

Lemme give you an example. You know, we right now could wire up and we are wire UPIs with perfect recall of everything. And it turns out that doesn't work.

Just like the same reason. It doesn't work in your head. It's not because we're not good enough.

It's because infinite information is nothing. It has no value at all. So curating our data sets and putting them in context where you can realistically actually use them as opposed to just storing them is, you know, a told you so from every data scientist, I suppose the last 50 years, saying, that's what you get for filing things this way.

Yeah. Boy, Chris, I Go ahead guy. Sorry.

There, there's, There's always a lot of talk in with new technologies, especially disruptive ones that clearly this is like the most disruptive technology we've ever experienced. There's talk about complexity, um, and I I'm hearing a lot that, no, no one's used the word, so I don't wanna put the word in anyone's mouth, but that, um, taking advantage of AI in all its forms, whether development or a or just incorporation into services or apps or whatever is, is, you know, there's a lot to think about. But I think actually there's a difference between complexity and difficulty.

Um, something could be complex, uh, hard to imagine complex and easy, but something that's really simple to imagine can be really hard. Like, for example, you know, lifting a car with your own hands, that's really hard, but a simple action. So what I'm getting at is, um, and I keep coming back to the way AI is a simulation and not actual intelligence.

And because it's a simulation and its designed point is to simulate thinking, it's really hard to understand how to use it and develop it. We keep coming up against this with, I think the actual experience of developing and using it, having a disconnect from what the expectations were going in. I kind of felt like being all kind, uh, like you be, you know, a slammer and go and say, oh, do it, do it.

So I'm really surprised everybody sucks at developing ai. Um, but I think that, that it's following the same kind of learning curve as everything else does. It's just because it looks like it's so promising.

Um, there's more frustration there would be, I mean, what is it, three years now? And we have this, this huge proportion of enterprises trying to, trying to adopt and develop it. Maybe we just need a little more patience.

I, I think the point, my expectations is exactly right, right? Because our expectation is the sci-fi, you know, you know, thing where you walk up, you know, they find the ancient alien computer that knows everything, and it says, I have considered every thought in the universe, and we expect that. But that's not, even when you think about how thought works, that's not logical.

It's not possible for anyone, any, like, not, I mean any, I mean, billions of years from now, no one's ever gonna be able to do. That's not how thinking works. And we expect these ais to do things that would do the super human model that we have in our head, which doesn't really exist.

So it comes down to compression of information, not expansion, you know, getting it down to small enough bits so you can reference it like we do in our heads. And, you know, we've been modeling, you know, a a i, I won't repeat it. You you phrased it really well at the beginning of this, you know, these are not intelligence, but they're, they're a simulation up.

So we're literally putting together code that simulates the parts of our brains and thinking about how that works. You know, the compression that, that we use so we can access the infinite information of the, of the universe and walk around, you know, as agent creatures has a lot to say about what any AI we're gonna create is even capable of. They can't do more than we can.

We, that's not the way thinking works. I think Kate's spot on though. It is about the context.

And I'll give you an example of something I was playing with, um, last few days. I took the earnings call from a vendor and I got the transcript and I shoved it in the chat GPT, and I had it quote, craft a story. Then I listened to the actual earnings call myself.

And I was, you know, the facts were in the story, but the context was gone. There was no, um, sense of, uh, where this all kind of fit or where it was in the larger Mike economy and Mike, wait, wait, Mike, and then wait, wait. And then I kept playing with the prompts to kind of tailor it to what I needed it to be and adding more data and all this other stuff.

And after like an hour, I was like, I'm spending an hour crafting prompts for something I could have done in like 15 minutes. And that's kind of where I'm like, eh, this isn't quite where it needs to be. But is that the AI's fault or is it the transcript of the call's fault?

The transcript of the call was highly accurate. It was com all the, every word that was said in the actual meeting was in the transcript. It's just the trans, but the transcript doesn't capture tone.

It doesn't capture, um, any of The other. And so are you blaming AI for that? Well, I'm not blaming though.

I'm just saying as a practical use case, though, it was gonna still gonna be faster for me to just look at the transcript or listen to the call and then craft something out of it than it would for me to sit around and tweak all these prompts to get to something that still wouldn't be a hundred percent. I think what you're Bringing up, that's what I mean. Go ahead.

What you're bringing up is, is pro, I call it prompting or using AI in that way, just as it is with, for developers. It's a skill you've gotta develop. I mean, we've all been prompting with Google, right?

We figured out our little human algorithms of how to get, you know, how do I get the LinkedIn page the easiest for somebody's name and company? I have a way of pasting that in really easy, and I it will come up usually first. We, it's on a much greater scale.

There are different ways of prompting. There's like structuring things highly structured of JSON prompting. There's contract prompting and people are discovering different models, like different types of approaches.

But it's, it's around that context point that we brought up earlier is there's instructions and then there's what we know about solving the problem or, or what we need to solve the problem, and how do we encapsulate that in some way to give the AI better direction as well. Sorry, I don think this is a prompt engineering problem. I think, Mike, what you are really saying is AI doesn't pick up the inferences, the nuances, the humanity.

That's, that's part of it. It's Also things are said, It's a combination of all those things, right? It's partially what Mitch is saying, the prompts and the skills and, and everybody on this call is pretty proficient with logic and how to put together prompts.

It's, you know, but do that in the average employee workplace. And those people are not gonna be sitting around going, let me become a prompt expert. That's just not gonna happen.

And they're gonna look at this stuff and go, this is kind of broken, and they're not gonna drive it as hard as you think that they would. That that was my, that was, that was a, a much nicer way of, of my reaction to what Mitch said. Mitch is taking a, a developer's and engineers approach to it.

And that's 1% at best of, of the user population. Uh, it's just not a case of, um, it just in general of, uh, using something that, uh, does work for you. I don't know.

Like it does do work for you, but it doesn't do good work for you. It, it just can't. It just can't.

It's, and if here's the, here's the other, here's what I'm getting at. Mike. You have been doing this for years.

You're a journalist and editor and so forth. 1% of the population and other people who sit there and go, oh, well I can produce a report. I'm gonna start my YouTube, I'm gonna start my own substack, and I'm just gonna AI the crap out of this stuff 'cause I'm really interested in it.

And that could produce volume. And a lot of the readership is not even gonna know that. It's not, I think we call that AI slop these days, right?

But this is an example of the layers of the problems because just that one thing and, and Mike we're building, you know, a, a, a AI that can hear the tone right now, but, uh, imagine giving just a transcript of a meeting to an intern as opposed to actually have them listen. You know, when you take context and tone out, this is what you get. So it's not that LM models or the use case, it is just that we have such a layer of inefficiencies and learning curves and, and dumb ways to do things.

So I I just, yes, we have a layer of inefficiency. You know what it's called humanity. When you look at, and I I I love reading about Neanderthals and Deno de Deso and you know, how, how language and communication developed in humanity.

It's not all from written languages. There were, there were, there were sounds, certain sounds not words conveyed, certain things, if you said a certain word by, but also making a hand gesture, it changed the meaning of the word, right? This is how humans communicate.

It's not all words. It's not just words. And unless we're gonna make our AI savvy to that, you can't blame it for not being able to, to translate it right?

Until the point where we can have an ai, watch the video, listen to the words, and see the, and being able to make guesses at the inferences and the subtleties and the nuances. That's, that's the difference between communication and words. Mm-hmm.

I don't think anybody's blaming AI here. I think we're just coming to a conclusion that, you know, it, it has fundamental limitations and I don't think we're abandoning it, but I think you really do need to think long and hard about how much time are you gonna spend trying to prompt something out of it versus just doing it yourself. This is, that's the problem.

It has fundamental limitations that are really hard for almost anybody to see. All right, we're gonna end it on that one. We'll come back.

We've got more to talk about guys. Let's talk about a IPCs garbage. I don't know.

You're watching. Textron inc. com is the leading resource for news analysis and education on challenges facing the cybersecurity industry.

com covers all aspects of cybersecurity, including data security, DevSecOps, cloud security, application security, network security, security threats, and more. com has the largest selection of security content featuring breaking news, blog posts, podcasts, and more. com to learn more.

com. Home of Security Bloggers Network. Hey folks, we're back.

And yes, we're talking about a IPCs. The Futurum Group has a new report out talking about what the latest rate of adoption is, and this is their second report on this topic. And well, not surprisingly, it may not be going as quickly as we thought, but there are people buying at least the lower end models of these things.

Guy, tell us what's going on here with these things and where are we on this curve? Well, um, it terrific analysis by my colleague Lyer Blanchard, um, who's been covering, uh, AI and devices for, you know, since the beginning of this craze three years ago. Um, one, one key thing to look at is, uh, that according to the research, uh, something like 90% of it, uh, uh, leaders, um, or at least, you know, PC decision, um, I think that a IPCs are, are worthwhile and, uh, the, the basic knowledge workers platform in the future, if not anybody using a pc, um, why is that?

Well, there's a craze going on. People are using ai. Uh, just to be clear, like, I'm not gonna define the AI PC exactly here, but it's something that can do inference on the PC and on the device.

So that also points to the problem and the problem. I wanna pose the problem as a question to, um, well, to you, Mike, actually, it's equally for Mitch, uh, or anybody, which is, um, uh, how come developers can't gr client server 40 years in? Why is it that when I'm using a mobile app, which I, this is what, what I've been complaining about for 20 years at least, and, uh, uh, you know, I go into a tunnel or out of a service area, suddenly the app stops working even though I'm just doing something that, like maybe just writing an email or who knows what, um, they're better than they used to be.

But somehow or other, um, we still wind up thinking that there's just one place like the mainframe computer on this USS enterprise where everything happens. And in our case now it's the cloud. So what Blanchard points out is that, um, most AI is still developed and hosted and delivered from the cloud.

So since a I PCs can't carry fiber with them wherever you go, there's a connectivity question and issue because that fun, uh, neural net, uh, inference chip on the PC with its associated memory and hardware is just not being utilized. There's maybe a few dozen commercial off the shelf applications right now. I mean, some are really important and big, um, like Microsoft applications, Adobe applications, but there's, there's really not very many that utilize it and forget about, um, you know, the, where a lot of people and how a lot of people use apps right now, which is in SaaS, all of this processing is happening in big cloud data centers.

So that's where the problem is. They're being adopted at, at a lightning pace. I think 90% is pretty high, um, but there isn't the software there for it.

And frankly, I despair, um, to a great degree of that happening anytime soon because, you know, after Client server, which was basically a way to overcome the limitations of the PC-based server at the time. Ever since then, we've had the internet and we've had cloud and we've had all this stuff. And it just seems to be yet another way for developers to do the easy rather than the good, which is to have their cute little development environments and not worry about distributed systems.

Go. So, you know, guy, first of all, I want you reference these Star Trek Enterprise Mainframe. I don't know if it was Julia a mainframe, but it was Quantum.

But the day stream, if you remember the episode, the day stream computer, the, the DStream Institute, remember they installed a, uh, Mitch remember the, the PhD fella installed a, a, a beta of a new operating system and it kind of mm-hmm went outta control. It reminds me of these ai, I I, I forget what they, they, the, the computer had a name and, uh, but it was kind of like an AI kind of thing. But in any of it, I'll look it up and we'll put it in there.

But here's, look, let's cut to the chase. The A i PCs won't be valuable till there's a killer app for it. And right now there's not period.

The end, there's no killer app. What, what, what am I getting on my A IPC that I'm not getting on my other PCs Max or Phones? Yes.

Thank you for saying that. That's a much better way to put it, which is the user doesn't know where the inference is happening. They just know that it doesn't work and they're gonna bang away at the thing.

'cause it's not working. There's no, there's Nothing to it. So, Mitch, wait a minute.

Soft, Mitch, let's get Mitch in here for a minute. 'cause you know, what we're kind of saying is, you know, software developers, you know, Mitch's people are holding the whole show up. So what's up with that, my friend Here, we, here we go again.

The famous word, Donald, Ronald Reagan, and William Software developers worth the issue. I thought you were gonna Donald Trump there for a minute. Well, you know, hey, that too.

Um, it is actually one of the big issues is, you know, you just don't put hardware in front of it and always get a great, you know, bump in, in performance or get a great new capability. You've gotta have the software take advantage of it. And developers are learning how to build applications with AI both in the cloud.

But now how, now that we've got this, uh, capability on, on the device, the, the best example is Apple. The one who's the farthest behind, at least publicly, right? They probably have the, have had the most inferencing power on the devices that we use, that who knows what it's used for, but it's certainly not an elevated or accelerated their AI strategy.

It's easy to dump on Apple right now. Maybe they'll have some meaningful announcements down the road around ai, but you know, software still is still eating the world. Folks, That, that's a perfect example of what I'm talking about.

I'm gonna start using that instead of the subway I used to use. I go into the subway and my game doesn't work. That's an example.

How many, you still have Siri even on, you know, the latest Apple hardware? Siri tells you I can't answer that right now. Why?

Because every freaking time you ask it anything, it's gotta go ping some server somewhere instead of just being able to, to do something locally. It just doesn't know how to do that. Why doesn't know how to do that with no is the wrong word.

It was not programmed to do that. Why not Mitch, tell me, tell me why are developers not doing this? Is it just too complicated to think about two environments running an application, it's just client server.

Where's the problem? Well, networks are like the future. Um, they're everywhere, but there's just uneven.

That's your problem in the subway. Uh, I'm gonna, I'm gonna take the opportunity, you know, for three, uh, separate segments to say this is another I told you so thing, right? You know, so we're gonna give people PCs, home PCs with 8 trillion lines of code and settings even at the, you know, the tab level where we actually have access to settings that none of us can decipher at all.

Um, and then that's how we get answered and that's how we get here. But I will say there was an open source project, Lumina os this is the last Windows device in my office here. That's still because I, I was tempted to, to switch over.

Uh, but yeah, you have to start off an entire different stack and it operates differently. You know, a a a PC hardware is totally hobbyist right now. You gotta be totally into it.

We'll see. Things are moving fast maybe in a year. You know, people are running our, you know, open source take on it in a couple spots in the world and it seems to be going well, but it's, as we said in in every other, uh, segment today, there are a lot of layers to this.

It's not magic pixie dust. It will take a while to work through. But you know, as with power grids, as with, you know, how we've been storing data, um, you know, how we've been considering our personal device needs reconsidering.

If we really want it to be smart enough to do LLM tasks and things, we expect, you know, without some massive data center somewhere. But we can do it. One thing that the hype cycles and the AI adoption curves don't properly represent is in this adoption curve, there's a point beginning of that curve where we're using the technology, the advanced technology, whatever it is, cloud, IPCs, et cetera, AI to solve problems in the way we solved it before we had that capability, right?

We're using the kind of the same mental model of how we would apply it or the same methods, same software that we're building. And there's a point like in the cloud where we realize that, oh, I can do things a lot differently in terms of managing my infrastructure if I have, you know, three times the amount of servers available when I do my upgrades, I can do rolling upgrades, I can do AB testing and do lots of stuff that I weren't really possible before. Just as one simple example.

And you start to rethink and do things a a little bit sometimes greatly d differently. We're still in this solving the way we've been solving it with ai. We have a, that's why, you know, we struggle with prompting Mike.

That's why we struggle with why, why don't our apps have better AI in 'em yet? Why don't we have apps with AI or apps that are AI and why can't they connect? We're we're in the sub subway.

All of those, you know, we're in this early cycle. I'm not saying all those problems will be solved, you know, at once, but it's an evolution. So we're, we're wanting the benefit and we're still solving problems that we, that we, we got here.

So I have, I have question. So do you think I'm being too pessimistic? 'cause I just, I, I just, this, I feel like this particular problem is different.

That there's some, there's some cultural or, or mental disconnect. Not mental, but there's some disconnect. I where like I'm saying like it requires development in two environments.

That just seems to be one too many. Yeah, I, I agree actually with Guy. I, I, I totally agree with you here, guy.

I mean, we, we need to, we need to, I don't know why we continue to even think about that whole two-way relationship. I think it's just, you know, let's start developing for that, for that specific chip for, for this ai chip on PCs and everything else. And forget about legacy.

Legacy is crushing us all the way around. I I I know It's spoken like a real Mac person. I Need to throw it out and rebuild from scratch Fast among the things.

Yeah, let's get at AI and, and, and Mac and all that stuff. What about web assembly? Web Assembly is a way to do exactly this.

Go wassom to Put a package on a Yeah. To put a PA and, and, and I'm like huge web assembly enthusiasm. Probably because I like tilting at windmills.

I mean, maybe, I'm not sure which is harder writing in Wasm or building an AI software. 'cause they're both kind of difficult, you know. Well, I think we're gonna have to continue, but do touche.

Okay, Fine. I I do have a question though. I have a question for the group.

So who here thinks that they want to have it's September, you know, and believe it or not, the holidays are two months away. So who wants an A IPC for the holidays? Who's got that on their list?

Anybody? Or are you gonna wait till next year? I Already have one.

No, I want it. Ah, see, there you go. There's all Kate.

Kate just summarized the problem. We're all buying them because we're FOMO or something. I already have three of them not counting my phone.

Yeah, I ju I'm just a big talker. Yeah, More hardware Is always good. Now, now you guys can start to, you know, now you know what to get me for.

All right. Put it out on Kate's list, guys. We gotta we gotta wrap up here.

We're in 45 minutes and we've got a lot of text drunk TV to come today. Kate, guy, Chris, Mitch, Mike, thanks for joining. Thank you for watching.

Stay tuned for Techstrong tv. Um, if you're not watching this in our stream, uh, feel free to go wherever you want after this. But I would suggest either the YouTube channel for Text Drunk TV or the Text Drunk TV website.

Or if you have nothing to do, download the OTT app on Apple, Roku, uh, Amazon, Google Play. And you could watch more like this till your heart's content until tomorrow. I'm Alan Shimel, we're out.

Hey everyone. Welcome back here to Tech Drunk tv. You know, the thing about Black Hat, it's, it's really two days of the show, right?

Like, uh, uh, Wednesday and Thursday are the actual show. There's trainings the weekend before and that Monday and Tuesday and in two days it's just hard to fit everyone in. So unfortunately, I didn't get a chance to meet up with our next guest while we were out in Vegas.

Or maybe it was fortunately. 'cause this is a lot more relaxed. It's not like you're in the convection oven of 110 degrees out there unless you're in the air conditioning.

Um, but let me introduce you to Mark Lambert. He's the Chief Product Officer over at Armor Code. Hey Mark, welcome back to Tech Drunk tv.

It's great to see you. Great to see you too, Alan. Thanks for getting me on.

And yeah, it's, so Shane, we didn't get a help code in, uh, Vegas this year, but I know both of us were running around like crazy. Yeah, yeah. As we were talking off camera, it was, it was a lot of stuff.

Even my nights were filled up with stuff and I, I'll be honest with you, I try not to like do the late night things and, you know, go to a gazillion parties. But a lot of vendors do business over dinner out in Vegas now, right? They have all these kind of round tables and it's, it's not just, it used to be when I was younger, go out for drinks.

You actually have, there's good business being done. Good discussions, good people. So anyway, enough about Black Hat 'cause it's r a's early this year.

We gotta get ready for that. But Mark, let's talk about you, right? You're, as I mentioned, you're CPO over at Armor Code.

Let, before we get into Armor Code and we're gonna talk some Mag agentic ai 'cause what else is there to talk about? Um, let's talk about you, mark. How did you get to become CPO over here?

What's your journey been? Well, I, I mean I've, I've been in the application security space for over two decades now. Um, originally, uh, working for an organization where we were one of the early, uh, SaaS tool vendors.

Um, you know, we also had a broader portfolio of DAST and ISAs and API security as well. Um, and you know, the thing that I saw during that, that time is that that tool landscape is very diverse and there's just a lot of, you know, frankly, a lot of competition, a lot of time spent trying to figure out who's got the best scanning capabilities. And what I realized is it was just generating a whole bunch of, you know, I don't wanna say noise, but just like people being overwhelmed by, uh, findings, vulnerabilities and alerts and not being able to make sense of it all.

So that was actually where, around about the time that I met, uh, Nikhil Ktar, our, our CEO and founder here at ER code. And, uh, he'd just come outta stealth with Armor Code at that time. And the thing that we do here at ER code is we bring all of those signals together to filter out the noise so you can actually figure out what to do and how to do it.

Um, and that was kinda like what got me over here, um, and, uh, yeah, in, in the chief product officer role here. Brilliant. So how long have you been in Armor code then?

I'm knocking on for, well, well over three and a half years now. So yeah. G getting on towards my foot.

You Were, yeah, you were really early. 'cause I remember, I, you know, I remember when Nick Hill Neil launched the Armor code and the whole, you know, verbal books and, and everything that goes with it. And, you know, we followed along.

I was gonna say it had to be during COVID or thereabouts. Yeah, we were still in the throes of COVID, uh mm-hmm. Uh, you know, log four J had like, just come out.

Yeah, it was, all of those things were kind of like bringing kinda like that critical mass of, of things together. So, you know, we were going through cloud and digital transformation. We were seeing open source security attacks becoming real, uh, software supply chain, uh, chain attacks being real.

Um, and that's really driven the, the industry's focus on how do we kinda solve this problem of, you know what, like I don't, I, I need another scanner to find the thing, but how do I bring that into a workflow where I can actually consolidate the data from across all these different sources? Um, but yeah, it's been a, a phenomenal journey. We we're really blessed to have some great, um, organizations that we work with, um, name brands within the Fortune, uh, 500 and Global 1000, uh, really leveraging the platform at scale.

So we know we've got, um, over 320 different integrations now with different scanners of different sources. Everything from threat modeling through the AppSec stack cloud infrastructure. Um, then actually ingesting now over 200 million findings a day in the platform.

We've processed over 40 billion and we're processing, sorry, we're supporting over 4,300 security practitioners supporting, um, wow. An estimated 200,000 developers. So you, we've got phenomenal, um, you know, community and you referenced the Purple Book community as well, which is, um, you know, a, a non-A code, um, you know, o organization that we support from the point of view of orchestrating it.

But really it's, it's a group of security leaders meeting on a regular basis to talk about things like the impact of AI or the Cyber Resilience Act, which are the two things that we did as, as Purple Book community panels, um, at the, uh, at the blackout event. Absolutely. You know, mark, it strikes me as I'm sitting here listening to you, you and I have both been around the block.

We've been in security, as you mentioned, you're in absec two decades. I'm three decades into security, you know, and we think of, oh yeah, that was during COVID mm-hmm. And during COVID we saw lock four J and software supply chains and S bombs.

And, and that was kind of, you know, when I was a little kid, I used to watch happy days and you know, what the kid, what they wore then versus what I was wearing as a teenager then. And you know, we, we think of, okay, those were the COVID days, but of course now everything is ai, right? And it's like that stuff is almost back in the drawer so to speak, though.

It's still real problems that we real need real solutions to. But now we think in terms of, okay, well how can AI do that? And what is else is AI bringing to us and what new problems do we have maybe as a result of this?

And, you know, it seems like it's AI all the time. Um, and, and of course it's no different, right? Armor code recently announced significant advan advancements and it's agentic ai, I dunno if you called it tool or service or product on my notes.

It's called Champion. Yeah. Called Anya.

Tell us about Anya, mark. Yeah. So, so you, you touched upon the right thing there is like we, we go through these waves of yeah.

Detection, right? So we had open source as a wave wave of disruption, software composition analysis came in to help put guardrails around that. Uh, we then had Cloud CSPs and Synapse kind of helping there.

Now we have this explosion of ai. Um, and, and you really have to ask yourselves two questions is like, first of all, how can you know, uh, you know, an organization benefit from the use of AI for their internal processes, from the point of view of internal, of improving their developer productivity, for example. But then also how can the mechanisms that we leverage to defend and help, um, do things like prioritization like we do at code?

Um, how can we leverage AI there to basically scale and go faster? Um, so at ALMA Code, we've been introducing AI powered capabilities now for a little over actually getting off for two years now. Um, and um, you know, that started with Correlation, trying to identify kinda like patterns in the findings that we're ingesting to really understand what are the unique issues.

Uh, we then brought in a remediation LLM that started providing remediation guidance. We're leveraging AI for ingesting pen test results from PDF files. But that was kinda like the first wave of like, okay, how can we leverage AI to move things faster?

The thing that we're now seeing, which has driven like the next generation of functionality within normal code is as organizations are adopting AI themselves, the development teams are becoming a lot more productive. They're generating a lot more code. How can we build a workflow and a process that really enables the security teams to scale and keep up with that pace of innovation that the development team now have They got 40%, uh, 40% more productive?

Or if you read some things like Ford or five times more productive, you know, that just means there's it's downstream of, of work that needs to be handled from the point of view of the security teams. Um, now I will say that there's a, there's an ev uh, evolution of that work, which we'll talk about in a, in a minute, hopefully. But what we're doing in Armco to meet that challenge is we're introducing, um, a whole suite of Agen age agentic driven functionality within the platform.

That started with a launch of Anya at RSA this year. So Anya is a natural language interface. You're right, we call it our, your virtual security champion.

You can ask natural language questions, you can ask natural language, um, um, actions for it to take. And we're connecting it in to the ecosystem that already exists with Anya, uh, sorry, with Armika. So Anya being this native AI capability and, um, agent interface that sits on top of all of that data that we're already ingesting for problems that are real and still there.

But how can we supercharge that with an agent AI like Kanye? So Mark, guest development seems to be moving ahead, right? Security can't afford, security can't jump be the man who jumps out in front of the railroad tracks and says, stop, stop, stop.

'cause that train's just gonna run you right over. Right? We've learned that the hard way.

Um, so we, we need to, to, you know, kind of fight fire with fire or to be on par, right? Uh, it's kind of cold war. It's mutually assure destruction.

We gotta make sure we're doing what they're doing. But we're also seeing Mark, you know, a couple of studies have come out, some of these AI initiatives are not being successful, are not delivering what we thought they, what people thought they might deliver. And some of it may be because the expectations were so ridiculous, right?

And so overhyped, um, I'm wondering what, you know, you guys are obviously, 'cause Anya has to, I'm imagining closely track with what developers and DevOps teams and platform engineers and these folks are doing. How are you seeing, you know, separating the hype from the reality of it? What are you seeing?

Uh, that's a, that's a great question. And, you know, I think, you know, we're, as an industry, and I'm not just talking about security, I'm talking about just software in general. We're figuring out the ways to leverage AI to be the most effective.

And certainly, you know, we we're leveraging AI ourselves at Armor code. You know, we're using things like Claude Code Cursor, windsurf interfaces that enable our engineering teams to move faster. What ends up happening though, the, the work changes, it doesn't go away.

You know, people concerned about, Hey, AI taking my job, kind of thing. No, it's actually changing the way you work. And you are, you are making yourself more productive.

You know, a good example of that is one of the things that we, we see a lot of people focusing on today from the point of view of the risk of AI is, and it's like, Hey, you're gonna be generating more code, but that code is gonna have more vulnerabilities and therefore it's gonna overwhelm you downstream. And while that's kind of true to a certain extent, right now, that problem's gonna be relatively short-lived. The, the code generators are getting better.

The, the tools like the EM grip are getting embedded into the pipelines. You know, that feedback loop is tighter. So the code quality, I actually think ultimately the code will actually be generated, will be actually better than if a human had generated.

The problem that we're gonna start to see though, is this generates another problem. And that other problem is it now just takes you minutes to spin up an application and you can create almost disposable applications that potentially don't have the same level of governance in place that your traditional, um, uh, software development approach would be. So how can we evolve our security practices to not stand in front of the train tracks so that you get the visibility of what's going on, and so that you can actually focus your resources correctly?

And this is actually one of the things that we're very focused on within the category of application security, posture management, which is really the one thing that Amma code does. But certainly the primary thing that we, that we talk to from a market category perspective is as part of A SPM, looking at the code repositories, identifying those that are leveraging ai, identifying those, or, or just basically identifying applications that are deployed quicker. Because what's gonna happen is your volume of vulnerabilities from that generated code are gonna drop, but the number of applications are gonna increase, and those applications are going to be spun up and then potentially forgotten about, which means you've got a growing attack surface.

And you and I both know that vulnera, you know, security is not a point in time. I mean, you know, log four J taught us that, um, you know, you'll have an application that spins up on a certain stack. There's a vulnera, you know, leveraging a, a framework or a library.

And I need to track that. I need to be able to be able to audit it. I need to, in essence, capture its sbo om so that I can know when a vulnerability appears that I have an asset out there that's vulnerable, that's maybe not being scanned anymore because that code repository has been archived because the team are no longer, uh, working on that.

So with this sprawling attack surface, and having the controls and visibility in place is going to be the real problem that I think we're gonna start up, uh, seeing in the next 12 months once we've got over this ai a AI generated code concern. Yeah. Yeah.

Mark, it's only a 15 minute interview. We've used a lot of it. Let, let's talk, you know, practically speaking, now people say, this sounds good.

I'd like to check out what Anya is. We're all, you know, we're always looking for solutions. What's the best kind of on-ramp forum?

Certainly. So the, the best thing is the website as it is with, with almost everything these days. com, uh, you'll see the request, the demo button there.

Um, you know, we're more than, you know, more than happy to share that with you guys. Um, there are a whole bunch of other resources on the website as well, including little videos that'll give you kinda like quick tours, so to speak, of, of that capability as well. Um, and then, you know, we also run, uh, pilots and PSEs like, uh, you know, like everybody else does in the industry as well.

But we're gonna connect that to your real data. We're gonna show you what Anya is able to do on your real data within a fully controlled sandbox environment. Love it.

Excellent. Well, mark, look next big one on the list. I, I don't know if you'll be at reinvent in December if you guys are there.

I Won't personally be there, but we'll have a couple of folks there for sure. Well, RSAI am sure you'll be at, and it's earlier this year. I, I think RS a's in March.

I'm not sure. Yes. But let's make sure we don't miss that one.

Okay, that Sounds great. Look forward to it, Alan. Alrighty.

Mark Lambert, CE CPO at Armor Code, uh, talking about Anya, go check it out. com, mark, take care. We'll see you soon.

We're gonna take a break here on Techstrong tv. Hey guys, thanks for the throw. We're here with Barma Kuna Paju, who's the CEO of Ops ramp and arm of Hewlett Packard Enterprise.

And we're having a little chat about AIOps and observability and cloud native application environments because, well, it's complicated. Pharma, welcome to show. Thank you.

Thank you for having me. Observability has been tough, and I think in a lot of cases, all we ever really managed to achieve was some basic monitoring in the first place. But I mean, cloud native environments, it's even harder.

'cause there's all these microservices that get spun up, they get spun down. Nobody seems to know which way is going. And, um, by the time you take a look at it and start to sort it out, it makes your head hurt.

So as we kind of move into the age of AIOps, is this gonna get better? And how does AIOps kind of change the way we need to think about observing these environments? Oh, great.

Uh, great point. You know, first of all, the, the complexity of cloud native and AI native applications increasing, right? You know, more and more distributor microservices are playing a bigger role in making it to deliver business services, you know, to the lines of business.

So if you look at that and look at what the AI ops era, if you looked at original AI ops era that started pre transformer days where, you know, the LLMs and, and the transformers were not there, the original ops journey was all about machine learning algorithms, trying to kind of find, you know, those patterns and analyze those patterns using historical data to figure out what is a deviation from a normal behavior. And the second application of that, uh, AA ops and the pre transformer days are all about, you know, potentially looking for, you know, a large volume of alerts and, you know, ingest all the, all those alerts as signal, and then figure out which is signal and which is noise. And, and really try trying to kind of troubleshoot in terms of figuring out, extracting a signal from a bunch of, uh, noise, right?

So those are the pre early days of pre transformer ai, um, techniques that are used for AI ops. But with LLMs and with the transformers, the AI applications, and ai, the way that the AI is used in the observability is massively shifted, you know, not just at the large language models, but foundational models that are domain specific that can be fully fine tuned for being able to kind of do that probable root cause. So where I see the AI and Agent K AIOps in the, in the context of post transformer and post LLM and, you know, the current generation is all about how do you kind of really get the probable root cause or a, a real close to the root cause by using the, the foundational models and by using more natural language queries to kind of make sure the human operator is getting the assistance from context syn two and domain specific intelligence around the observability data.

Mm-hmm. Am I still having to instrument those microservices to collect that data, or am I just pulling the telemetry data in its kinda raw format and dumping it into something that the algorithms then makes sense of? Yeah, good question.

You know, you know, in the, with open telemetry, you know, the collection of the data is really making it, you know, much more seamless, you know, as opposed to in the old days where, you know, you have application specific instrumentation with agents deployed to, to really get that observability data to now, um, with Open Telemetry and eeb PF, the advances in EBPF really makes, you know, in certain cases, auto instrumentation to get a, a really good set of telemetry data that you could ingest for your models and your, uh, foundational models to absorb if you wanted to apply some customized foundational models for observability or even otherwise, you know, regular observability data, the instrumentation, um, both with open telemetry and the advancements in EBPF, um, you know, the instrumentation is becoming much, much more uniform and open. Yeah. Early on, AIOps makes use of a lot of machine learning algorithms, and it took a while for those machine learning algorithms to learn the environment.

And in the age of cloud native, the environment's more dynamic than ever. So how quickly do, do those ML algorithms first learn the environment, and then how do they keep track of all the changes? Yeah, so the original machine learning algorithms are all based on pattern recognition, right?

So you have to kind of really do instrumentation, instrumentation absorbed data to feed along with all the historical information to really recognize the patterns. Lot of those machine learning algorithms in the first generation are all based on, you know, learning from the past data, very little chance to do Jira shot, uh, being able to kind of, uh, for the first time encountering some things. But with the new, um, transformer based models, you have a very good shot at being able to kind of go after Jira shot, um, predictions and, and predictions that you have not determined that you have observed before because the models are trained and the models can really come back and, and, and, and, uh, significantly, um, pinpoint saying that this is potentially the, the, the reason why this is happening, even though that pattern did not account encounter before.

So that's the big difference that we see with, uh, the last two years of advances in ai. Mm-hmm. As I understand it with AI too, there's generative and there's predictive and causal algorithms, and am I gonna be using a mix of these things in an AIOps platform?

And which type do I use for what when? Yeah, no. Um, if you look at, um, the, the entire instrumentation and observability data, you know, I think there is room for playing a number of these, um, AI techniques, right?

You know, you are not replacing the machine learning completely with LLMs, and you know, you're not replacing full-blown transformers. So if you look at the, the, uh, advances in the last two years, you made the human interaction with the observability data a lot more natural language specific. You can do prompts and prompt based techniques to kind of interact where a virtual operator from, you know, um, agent or ai, um, you know, LLM driven AI models to gear that in human interface is more prompt driven.

That's number one. But it doesn't mean that, you know, you are completely replaced with the old ML technique. So it is complimentary to what took place in the past.

And, you know, leveraging those, uh, algorithms and moving the data to, in some cases, frequency domain and analyzing that data in the frequency domain to kind of freely understand those, those ultimately is the needle in the haystack to determine the po potential root cause still exists. So, to Sean, answer is, it's a combination, you know, Well, we still need traditional monitoring tools. We've had 'em for decades.

They kind of track a bunch of predefined metrics, but observability in my mind was always about, you can dive in and analyze stuff and tease out root cause issues. Uh, are these things converging or are they always gonna be somewhat, uh, orthogonal to each other? How do you see this evolving?

Yeah, I think, you know, most often the industry says, you know, monitoring is all about, you know, uh, finding a problem, right? By, by putting some thresholds and determining, and that's how the monitoring definition came in the early days, more and more, the observability is all about how do I reason, how do I find more cause that actual problem that because you're now providing more context, Jan, Jan, than saying that something failed because of the, the monitoring alert that showed. So to answer your question, in my mind, observability is a more super set, you know, and, you know, and observability came in the, in the cloud, cloud native stacks with logs, metrics and traces all coming together to give, but it also, you know, doesn't preclude for making the same observability extend to your network, extend that to the edge, where the actual consumer of those applications are really sitting and being able to kind of really find why an application performance or an outcome that the business user is expecting is not getting delivered.

So, in my view, this is a, a super encompassing thing. When you kind of really take cloud cloud native or a NA to stacks with Edge and the edge consumer, um, of those applications and, and the data that is needed for finding that root cause is complemented, then that's when you get the full, full view of the entire observability. So in some ways, monitoring is a subset of the overall observability that is happening today.

Early on, the folks who built these cloud native applications on things like Kubernetes were small groups tucked in the corner somewhere full of specialists, but it seems like in the last year or two or so, these apps are now going mainstream, and they are now part of the general IT landscape. Are we gonna see some unification of the way we manage legacy monolithic apps in these new cloud native apps in some way? Because otherwise we still got two separate teams just managing more stuff than ever.

Yeah. I think these silos and and existence of these silos happen over a period of time, because more and more cloud and AI native applications are becoming relevant for new stacks, but traditional enterprise applications still sits in the enterprise. Now, how do you bring these two silos together with, uh, call it digital operations command center, you know, for lack of any other words, where you bring the traditional applications, the edge infrastructures, and the observability around those consumer, uh, of those applications and those edge infrastructures and the observability data along with modern cloud, cloud native applications, uh, instrumentation, bringing that together is where, in my view, the actual it, and it is proactive as to make sure that the business users and the outcomes that they're expecting are really, you know, is getting delivered, is going to play.

And how those two things come together in my mind is, uh, uh, uh, um, a concept of digital operations. Our digital operations command center, where you're bringing all the entire, entire state of the IT ex across bottom of the infrastructure, all the way to the applications, and taking traditional and modern applications together into one single place. You know, AI ops in the original days was all about alert correlation to create that digital command center.

You know, the so-called AI ops tools were originally designed around, you know, Hey, throw me all the alerts from your traditional applications and modern applications, and we will process and give you a, a qualified incident that you can pass it to your IT teams. But the concept of DevOps, tech ops, IT, ops, SE ops, all of them coming together in the modern digital operations command center, calls for a new way of looking at that command center. And that's where observability is leading to now.

Mm-hmm. Do we need to rethink, therefore, the way the IT organization is structured? Because historically, we had all these silos, we had the networking people over here, the storage people, the servers, bunch of folks managing applications over there, and they kind of created their own little fiefdoms.

Um, do we need to kind of just make a concerted effort to start breaking those walls down? Yeah, no, I think, uh, um, you, you, you brought a very important point in the, in the context of pandemic and how cloud and cloud adoption accelerated those silos needs to be completely broken. Because at the end of the day, the business outcome and a business service impact is what it is expected to deliver to, you know, lines of business, almost like it acting like a service provider, right?

So if you wanted to make it act like a service provider, you know, throwing 20 people on a bridge to determine what happened is not going to kind of help. So one needs to kind of provide a, a qualified observability data, not just in the full stack observability at the application level, but all the way to the edge where the actual business user is consuming. You know, whether it is rum, whether it is application performance data, whether it is full stack observability data along with network that is going to play a major role in terms of consuming this application.

All of that needs to come together to really find, you know, a, a unified IT that acts like a single service provider to, to cure the business outcome, right? That's where I think the, the advances in AI advances in observability, the advances in machine learning, all ultimately resulting in bringing that probable root cause and making the outcome, uh, to the business application user is what needs to be broken. And that breaking down is happening as we speak.

So ultimately, what's your best advice to organizations that are rapidly embracing cloud native computing, but they definitely have legacy applications. Is there some smart way of getting started? I think a lot of them have thrown tech at the wall for many years now, but as we look at AIOps, sometimes it's intimidating and other people are fully embraced, but how should they think about getting started?

Yeah, I think the best way to start, in my view is, uh, an approach of integrate to consolidate as opposed to rip and replace, right? So you have this modern, new, more cloud, cloud native applications that you, IT organizations did full stack observability, let's say. And then they have the old, you know, the traditional applications which are happening in, in, in, in either their data centers are in the edge locations.

More and more edge locations are becoming like a mini data center. If you go to a store, you know, behind a large, a grocery store or a, you know, or any other store, there is a, a small mini data center that is running for what needs to be run at the edge. So those applications and those, uh, edge edge applications along with their cloud, cloud native applications, they need to be brought together with an integrated to consolidate approach where you bring this New Year's, um, agent ai, um, slash operations management command center, kind of a solutions, and integrate some of these traditional old instrumentation that one would've done along with the new observability stacks.

And then over a period of time modernize that entire stack. So that way you have one single, um, operations console for most IT organizations to be able to deliver business outcomes. All right, well, folks, you heard in here, cloud native is everywhere.

It's not just in the cloud, it's all the way at the edge. And while that can seem a little scary, it also creates an opportunity to rethink how we manage it altogether. Hey, Norma, thanks for being on the show.

Thank you. Thank you for having me. All right.

And back to you guys in the studio. Hey everyone. Alex Smith here and welcome to Textron tv, and I am delighted to be joined by DA Vu, managing director of VA Google Cloud Marketplace.

Di thank you for joining the show today. Yeah, great to be here with you, Alex. So, Dai, we at the Futurum Group did a research study with Google Cloud on the marketplace.

We spoke with a ton of your partners, um, earlier on in the year. And, and we'll get to that in a little, in a little bit. But just to kind of start off, give us a bit of a big picture, you know, for ISVs and channel partners that are new to this, you know, how do you see Google Cloud marketplace, you know, kind of changing the way in which companies ISVs go to market?

Yeah, absolutely. So at the core, uh, cloud marketplaces are fundamentally to change the way ISVs and channel partners go to market by shifting this traditional, like, direct sales motion to something that's more digital, online and scalable, and also collaborative across the ecosystem. And so, you know, think of it as, you know, the central hub where customers can search, discover, trial, procure, and deploy software.

And for partners it's a great opportunity 'cause it streamlines the sales process. It opens up, uh, new revenue streams and fosters deeper collaboration across the ecosystem. Yeah, and from our side, you know, our research shows that cloud marketplaces are becoming an increasingly major route to market.

Um, and in fact, a survey that we conducted at the start of the year, uh, showed that 97% of partners are saying that some of their revenue is tied to marketplace. So, I, I don't know from your perspective, what, what do you see as some of the driving forces behind this growing shift? Yeah, I think, uh, I like to kind of start with the customer.

So, you know, ultimately partners want to sell where buyers are buying, and increasingly that's marketplace. So, you know, a lot of companies are scaling their usage, but I would say, you know, nearly 90, 94% of customers are actively carrying marketplace in some form. And, uh, you know, with customers, what they're doing is they're making larger and larger cloud commits and marketplace spend helps de-risk that minimum committed spend.

Uh, but once they start going on marketplace, our data shows that once they get a few deals under their belt, they scale their usage considerably. And, uh, you know, customers love the ability to procure very quickly. They can consolidate some of the billing relationships, uh, they can get the value faster.

And they know that a lot of the, uh, platform features around, like things like governance, customization, and access control can really help manage compliance and, uh, software consumption. So on the flip side, you know, partners, uh, love marketplace because, you know, it's not just another sales channel, but it becomes this really strategic imperative to stay competitive, unlock new revenue streams, and provide the value of sort of like a modern, uh, sort of distribution channel. And, you know, it's, whether it's the partners getting access to that committed cloud spend or accelerating sales cycle time, or just enabling this sort of co-sell motion with Google Cloud, it's really a great opportunity for them to sort of grow, uh, from a strategic standpoint, uh, the overall opportunity.

You know, so the way I think about it's like a partner that's not leveraging cloud marketplace would be the equivalent of like a retail business who is not leveraging an online store in today's digital economy. So it's just something you just have to do. Yeah.

Absolute necessity. Um, and you know, you, you, you rattled off some of the, the benefits there. Um, the, I cited at the beginning of the conversation, you know, we did a study with Google Cloud and, you know, let me just read off some of the stats that we found, um, in doing this study is BS seeing, uh, 112% increase in their average deal size.

Uh, they're also seeing a 14% improvement in customer retention, um, uh, again, for the ISBs and across all partners, um, deals are closing faster, up to 50% in time savings, um, and 70% of partners reporting that multi-year deals are more common through the marketplace. So just, those are just some of the key stats that we found. Um, you've obviously highlighted some of the data points you have at, uh, at Google Cloud, but, you know, how does, does this all stack up with kind of what you're seeing and hearing every day as you're talking with the partners engaging in the marketplace?

Yeah, yeah, absolutely. So first of all, amazing stats. I love it.

Um, very consistent with what partners are telling us in terms of the tangible benefits, but let me touch on a few of those. So, you know, on deal sizes, as you know, private offers has provided that sort of seamless transition for many partners to go from that traditional sales led motion and bringing it online. And we are consistently seeing, you know, deal sizes, like total contract value of millions and tens of millions of dollars.

In fact, we're doing multiple nine figure deals. So over a hundred million in contract value, uh, over the past year, uh, on faster deal cycle times. I think this is driven by, you know, standardized agreements, simplified negotiations, and really empowering the customers to procure a solution without engaging sort of that lengthy procurement and vendor, uh, review cycle time.

So really accelerating time to value for everyone. And then for multi-year deals, you know, we provide, uh, support for up to 10 years upfront, multi-year prepay for five years. And of course, this pricing flexibility aligns with customers who want to have greater discounts and greater cost predictability that comes with these longer term commitments.

And then lastly, on the, the retention rates, I think what we're finding is deals that happen on marketplace tend to have better renewal rates and better expansion opportunities because they kind of deeply integrate into the customer cloud ecosystem and financial commitment. So those opportunities to grow the business is considerably there. So it's no surprise that partners are shifting more and more of their business through marketplaces.

And what we're finding is some of our top partners are driving 50%, 60%, 77% of their business through cloud marketplaces. Yeah, Incredible multi, multi-dimensional benefits there. Um, something as well that you touched on earlier is the, this notion of cloud commits, the committed spending that exists and the ability for partners to be able to kind of tap into some of that opportunity.

Um, and I think, you know, we found that in the study that we did, you know, it was a definitely an important factor. Um, I also think historically there that was kind of more of a reactive, um, approach to the market, but you know, now we're seeing partners being more proactive here working with, um, you know, Google, FSR, you know, teams to kind of, you know, help, uh, you know, just maximize those opportunities. So, uh, anything you could share around, you know, what you're doing on that side of things and, and how you're helping partners understand the cloud commit landscape and, you know, working kind of in this co-sell tandem motion there.

Yeah, absolutely. So I would say we're doing a number of things. Lemme just highlight a few.

So I think one is, you know, we're providing, you know, data and visibility and tooling, uh, incentives, uh, various go-to-market initiatives. So for example, uh, deal registration. So this, uh, our solution connect platform enable ISVs to register deals and basically enable them to connect with, uh, reps, uh, our cloud reps on a particular opportunity.

So this really ensures a very coordinated joint sales efforts. Uh, and of course our reps have quota attainment, uh, for marketplace transactions. So this creates a very powerful alignment and really encourages them to, uh, engage with ISVs and their products because, uh, you know, they're already incented or motivated to engage with ISVs because, you know, it's a critical part of customer workloads.

They can accelerate customer migrations. And, uh, sometimes it's part of this, uh, you know, this platform consumption capability. Uh, we also provide incentives.

Uh, so for example, we have something called the Marketplace Customer Credit Program. And this gives net new deals to marketplace and customers the equivalent of a 3% first year a CV Google Cloud credit. So this has been very effective to accelerate customers purchasing a specific ISV solution on marketplace for the first time.

And then I would say we also are rolling out other tools like propensity to buy tooling. So this is leveraging our data on customer usage, spending behavior, and effectively partners can give us a list of target accounts and we can generate a propensity score. And this enables them to have a very much more targeted, uh, uh, efforts in terms of their selling efforts and have higher probability conversion.

And then last thing I would say is we're doing a bunch of things around marketing as well. So there's broad, uh, sort of co-marketing, uh, opportunities, whether it's creating co-branded campaigns and taking advantage of incentive funds to create healthy pipeline or defray costs where they can just leverage best practices and go to market guides to help guide, uh, build and grow that marketplace business. So, uh, so it is, they say it's not just a, you know, listing products, but it's really becoming this proactive sales engine.

Mm-hmm. And we're providing the data and tooling to support them. Yeah, Lots of great programs and tools there.

Um, so now let's talk about the, the channel. Um, mm-hmm. Obviously Google Cloud has a unique channel centric model with its, um, with its marketplace, and especially with the, uh, marketplace channel private offer or MCPO program that was launched.

Te tell us a little bit about the thinking of putting channel partners, you know, at the core of your strategy and, you know, what are some of the benefits that you see from this model? Yeah, so, uh, you know, as you know, there's, there's a lot of chatter a few years ago that hey, marketplaces and channel partners, we're gonna be competing channels. But what we're finding is a lot of enterprise deals are, they have complex sales processes, negotiations involve multiple partners, and then as customers scale up their usage or marketplace, they're gonna look to their sell and services partners to help them, uh, you know, discover, procure, and deploy a very broad set of technologies.

And of course, they're gonna leverage the expertise, the sales and services expertise of the partners, um, as they manage through the customer, uh, lifecycle. So I believe, and I think it's validated throughout the industry, is that the channel partners are gonna play a critical role in driving marketplace growth. Um, so customers really demand that.

And so, you know, we think about the things that we're doing with resellers, it's very consistent with how we have a very open ecosystem. So it's whether customers can choose to work with direct or channel partners of their choice, or determine whether it's a first party or third party service that they want to, uh, uh, uh, procure at the marketplace to address a particular business challenge. And, you know, what we're finding with our, uh, traditional resellers is we're going through a little bit of an evolution.

So, you know, as they embrace and work with cloud marketplace, they're not gonna take their traditional sort of resell fulfillment licensing model and bring it that online, but instead, they're going to expand their role and value proposition because, you know, what we're doing is we're streamlining some of the billing and operations so they can focus on more higher value added services, right? Whether it's, uh, bundling services with marketplace solutions or bringing their own, uh, professional services capability or managing the cloud spending. I think this enables sort of this broader sort of business outcome, uh, and, uh, an impact, uh, working with the broader ecosystem, working with our customers.

Yeah, I totally agree. But at the same time, we also sometimes have to be a little realistic, and there are times when the ISV reseller connection is, you know, just not as, uh, as smooth as, uh, we might want it to be. Um, could be an ISV that doesn't really know how to work with resellers, um, or yeah, or vice versa, or reseller.

It might not be proficient in a particular ISVs technology. So how, how are you thinking about, you know, kind of managing the potential clashes that, you know, might have kind of in this engagement model when you're kind of really now at the, at the center of this ecosystem? Yeah.

Yeah. I think we're doing a few things. So I think particularly on a particular deal, I think what we're doing is a bunch of things to enable early engagement in a deal.

So, uh, you know, certainly if that engagement happens at the 11th hour where like an ISV is already quoted to the customer, that can cause some friction. So we're providing some tools, uh, to our partners to enable, you know, telemetry and visibility earlier in the sales cycle. So the ISVs and reseller can align on things like commercials and they can dev jointly develop the opportunity.

We're also, you know, doing some things around robust training and enablement. So like, for example, marketplace, marketplace specific training where both ISVs and resellers can access training that focus us on how to, you know, transact on marketplaces best practices for creating private offers, managing reseller relationships, and navigating the whole entire co-sell programs. And then what I would also say is there's some other things that we're doing, like standard reseller agreement.

So like, for example, if a reseller and ISV haven't worked together, there's an ability to sort of grab standard reseller templates, enable that collaboration and really close deals faster, and it scales in a very efficient way. And of course, there's a bunch of tools that we're providing the ISVs on the platform, whether it's like granular discounting, uh, you know, enabling entitlement transfers or being able to bring their own channel partners from their own channel network in a very, uh, you know, frictionless way. There's a bunch of tools that we're looking at.

Uh, also improving things like enhanced reporting. So imagine, uh, providing granular reports to both the ISV and the reseller for the respective deals, including customer data, reseller, data consumption details, and progress against committed spend. So I think what we're doing is really create more of a partner centric marketplace.

So moving beyond sort of the basic functions of, of, of a storefront, but really enabling that ISV reseller relationship that's not in a transaction, but more of a successful collaborative partnership. Yeah, the under the hood stuff is so critical. Um, now you have a lot of success stories, um, you know, in, in the Google cloud marketplace.

I think the one that has gonna a lot, gotten a lot of airtime this year has been Palo Alto Networks. 5 billion in sales through the Google Cloud marketplace. Yeah.

Um, what are some of the things that you see, you know, companies like Palo Alto Networks or others doing, you know, to really be successful, uh, in the marketplace? Yeah, so Palo Alto Network, specifically the way to think about them is it's a very strategic and collaborative approach they've had from, uh, from going back a number of years. It isn't just sort of simply listing their products on marketplace, but really just integrated their, their business sales motion and technology with Google Cloud.

So I think, I think it all starts with co-innovation. So, you know, we have a number of, uh, solution integrations across a number of different areas, and, you know, of course that enables customers to experience solutions that feel very cloud native to their Google Cloud environment. And it really creates a very massive selling point for customers who want a very seamless, non-disruptive security solution.

The other thing that they've done is they've leaned in very heavily in terms of, uh, the go-to market and partner ecosystem and creating this sort of co-sell motion that's sort of best in class for us. So they've embraced marketplaces, the sales channel, they have, uh, over 30 listings on the marketplace. They have very comprehensive tech technical documentation and reference architectures to help customers, uh, with seamless deployment.

And of course, um, you know, the way to think about this is that, uh, this was sort of a multi-year journey for, for Palo Alto Networks, which was, you know, getting listed on marketplace was, was relatively easy. But, you know, there is no channel where any partner can just get listed and all of a sudden you get all these deals in pipeline. You had to be very intentional and invest.

And, you know, what we're seeing is, uh, you have to view this as a long-term strategic growth opportunity that may take a couple of years to scale, you know, and what we'll see is, you know, over, over the couple of years partners that are doing it very well, they are doing things like product integration, internal organizational alignment, sales enablement, you know, having the right policies in terms of like pricing and how they comp their reps. They get invest in people, you know, maybe some operational capabilities like a deal desk. And these are the type of things you have to do to get to your first like 10 deals and get that flywheel going, and then ultimately invest at scale where the marketplace ultimately represents 30, 40, 50% of your business.

Yeah. It's like you said a couple times, kinda like anything in life, but it's not just listing on the marketplace in order to, you know, see good results. You, you, you, you need to invest.

Uh, kind of further into that, and you've touched on, um, a lot of the, you know, the good things that you see partners doing there. You know, our research showed things like successful partners at minimum have it dedicated, you know, cloud, uh, deal desk and sales team to kind of help, um, operationalize it. And even things like comp mutual plans to ensure that the sales organizations are are, are bought in.

But, you know, for, for kind of new partners out there, um, what what would be your kind of one, two pieces of advice for a partner that's kind of just getting started or thinking about, um, you know, new into the marketplace and, you know, how do they think about driving kind of long-term success? Yeah, yeah. So again, I would point at the foundation has to be this having a very differentiated offering and a very better together story.

So what is the joint value proposition? Why does your product align well with Google Cloud and how does the marriage between your offering and Google Cloud really provides this great benefits to the end customer? And you know, what seems to go very well is if there are strong integrations with our first party services, whether it's like AI and data and analytics and security, uh, you know, this also drives, uh, a big part of that success.

And, uh, you know, you gotta make sure that once you have this better together story, that it becomes very easily and enabled to not only your own sellers, but our, uh, our sellers as well. Mm-hmm. Uh, the other piece that you mentioned is investing in people, processes, uh, marketing, uh, relationships and enablement, because I think what companies do are, you know, they have to modify maybe their systems and processes to align with marketplace.

We mentioned the, um, the, uh, the deal desk. We have to make sure that you evolve and get more of a sales or revenue leader, uh, function. Maybe as you start, it becomes a little bit more of an alliance led, uh, motion, but over time as you get more success, you get, uh, executive, uh, uh, sponsorship with the chief revenue officer or the sales leader in the organization as well.

Yeah. And then from a, from a technical, uh, or tactical standpoint, you know, you gotta register deals, uh, and when you register deals, you know, one of the things I recommended as companies get started is you need to establish that track record of success. So, you know, identify like a, a geography, a customer segment or an industry where you had some early success.

And then once you have a couple of wins underneath your belt, you can work with your advocates and sponsors within Google Cloud to, uh, elevate these, these wins, these win wires. And then what what happens is it becomes a little bit of a self-feeding process where you build momentum, you get some differentiation, and you get some wins, and then you can scale, uh, pretty significantly thereafter. Yeah, absolutely.

That internal selling is so critical in a lot for a lot of these, uh, uh, companies like looking to get buy in, like in anything in life. Right. Um, so now, like, kind of forecasting a little bit, um, give us a sneak peek.

What are some of the things that, um, you know, you and your team are thinking about or, or, or working on? Any, anything that you know might wanna highlight that you, is, that you're able to share that might be coming down the road that would benefit some of your ISV and channel partners? Yeah, maybe, maybe I'll just highlight a few things we recently launched.

So, I mean, certainly one thing we launched was, uh, introduce a new variable rev share model. So for eligible partners in their deals, rev share can go as low as one point a half percent for things like renewals or large contract, uh, uh, uh, uh, deal sizes. Uh, we, we also, as I mentioned before, we went general availability with this end customer, uh, incentive program, 3% for the first year a CV.

So this has been great to unlock and acquire new customers. And then earlier this year we also launched, uh, professional services. Uh, so professional services is a formal solution type on marketplace.

So at first, the ability to cross-sell or upsell things like implementation services, training assessments, and managed services. But I think what this does is it creates a nice building block for us to drive more business outcome and solutions for end customers because, uh, you know, between the ability of a sell and services partners to focus on, uh, solutions between like different ISV solutions, multi-vendor private offers, marketplace becomes this potential connective tissue between the different partners who participate in marketplace delivery and value add. So imagine of shifting from simple products and SKUs to more customer outcomes and complete solutions.

And I think marketplace will be a key enabler for that when you think about these multi-vendor private offers. The other area that I would say that we're investing in quite a bit is, um, activating, uh, product-led growth. So PLG.

So, uh, as you can imagine, this is the ability to sort of self-serve. Uh, this was the original promise of marketplace, but what we're finding is that there are a lot of capabilities between personalization and AI that will make this a little bit more real. So we're improving the search experience, we're gonna improve some analytics so that you can drive a campaign directly to a marketplace listing mm-hmm.

And track where they are in the funnel. And then we'll also surface, uh, third party solutions in context across the broader cloud console. Uh, so for example, if you're a Vertex AI developer, ML practitioner, you should be able to see related solutions from our marketplace within your experience.

And then lastly, what I would say is, um, we're gonna do some things to really automate the partner co-selling journey from lead to cash. So some of the things we're doing around like APIs in terms of like deal registration and private offer creation will create some of the automation, uh, that our partners have been asking for. So a lot of great opportunities and a lot of great areas of innovation that we're driving.

Yeah, lots of innovation, both I'd say on the front and back end there and, uh, and raw expansive of the, of the program. Uh, that's great. So look, we're getting close to wrapping up now.

Maybe just a final kind of thought, even looking further out and, you know, the mm-hmm. The theme of, uh, this year in the, in the technology industry really has been, um, obviously ai, but I think even more so ag agentic ai. Yeah.

And just want to think about, you know, how do you see, you know, marketplaces playing, you know, an important role in a world of agentic ai? Um, you know, you obviously launched a new category this year, so clearly it's something that, uh, that, uh, is, is, uh, important in the, in, in the halls of Google Cloud marketplace. Yeah, yeah.

So, uh, listen, I, I don't need to tell you that the market opportunity for Gentech AI is just massive and, uh, you know, there's a lot of growth. It's, it's really shifting from this reactive, uh, to something that's a little bit more proactive in goal-oriented solutions. And I think because the AI agents can, they can reason, they can plan, they can act autonomously across a, a complex set of tasks.

And I think what we're gonna see is, uh, maybe evolution of AI agents as a solution, right? So certainly AI models and services has been available in cloud marketplaces for some time now, but AI agents represent that next evolution. So, uh, you know, the simple model performs a single task to an agent of software that could perceive an environment, reason, make decisions, and act autonomously creates a tremendous opportunity.

And the reason why I think cloud marketplaces really that go to market and commercialization innovation model is when you think about, um, you know, where's all the innovation gonna happen? It's all gonna be across the ecosystem. So, you know, we might have, you know, 5,000, 10,000 agents on marketplace within a couple of years where, you know, you need to be able to search and discover and look for business outcomes.

You need simplified procurement, you need quality signals around trust and security. You need scalability capabilities and integration. And, you know, I think all these things come to marketplace as that primary solution and route to market for this, for this, for this area.

Now, in the near term, as you mentioned, we launched an AI agent marketplace, uh, in April. And now partners have the ability to not only list and monetize, uh, their agents, but now potentially integrate into agent space. So agent space is our solution for the general business and knowledge worker, where you can basically have, uh, business users not only work with agents that are first party custom agents, but also a rich ecosystem of agents that they might acquire through a marketplace.

So it really creates a great opportunity to extend the reach and drive innovation, uh, with, uh, with, with ai undoubtedly. So it's a huge opportunity. Yeah.

Yeah. This space just continues to get more and more exciting. Um, so Dai, thank you so much for, um, you know, hopping on here and talking all, you know, good things, uh, marketplace, um, really, you know, enjoyable conversation.

And I always learn a lot when, uh, when speaking with you on this topic. Um, and to all of our, uh, uh, viewers out there, thank you for taking the time and, uh, have a great rest of the day. Guy's gonna apologize for not apologizing and he's gonna tell us that AI is good for AIing things.

Jay would like everything as code, including the entire business. I'm not sure he wants his weekends as code that join me on the Tech Field Day podcast as we discuss with and twiddling all the dials, is really what your business is about. Welcome to the Tech Field Day podcast, where we bring together a group of ITE technical experts discussing a single idea around key concepts in the industry.

This podcast features a variety of perspectives from members of the tech field aid delegate community, and it's often recorded in association with one of our events. Tech Field Day is part of the Futurum Group, and this podcast is also published on our sister company Site Techstrong tv. On this episode, we'll be discussing the premise that you aren't in the business of twiddling the dials.

But before we dive into the discussion, let's see who's on the panel today. And my top is Guy. Yes.

Hi Al. Um, good to be here. Guy Courier, I'm a, uh, analyst at Futureum Group and, um, uh, infrastructure and Dial Twiddling is one of the things I cover regularly.

I'm Jay. Hi, I'm Jay Kroll. I'm the Chief Product Officer, nexus Tech, and we twiddle dial so that our clients don't have to happy to be here.

And of course, I'm Alistair Kok. I'm an event lead here at Tick Field Day and long time in vi dial twiddler. I've had my career in, uh, twiddling the dials and then teaching other people how to twiddle the dials on various technologies for the last 20 years.

And it does seem like the dial twiddling the dips of the technical knowledge is less valuable than it used to be. That going back to the days back when I used to have here and, uh, long before I was involved in tech field day, uh, I would be assembling servers. I'd be actually putting the CPUs inside the servers when there was an upgrade or adding additional RAM to them and hand delivering those servers into the data centers.

Well, here in rural or semi-rural New Zealand, you might be lucky and have a rack, otherwise it was just a cupboard. Uh, but I'd be delivering this hardware directly into the customers and going through the entire build out process of installing operating systems and doing all of the migration of their data. And there's a huge amount of very detailed knowledge in there.

And yet as I progressed through those 20 years since building automation to hide all of the dials, whittling, and then consuming services where somebody else has automatically twiddled the dials for me seems to have been a big thing. But I've kind of noticed over the last few years that the cloud started off being a very simple thing where you just bought in t-shirt sizes off the rack and has now become pretty complicated with a whole lot of dials to twiddle guy. You also cover dial twiddling.

Is there less or is there more dial twiddling? And is that actually valuable to a business More and no, it's not valuable. Shall I expand?

Um, so, uh, yeah, yeah, I think, I think one of the things I've identified is, uh, uh, really interesting. And the other thing that you've hinted at is even more interesting. The thing that's really interesting is the dial tool.

Loing has gone upscale, if you like. Um, it's not swapping out subsystems anymore or, or, or CPU or what have you. Um, but, uh, um, there's a fair amount of on the ground operational work that's, uh, done in your average data center right now, let's call it the level below site reliability engineer.

Um, and so maybe that's the floor and it's a higher floor than it used to be. Um, but the other thing that you hinted at, I didn't quite say directly, is that the, um, ification, the increasing abstraction of services, including really low level services up into cloud-like platforms and orchestration systems, has made it a lot easier to get access to a whole lot more and fine grained dials to Twitter. And so that's really to me, where a lot of operational efficiencies get s sapped.

And in fact, I don't think we're really answering the problem now with ai. I think we maybe even are exacerbating the problem in some ways because AI has looked to as this answer to, oh, there's so many dial total. Now it's also complicated.

Oh, woe is us, what are we gonna do? And, you know, incomes management or whoever else saying, oh, we're just gonna have an agent do it for you. I don't really think that addresses the problem.

But from the standpoint of more dials to twiddle and more dial twiddling going on, absolutely, it's exponentially inverse and it's a problem. And Jay will explain why. So with that setup, I will tell you a small story.

Uh, I'll change all the names to protect the innocent, but, uh, in reviewing a private cloud, I, uh, did some plunking around, found out that in this situation, uh, the, the client's actually standing it up the old fashioned way. They, their home rolling, not their encryption. That'd be awful.

Not anyone's ever written that, but their, their home rolling their Kubernetes stack and of, because of that, they're, uh, bypassing all the wonders of, uh, insert vendor name, cloud foundation. Uh, they're bypassing all the wonders of blueprints and other available automations that would've simplified their lives, not because they're trying to do it one sprocket at a time if you, and so when I found out this, I realized that's really taking away a lot of what the business is actually requiring of, of them, which is time to market. And so for them putting it together one stick at a time, uh, we're, we're really robbing the future of what the possibilities would've been for them to execute upon that to get those business aligned applications up and running in that environment.

So I think of it just as, what's the time to value? Um, what's the speed of return on the invested capital? And you are investing significant capital, so how can we shorten or lessen that time window to get that return?

So I describe the environment where there's more dials to twiddle. But, but, uh, uh, a and Jay, um, I, I want, I wanna ask you guys, um, to me there are two reasons for this phenomenon of what that horrifying situation Jay just described of homegrown Kubernetes. One element of it is we want it to be exactly how we want it to be, and off the shelf, or even customized or whatever, it's all expense and it comes out to the same dollar amount in the end or whatever.

That's not really true, but we'll leave that aside. It has to be exactly how we want. That to me, is like, let's say the good reason, the good motivation, the bad motivation is the unconscious desire to twiddle vibes.

It's fun, it's interesting. I'm gonna learn more. So I'm asking the two of you.

Um, sorry, Al, I turned the tables on you. I'm asking questions. Um, which one, what, what's the mix of those right now in your experience?

Because I still feel a whole lot that it's emotion driving it, it's just fun to do. So we come up with the justifications for it. I definitely wouldn't underestimate that social part, that part where it is.

I've always twiddled with the novel knobs. Uh, the thing that defines me as a professional is that I have a deep knowledge of where all the knobs are and where to twiddle them. That absolutely is one of the factors in, in overcomplicating systems, and I think that's one to, to face up to and to start to, to view is that the knobs I want to twiddle need to be more abstracted.

So they need to be less about the underlying technology and more about how they serve business. And that, I think is the, the kind of message from Jay as well is that there's that time to value and, and that, um, return on investment, uh, this is why we're in business is to, to return some value, right? Uh, I'd add an another complexion to Jay's comment about time to value isn't sufficient.

It's also, uh, the sort of return on investment or the return rate on that investment. And so this is what sometimes leads us into the feeling that we need that perfectly crafted solution that is gonna be the most efficient return on the money we are spending. And there's, there's always this balance of how much you overanalyze, getting things perfect and the wasted energy and, and effectively money and time that you spent perfecting something versus the return you get from it.

Uh, it really is one of those, those things where there isn't a single answer of, of you should always work with a standardized tool because the standardized tool is cheap to deploy, but it's then more expensive long term to operate, maybe if your specialized use case. But I think there's also a psychological thing. The same thing is that desire to twiddle all the knobs and be the, the expert in the knob twiddling is also the feeling that we're all unique snowflakes.

Our business requirements are unique snowflakes. And so we've gotta build a unique snowflake of an environment or an application to fulfill us and not sure that really fits together very well. Jay, uh, what about, what else, uh, do you see in this at, at different levels within organizations as well?

Different skill levels and different, um, management focus as well. I think there's, there's more of a twiddling knobs as you are lower down towards the technology and, and less interest in the knobs as you move up into more of the executive and business oriented views. I think of it maybe as a slider.

So if we started off with the notion of that the very, very lowest of low level, you have a troop, uh, perfectionist twiddling every last knob to its most extent, uh, optimizing, constantly refining, going for it, then you have a practitioner, uh, that's there to maybe get some things done, um, maybe have a professional, uh, that's there with a business angle to try to really accomplish what the, what the business is asking for. And then you maybe have like a pragmatic, if that's a word we could use to describe someone that is not only aligned to your earlier point about just the return investment, but the ongoing compounding return from that capital that we deployed. It's now supporting a named application.

We can tie that back to top line revenue. Uh, we understand what the impact is to the bottom line, and they've really taken this down more of a TBM or technology business management perspective. So getting there is the hard part because, uh, as someone who also does love to tinker, uh, I, I do, I do cherish my weekends, the finite amount of time I get to go, go on the console and get things done personally for my own projects, it is, uh, it is a temptation that I fully understand and grasp.

Uh, but at the end of the day, you know, the things that we have to do for the business really require us to put on our pian hat and, and probably put some of that perfectionist hat, uh, away per, uh, was it, uh, uh, perfect as the enemy of progress. I think I've heard that other, uh, term of ps. So that's what I think about that.

Part of the trigger for this conversation was I was talking about my transition from being a very technical person doing building things, building things like my auto lab, you know, infrastructure automation and building solutions to companies, to my current role with Tech Shield Day, where I'm an enabler of people more than anything else. And that sort of feeling of fulfillment from working on the physical building, the solutions to technical solutions to, to business problems is, uh, the thing that, that goes away is the big, big risk is we bring somebody technical into this role. And it, it kind of feels that internal thing of, I, I really do feel internally that, uh, twiddling the knobs, understanding the depths of the technology is a vital thing for me.

But yet, as I transitioned from teaching on-premises infrastructure to teaching the web and, uh, teaching in particular AWS uh, training courses, it was, this stuff changes all the time. You're never going to be the most expert, the, even in a single product, you're never going to be the most expert, let alone across the hundreds of products that are out there. And that's kind of where I see the, the return of the, the knobs twiddling, even in cloud where everything's supposed to be a service and you're just not supposed to care, is assembling together.

All of these pieces feels a little bit like we've abstracted away what we used to care about, and now we've made something incredibly detailed to care about that is new and different. Are we actually making any progress in this guy? Or are we just redirecting our knob twiddling energy?

I think that, well, so let me bring AI back into this, um, this time of that. Apologies, I usually apologize. Um, it, so I'll start with this.

Um, I think it's evident that the best users of AI understand how AI works. It might not build models, but they understand the fundamentals of how a model works, and whether it's generated generative AI or predictive or whatever it is. Um, they know what, what mechanism is being used, and that allows them to use that tool better.

By the same token, that's not so obvious in ai for some reason. It may be 'cause it's simulative, so it, it fools you really easily, but it's relatively obvious in other disciplines that those who have practical hands-on experience in it like you do out in dial twiddling, um, can understand better how to direct others and what's important. And that's not important.

So, um, are we just twiddling dials higher up? Um, I don't think that's our goal. I think we wanna continue to abstract further and further, so to speak, not just with technology, but with operations or like culture and how the team works.

But I think that, um, that, so that takes us out of the, the, um, you know, fine-grained operations part of it. But nonetheless, we need to be able to evaluate and understand what's going on, and we need to accept, um, differences in ways to operate, differences in styles, differences in approach, and accept some of them to keep the grand picture in play. All of that does require a bit of a mi uh, dial twiddling mindset to, to, to, you know, to put not too fine a point on it.

I don't think you should ever really escape it, but the short answer to your question is, um, that is our goal, especially with the advent of ai to bring it back to that, is to do, do less with, um, tactics and more with policy to set policy. And then to Jay's point earlier, um, we can hobby with the things that give us pleasure and learn through doing, through twiddling those tiles. I'm getting tired of the word twiddling, but yeah, we can learn that way and make our policy decisions and our guidance more effective, whether we are guiding people or ai, sorry to say, or a mix.

So I think of this in a couple different ways. Uh, the first is that being able to know what's going on, whether it's an AI workload or any workload that is new, uh, to an organization, there's a, there's a freeing concept around knowing what's going on. But I also think that, uh, getting to observability traceability, uh, we, we actually may not have time for everyone to become fully instrumented aware and a PhD in big masks.

We may just actually have to use some of the AI ops capabilities that are being shipped by some of these vendors that allow us to plug into our on-premises infrastructure to tell us what we need to know using small language models that know exactly what one needs to know about what's going on in this environment, and may not necessarily know who, uh, won the 1985, you know, world Series. Um, uh, those of you who have chat, you boutique go figure that out. But, uh, I think it puts us on a policy as code path as we all aspire to be more businesses code.

Um, if, if we, if we don't go down that path, I do feel like we'll really, really be back at the beginning of our conversation where, you know, I'd really like to move forward, but I really don't understand fundamentally what's happening in silicon right now. I don't understand the electrons passing from, you know, point A to point B or statistically point A to point whatever. I think of it as we, we just have to set aside and trust that these platforms that are available in market right now packaged are gonna help us move faster and, and be better in our business.

And to your earlier points about moving or shifting up in the business value conversation, it makes it practitioners, uh, into more pragmatic going forward. That sounds a little over abstract though, Jan. I mean, when I hear businesses code, I wanna flee.

I I don't wanna be in the businesses code. Um, I want to be in the, well, you know, I'm, I'm not gonna get all mushy and say, oh, it's about people. Um, which it is, but I just don't wanna get all mushy about that.

But that's, that's not what I was saying. What I was saying was that, um, your, your, your dips and dives and hobbying are really necessary in order to make the, the, the best kinds of higher level decisions and guidance. And I mean, don't we want our team members to be able to do that?

So that, you know, to put to, to, to put AL'S hat on right now, you know, if he didn't make it clear, he's, he's now a puppet master orchestrating all kinds of other people to go and do things with tech field day. That's what he does now. Um, he used to be a real worker and now he's just motivating people and coordinating and things like that.

But no, but seriously though, um, one of the things that Al can do as a team leader is to enable the team members team to similarly be puppet masters for whatever tools that they have, but they can't really do that if they dunno how those tools work. And I feel like you're going a little too far, Jay, in, in terms of trying to abstract out what the teams are trying to accomplish to, to the level of, you know, business guidance as opposed to, you know, serious architectural and operational decisions in the technology. I, I, I, I get it.

I get it. It's, it's high level. But I would also counter that the business' code would be what are the essential operating KPIs, OKRs that we all agree as a business that we're gonna chase after.

Now, how do I take that, make it real for the amount of infrastructure that we've deployed on premises? Do I know exactly where that capitalist went? What's, what's that mean for I business?

And is it tied back to a metric that matters? If it is simply like we have the capability, we have the ability we can do, if we can do anything, I go back to all my s enables possible, permissible, sustainable, repeatable, advisable, and if we're just sticking here in the art of possible, we're never gonna get advisable. And so that, that's my only pushback is if it's not advisable for the business, why are we even going down the path?

But I also seed, I, I admit there is some experimentation that makes us all better over time too. I just, I'm trying to figure out where, where's the balance in this. Um, so that's, that's sort of my rhetoric.

And I don't think there's a single balance point. And I think that's one of the things, as within a team, you are going to have those different personas, those different roles. And the larger the team, the more diverse the roles are gonna be.

There is probably always going to be a role for somebody who has to optimize the thing that is business critical, really expensive and needs some very careful dial control. But there's also gonna be a role for the person who sets very abstracted policies that then filter through, I think, you know, understanding that there isn't a single answer to what level of skills you require, uh, and what skill set do you require is, is really vital here. Um, particularly as we then start to see that there needs to be conflict between some of those roles as well, that there will naturally be a conflict.

And that's something I see in my current role, uh, between the, the different constituents. And, you know, one of the things we, we often think is everything's gonna be harmonious and everything is gonna be beautiful and all, all gonna be rainbows and unicorns. And that, that simply isn't the way groups of people work together.

And understanding that is an important part. Yeah, you don't even want that to be the case. That's, that's, you know, we're, we're talking about, um, certain kind of disruption and, and di dis disease and discomfort that, uh, leads to innovation.

And innovation is a word, right? But the idea is that you can gain, uh, a greater value the earlier you can move in new directions that are productive. There's moving in new directions that are not productive and the value that gets wasted there.

But, uh, I guess I'm coming around to your point, al, which is, you know, ultimately, like here we are, um, we, we need, um, multiple levels and multiple perspectives, and that creates, um, an unsettling environment. But as workers, we wanna, we wanna, um, deal with that. I, it's silly to say get comfortable with something uncomfortable, but we wanna deal with that and welcome it.

Um, and, uh, and as, as leaders, uh, we want to both foster it and help everyone, you know, deal with it. And then, then you have more of a virtuous system there. So, so to get back to dial twiddling, it sounds like we wanna both twi, it's like the Schrodinger's dial twiddling, we wanna do it, and we also wanna not do it all at the same time.

Schrodinger's dial the dial that you're both twiddling and not twiddling at the same time, depending upon whether you observe it afterwards. Yes, there you go. Schrodinger's dial.

Oh, that's Great. That's great. And of course, you only know afterwards whether it was worth making that change on the dial because you result No, that Would be the Heisenberg.

No, that would be the Heisenberg dial Heisenberg's style. So we're coming to the conclusion, I think that it's, it's that uncertainty of, of the dial, the need for all of those dials, all of the knobs and the switches, and that there is definitely going to be a desire for simplicity and a desire for getting to a state where we can just describe the result rather than having to look at every single setting that makes up that result. And that's the, the everything is code, policy is code, infrastructure is code.

Um, guy was loath having a business as code. But no, it's a direction to head towards. Uh, ye there's still a need to get into some of the weeds now and then, and there's a need for people who understand the weeds.

I think, uh, we are a long way away from being able to just hand all the reins over to an ai, and that's probably a good thing because the time we hand the reins over to AI completely, it means that the humans are no longer so unnecessary. Uh, and I think, uh, humans are kind of important to us. We certainly see far more innovation from that, again, conflict between the people and the, and the humans.

Now we're heading towards the, um, heading towards our time. And so I just, we give you an opportunity to have one last thought. Jay, do you have one last thought to add to this conversation?

I do think that in terms of the number of knobs, we all all have a choice, many more choices now, uh, around the constraints and the affordances. You know, we can procure a product experience, we can get an engineered product experience that has a lot less knobs, uh, because it's been pre tuned, pre optimized, you know, for our safety protection, um, uh, like the, uh, like the airplane where you have to have to have the dog in the cockpit. What's the purpose is to basically, uh, you know, keep the, keep the pilot very, you know, you know, happy during the flight because they don't really do as much.

And then, uh, viciously attack them, should they try to, you know, arrest controls of the plane at any point in time, keep them away from those controls. So, uh, at, at the risk of saying, uh, it's, it's gonna get, uh, less knobs exposed to us over time. I think the appetite for knobs will actually decrease over time.

Um, any, any recent interview you might have done with a, a level one candidate and your organization to matriculate, uh, they may not have ever put together a pc. They don't know what a board stiffener reason is for a server. Um, so some of those concepts will, I, I think go the way of people that can identify token ring at a distance, people that know what, uh, different ethernet, uh, interfaces looked like.

Um, but over time, I think what the product experience will be for infrastructure buyers is much more refined, uh, again, much more of a product engineered experience. And I think that's a good thing. Um, and I think there will still be, you know, enough dos and dials and things to twiddle, uh, but, uh, a less, less so, um, going forward.

That's my perspective on the market. I dunno. Yeah, I guess, I think you're right, Jay, but I guess we might wind up reframing what didn't you used to look so much like a bunch of dials as dials now, and there will be, again, that floor will go higher up in terms of what we're doing.

I, I, I almost, I almost see it as, so it occurs to me that a lot of what we deal with, with the people, part of people process technology in, in, in, you know, our industry, um, a lot of what we deal with is mindsets that tend towards the more extreme positions. And we've all dealt with this, that passion seemed to run high among the, the, the, you know, um, the, the tech operators and architects and so forth. There's only one way to do this.

I mean, how many times have we heard this? This is the only real way to do it. All the other ones are know, we should be able Alice smiling.

He's, he's probably the most recent of us to be like in the room there, um, discussing these things. And, um, maybe the takeaway I'm getting from this is that that tension that we were talking about is something to try to foster a little bit in our, in our industry. This is one example of it, um, which is do you twiddle the dials or do you just, you know, use the platform where, where's the bending point?

Um, Jay, I think your guidance, um, in general around ensuring that you make good business decisions and that what level of detail you're getting into in operations and tooling, um, should be based on, you know, where your value point is and what your goals are and all that sort of thing. That's completely correct, and it's the sort of thing I say all the time, but I'm also kind of feeling like you can only get out of your team and your organization, what it's able to produce and, and, um, uh, sh shifting what it's able to produce takes a really long time. So in the meantime, fostering this idea that a tension between what you want to be doing and what you must do and recognizing that and realizing that it's all not, it's not all one or not all the other, that would be a really great thing to see more of in our industry because I think we do tend more towards more extreme statements and positions because that is what we believe in and what seems the most fulfilling to us.

Well, thank you all for joining me, particularly Guy and, and Jay, uh, and for this episode Field Day podcast guy, you clearly care and discuss them. Can people continue the conversation with you? com.

Um, you'll also find me at text on TV doing sometimes the podcasts over there or webinars, and of course, find me on LinkedIn. And Jay, where can people connect with you? org.

com. org. Awesome.

And of course, I'm Alistair Cook. com site on Tech Field Day, as well as on deta nz, which is my own personal much, uh, neglected blog too. Of course, LinkedIn is the place I've been publishing a lot more content.

Thank you for listening to this episode of the Tech Field Day podcast. If you enjoyed the discussion, please subscribe on YouTube or your favorite podcast application so you don't miss a single episode. Do consider giving us a rating in a nice review.

This podcast was brought to you by Tech Field Day, home of IT experts from across the enterprise and a part of the Futurum group for upcoming e episodes events. com and slash podcast will find the podcast. Of course, you can always find us on Techstrong tv.

Thanks for listening and we will see you next week. Hey everyone, welcome back to the six five Summit Cloud Infrastructure spotlight. Excited to have this conversation with Koran Botta, SVP at Oracle.

We're gonna talk a little bit about multi-cloud and ai. Koran, good to see you again, it's been a minute, but, uh, always love having our, our conversations. Yeah, thanks for having me, Daniel.

Excited to talk, uh, all things cloud. So I, I've spent a lot of time over the last few years, uh, having conversations, you know, Oracle went through this pretty significant transformation in its own strategy. The cloud business has been explosive among the fastest growing each and every quarter.

We track that very closely, but for a while it was kind of like among the peers. Everyone was sort of running their own race. It was like, Aw, WS Google, Microsoft, Oracle, we're gonna get as big as we can.

We are gonna try to dominate our markets. We are going to convince the world that everyone only needs our cloud to do everything. And then all of a sudden there was a bit of an aha.

And, uh, Oracle had a big aha, um, you know, over the last couple of years and last year in a really big way where you went kind of all in on multi-cloud, made some big partnerships, um, you know, announced with all the major hyperscalers. And I mean, I'm sure it had a lot to do with, uh, how important your database is to so many of their enterprise clients. But, um, you, because necessarily really stopped and kind of smelled the moment when that took place.

But, uh, you were one of the ones that kind of went from that. No, we're gonna be very much a wall garden to no, we're all in on multi-cloud happening really, really quickly, you know, as you expanded all your partnerships, kind of what drove that and, uh, how do you see that kind of just in the near term, you know, evolving into the application space? Yeah, yeah, definitely.

Thanks. Um, I think, you know, it, it's been a journey, right? I mean, multi-cloud, I think for us, uh, you know, I think I, I think it's simplistic now when we talk about database, but really we started this journey five years ago.

We didn't even think about database at the time. What we really wanted to accomplish was in the old, in the old days, you know, when you used to go build a private cloud, you used to be able to go buy different parts of things from different vendors. You could buy a switch, you could buy a host, you could buy some storage, you could buy some different pieces of software.

You could collaborate it together and build a private cloud with many different vendors. That world didn't really exist in the cloud, right? It was sort of like everybody was playing hungry, hungry Hippo, like you said.

Everybody was like, well, if you are gonna come to my cloud, you should use everything from my cloud and I'm gonna make it extremely expensive to move outta my cloud. You have to buy everything and you have to be all in on my cloud, right? So it became very hard for the customer to pick and choose one or two, uh, uh, you know, features or services from one cloud as a decision point.

We wanted to break down some of those barriers, right? We started with, let's say networking as an example. Customers should have the freedom to move their data around.

They started moving their data around. So, you know, we signed a interconnect agreement with Microsoft. No money exchanged hands, but it was the ability for customers to treat two regions as a singular entity so that customers could move data around.

You know, now fast forward to now, you know, we now have a host of data services now available through all three clouds. So I think that journey has taken a long time. It's been breaking down small little barriers.

Um, uh, but the goal really was to enable choice for our customers. Of course, our database is the market leading database out there, but also to give, uh, customers the ability to pick and choose the best of breed from every cloud and just use whatever cloud they want. So that means interoperability, integration, data, movement, you know, I'm not, I'm not even talking about databases yet, right?

So it's, it's all these other things that combine it to make it a multi-cloud experience for a customer. Yeah, I'm still playing hungry, hungry hippos in my head, thinking back to my youth. Um, no, but I, you know, it's interesting because now we're in this inflection with ai.

So you spent the last, you know, almost decade building out O-C-I-O-C-I Gen two. I think gen two was such a pinnacle moment, you know, where it really started to turn and that great growth ensued. But then AI came, and AI is like, oh, we're gonna have to build a whole new data center.

You know, you, you have a great partnership with Nvidia, but you listen to Jensen and it almost makes it sound like everything you've done in the past is almost like stop, pause, cutoff point. You now gotta build these new AI factories from the ground up, new architecture, new applications agents, you know, talk a little bit about that because you've gone big, you've got Stargate, of course, you were mentioned as part of that project. You know, talk about the approach there, because this seems like it's almost a moment to take everything you've learned, but to almost start over and try to become the, you know, the, the cloud company of the AI era, which almost seems like a new race.

Yeah, no, absolutely. I think you're absolutely right. I think, um, if you look at our past history of, you know, it'll be 10 years for OCI next year, like almost a decade, and, you know, we started with our, our core fundamentals of what makes OCI, you know, really great for infrastructure customers, which was extremely high performance for very affordable economics, right?

We wanted to marry the best of what you get with private cloud or your on-premise clusters and marry with the benefits that you get from a cloud. That was always the, the sort of the, the kind of a core tenet of how we built OCI. So when we built OCI that, that's the same thing with exodata, right?

We have IB and exodata as an example, where you have a, a database cluster. Um, we used the same technology to enable HPC, uh, several years ago where customers could come in, they could launch HPC clusters, and then the, you know, the AI advent sort of started of, you know, whatever it was like, or, you know, when Chad GBT came out and suddenly that became a thing and we used similar technology and expanded that, right? It became, we were already invested in the, down that path, which was extreme performance.

So I think the scale is just different now, right? We're talking about, you know, 120, 130,000 cluster, uh, GPUs per a single cluster when you're building something that big, um, the focus goes away from just sort of, you know, one part to the entire thing from all the way down from construction, uh, construction of the physical buildings, um, how they're interconnected, um, all the way up to things like liquid cooling and sort of the, the, the, the loop that goes around in the buildings, um, all the way to the network, uh, the network build out, and then even software, uh, software based, uh, um, you know, uh, uh, operation. So, you know, when you're running a hundred thousand GPUs in a cluster for a single training job, you gotta make sure everything runs, uh, you know, uh, cool run.

Everything runs at the, at the right time. Everything is, um, running without any issue. 'cause one GP goes down, your entire training job can go down, right?

So there's a lot of work and effort being put into, uh, software operations automation, making sure things are up and running, what your RMA process is gonna be, you know, if a GP goes down, how are you gonna be able to replace it quickly and restart the job? So yes, there's the sort of the core engineering aspect of, you know, working with your partners like Nvidia, but then there's an ecosystem of technology that we have to invest in. So there's a lot of work to do, and the scales continues to grow.

You know, of course, we've been talking a lot about, um, how our cluster sizes are growing, you know, with, with Blackwell, GPUs, there's liquid cooling. So there's a whole set of new problems to solve in the next several years. There are, and it's, it's exciting.

It's creating a huge opportunity on a global scale. Um, things that you've been doing, like regional, like zones like OCI, you know, seem to really be perfectly aligned to where we're gonna have to head with sovereign ai. You know, and it just kind of the way it's been thought about, um, there are some parallels and of course compute that's done on CPU is not going to disappear overnight.

I mean, in fact, there's still this very symbiotic relationship between CPUs and GPUs, but there is, it is very clearly a new architecture. The, the way it's being built, the way software's being designed, the way power has to be considered thermals. Everything that a data center used to be and what it will be, is going to be different.

Um, you know, one of the interesting things about Oracle though too, is you're not just an infrastructure company. You are not just a database. You're also an applications company.

And gen AI was kind of the rage for a few years. It still is, but you're sort of seeing this transformational moment now, another inflection where generative and agentic, it's sort of, it's, it's morphing into, you know, we're going to have basically agents working alongside us, and then maybe if we're lucky next year, you and I won't even have to come for this conversation. We'll just know you.

It'll know me. We'll do this, it'll, you know, come on, a a a couple of guys can dream. We could be out at the racetrack, right?

In other words, we wanna do, but like, talk a little bit about how Oracle is thinking about generative AI agents, um, you know, across your stack. Yeah, I mean, I think, I think you, you pointed out a really important thing that oracle's not just an infrastructure company or a cloud company. I mean, we're a cloud company, but we're also a applications company, right?

That's something that our competitors don't have, is we have the applications, um, and we also have the infrastructure side by side our fusion apps and our vertical applications such as healthcare with Cerner or retail and mic, cross and finance. Um, so the, the advantage that gives us is number one, all those applications can run on the same infrastructure side by side your clusters that are also running the training. And then on top of that, we have the world's most popular database where all of your data sets.

So, um, you know, imagine a world where, as a customer, as an enterprise customer, you can use our platform services. We have a gen AI service where we offer things like llama, mistral, the usual set of models we recently announced Brock as well. Um, so you can essentially do rag point your database to it, retrain your models, and then you can also build agents using our agent platform as well.

So there's, there's the external cus customer component, but also internally at Oracle Daniel, what we're doing is we're injecting all of that intelligence into our applications too. So we've got hundreds of agents that we've already built across all of our applications, infusion SaaS, like an agent to help you with supply chain, an agent to help you with patient data, um, an agent to help you with public safety. Um, these, these agents are already embedded in our applications, and they allow for much, much, uh, more optimized use of those applications.

So we've sort of got a multiple pronged strategy with the apps, the infrastructure and the platform, depending on who the user is. Yeah. And I like how you kind of brought that together on our side, and we've been evaluating the agent space from the very beginning, and there's sort of these different approaches of where ag agentic will be adopted in the organization, kron.

And so, you know, if you kind of look at first blush, a lot of the early came out of applications, it was like, we're gonna build an agent inside an application, and you're using this SaaS platform or this enterprise tool, we'll give you an agent. And then there's the bit of the other side is we'll deliver it from the, um, from the infrastructure level, build the agents at the platform level on the infrastructure, a little bit more agnostically to whatever application 'cause companies don't have. If you end up having to buy agents for every software you have in your enterprise, that probably doesn't scale super well, right?

So it, it starts to look a little bit like almost how we did like, um, like virtualization and stuff like where you kind of wanna have more of a platform and in a, and a middleware layer. And then you have these companies that are kind of saying, well, we'll do it just that way. We're not really either infrastructure or app, but will be the agent layer.

So it's gonna be kind of an interesting thing to watch how this sort of diffuses, because what I know for sure is like, people are gonna wanna streamline this. They're not gonna wanna have an agent with every application and every cloud that they're they're in, they're gonna wanna get that down. So that's gonna be really interesting.

Oracle being full stack sits in a pretty good situation where you can come at it through apps, you can come at it through data, the database and ERP layer, and you could also come at it through infrastructure. So that certainly positions you well, makes you a bit more of a wild card to, to really have an even bigger role than maybe what, you know, some of the names that kind of are, are sort of synonymous all the time with ai. Um, let's talk a little bit about your, your, your global scale.

I sort of teased out sovereign ai. Um, you guys have a massive sort of modular approach. I think you're at like 200 regions or something at this point.

Uh, and you've done it in a pretty short period of time. You're, you're sort of going zones, regions happened in a very short period of time. How have you kind of approached this?

How have you done it so quickly and what's your sort of strategy to addressing, uh, your regions and zone and how you're picking? Yeah, yeah. I mean, I think, uh, regional availability is a probably one of the most critical parts of, um, number one.

You know, it, it's, it's, it's really a big portion of why we are where we are today. Um, you know, when we started in 2016, in OCI, we could have all the greatest services in the world, but if they're not available where the customers want them, we would've not had a seat at the table, right? So we had a need to build, deploy regions as fast as possible in the locations that our customers wanted for various reasons, whether it's latency and performance, whether it's security, data sovereignty needs, lots of different reasons, right?

And so we invested over a decade's worth of engineering time and actually scaling down our region footprint, right? So we had a need ourselves. We wanted to deploy lots and lots of regions everywhere, right?

Think of a world map and like, there's dots in every single city on the planet, right? Not just big hubs or big regions in short number of like two regions in US and two in Europe. No, we wanted a region, an OCI region in every part of the world, in every city of the, uh, city of the world, right?

So for us, we had to spend a lot of time scaling down our region footprint so we could fit our entire public cloud into a single rack, let's say, as a, as a, as a North Star goal, right? We're very close now, 10 years in, we can do it in three racks, right? We announced dedicated region 25 last October, which basically means a customer can deploy three racks and get our entire public cloud footprint.

Now, of course, they can modularly scale that out. Um, of course we did it for ourselves to deploy more regions, but it just so happens that customers love that and they want to be able to do that themselves as well for themselves. So we ended up deploying dedicated region, and then we built on top of that, right?

Because customers were like, well, um, you know, with sort of the geopolitical space that you mentioned, uh, in Europe as an example, a lot of the countries are making these workloads, um, uh, you know, government workloads or mission critical workloads, uh, sovereign, right? So if you're, you know, in Germany and you're running an automotive, you're an automotive company, you cannot run, uh, in a public cloud that is, you know, a, a US company. You have to run a local public cloud.

Well, we have allway. So we could take the DRCC footprint, give a partner resell rights, and they can be a wholly owned subsidiary inside that country and provide a cloud, right? So truly, um, it's, it's actually opened up a lot of different avenues from us, from, um, you know, not just DRCC alloy, our own region footprint.

Um, edge no longer has to be scaled down to just a few services. Why can't edge be the whole cloud as an example, right? Why does, why, why can't you deploy everything in the edge?

As an example? Um, you know, also multi-cloud, right? There are some regions where we may not have a parent site where our partners are, well, that's great, now we can just just deploy a very quick region.

So it, it actually opens up a lot of different avenues, um, not just sovereign security, data sovereignty, but also just the, the, the reach of all of our cloud. Well, we have a, just a couple minutes left. I wanna just stay on this distributed cloud strategy a bit more, um, to kind of give some nuance out there, because one of the things that I, I've been tracking with is that, you know, you have dedicated regions, but you've also now announced alloy, and I think you were alluding a bit to that in your last answer, but, um, kind of would love for you to just sort of break that down, you know, being a little, how is that different?

Um, maybe double click into that then say the traditional managed service offering. Yeah, I mean, sort of the idea behind Alloy Daniel was that, you know, if you look at, um, similar cloud is becoming a really important part of, uh, critical infrastructure in general, right? Similar to sort of telecommunications or utilities.

Um, right? These are very regulated markets where there's three or four providers in a single geography, right? Like, you know, if you look at, um, uh, in the us like we have Verizon, at and t T-Mobile, right?

And they're regulated markets or utilities or energy. I think in the future, we truly believe cloud infrastructure is going to be critical infrastructure, right? Like your, your, uh, core banking services, your, you know, uh, major emergency systems are all running on the cloud, right?

That is critical infrastructure that needs to be regulated. And so what's gonna happen is you are gonna have countries across the globe that are, that are gonna wanna regulate this, that are gonna want to foster competition, um, and, and, and somehow control it and, and have the right boundaries. And so really what we found is, you know, every country's gonna end up having three or four cloud providers, right?

Um, and today, well, at least there wasn't a model before Alloy where you could just buy a cloud and run it yourself. There was, you know, people tried OpenStack, people tried other different types of things, but it didn't really work out, right? Um, you need a commercial capability of an existing big four cloud provider like us, but you need to be able to have your people run it, operate it, manage it, but you also need tools around the cloud to be able to manage your business.

How do you price it? How do you sell it? How do you support it?

How do you operate it, right? So what we do is we take, uh, we take our DRCC footprint, we give you resell rights, and then we add other parts of our business, which is why it makes sense for us as an application company. We bundle fusion with it.

So you can price your products, you can discount them, you can invoice your customers, you have a support system. So you essentially in a box, get a cloud that you can manage, run, and operate and sell to your customers. And the value that you're providing is the, is the fact that essentially, you know, you as a customer, the end customers mandated that you must use a local cloud provider, right?

Um, so that's, you know, that's been a, that's been a big success. You know, we have, we have I think over 20, um, don't quote me in the number, but I think we have over two dozen alloy, uh, customers now. Um, some are live, some are not yet, we announced this a couple of years ago.

You know, for example, one example is Fujitsu NRI actually is a great example where they're, they started with A-D-R-C-C for their internal workloads and then moved on to Alloy to build their financial applications and offer it to the government space in Japan, right? So, um, it's a, it's a, uh, it's a great stepping stone for, for those customers in partners. Well, Koran, that's really great progress, and I see, you know, I've mentioned a few times with, with the AI shift, with Sovereign Cloud shift, the ability for them to have sort of all the pieces that an Oracle can offer, but at the same time really own it and manage it, and not trying to build it, uh, and run it like an open source project, um, gives kind of all the tools required.

And I could see that being really, really successful. I do gotta stop there. Um, it's been a lot of fun chatting to you, Kiran.

Um, you know, congratulations on all the progress. It's gonna be great to watch these next few years as the cloud gen three in this terms of the next era, perhaps, um, you know, continues to roll out, continues to scale, and continues to meet the customer, meet the market, meet the world where it is in such an exciting and rapid transition. Kara Batta, SVP Oracle Cloud Infrastructure.

Thanks so much for joining me here at the six five Summit and this cloud Infrastructure Track Spotlight keynote. We'll look forward to seeing you again soon, sending it back to you in the studio.