Making Decisions on Infrastructure in Light of Tariffs – Infrastructure Matters EP78

This is Textron tv. Good morning everyone. Hey guys.

Welcome back to Infrastructure Matters. It's a dynamic week. Um, I think by the time this gets posted, we'll have some additional news that is going on as well.

But first and foremost, Diane, where are you today? Uh, well, I'm, uh, still, uh, just arrived in Europe, so Yeah, you're still there in La Negro. You haven't decided to go someplace else.

Okay. So Yeah, we're probably, yeah, we we're probably gonna head south to Albania, but Well, we, we will hear more about that. Okay, guys, so we've got some, we have to talk about what's going on with the economy and the tariffs and everything else, but we're gonna kind of twist that into looking at, you know, what does it mean for our CIOs and our infrastructure matters, people to make sure we cover that, as well as a couple of other announcements that we have to discuss.

So, um, there is a, um, d Diane has also put out a new CIO playbook on how to, uh, navigate these changing tariffs. So that's gonna be good. So let's start first, um, for those who have been hiding under a rock, um, just let you know that Mr.

President Trump announced some significant tariffs on Wednesday. Um, we've had all kinds of stuff all over. There's a lot of screaming and yelling.

People scream, very concerned, and then others are saying, Hey, everything's fine. Um, and let's hold the horses. So with that, I'm gonna turn over to Diane, kind of like, talk about the infrastructure and where you think this is gonna go.

Well, yes. So the, uh, Trump administration, uh, announced sweeping blanket tariffs, uh, uh, to virtually every country on in the world. There's a few that weren't on there, which, which would be interesting, but not significant for the tech world.

Um, the, um, in addition to, uh, the, uh, flat tariff, uh, depending on what they believe the trade imbalance is, uh, there, they also have, uh, additional high tariffs. So, so Taiwan, for example, and Vietnam and Cambodia, uh, set, uh, which all manufacturer tech components, uh, have, uh, especially high double digit tariffs. And that has dramatic implications for those who buy it infrastructure.

Those who build data centers, those who create the, uh, the new big AI clusters that, uh, do the training and the inference. Um, these are very costly capital investments, and they're costs are now going to skyrocket, uh, especially since China just announced reciprocal tariffs this morning. Um, and, uh, and that's not what the, the administration's hoping for.

They're hoping that most nations and, and maybe some co companies, uh, large companies, will negotiate tar, uh, the individual tariff, uh, exceptions to this. So the belief I, uh, from the administration, from my understanding is that they're, they're announcing these tariffs, but they're anticipating negotiating them rapidly down. Now, that hasn't really happened yet, uh, but right now, the market's feeling the forces, uh, uh, of the, of the tariffs on, uh, you know, driving it infrastructure 40, 50% higher overnight, um, you know, the, the iPhone market may, may be dramatically impacted in, in the coming weeks, uh, if no, no negotiations are, are, are being done.

And, and so for those who are building, um, private cloud, those are BU building public cloud, uh, and buy and have to buy lots of hardware to, uh, to, to fuel their ai, um, dreams and aspirations. This is a, a, a major issue, um, and it's gonna dramatically change budgets. And, and that's not what the, you know, the CFO uh, uh, wants to see.

Um, budgets are set in it year by year, and, uh, they have to go back and redo, um, your IT budgets. In a highly fluctuating market, it's gonna be very difficult. Uh, so it's gonna have impact to vendors, um, to, uh, suppliers, uh, to all companies who are building any kind of IT infrastructure.

And, and in fact, this, this is showing, however, one of the, the, the bright, um, uh, cl uh, the, the, the good signs of if you've been investing in, um, public cloud, you've kind of arbitrage, you've exported your, your risk to that to, to your cloud provider. Um, now your cost might go, might go up, your cloud bill might increase, but you're not gonna have the CapEx, uh, hit that the, that the others are gonna have. Uh, what are you guys seeing as in terms of the impact?

Well, a couple things, and I haven't dug into the gory details on the tariffs, but from what I understand and in listening to some of the news that's going on, is that while, you know, we saw the poster that was, you know, that was shown saying, you know, this country, this number, this number, and that's how it's being reported, is across the board kind of tariff watches that may not be particularly what is happening, because there are items that are not, like, for instance, um, NAFTA covers quite a bit of items that are coming over the border from Mexico and from Canada, and this has to do with car parts. So there's a bunch of, you know, everybody's, you know, screaming about the car kind of thing, but yet we still have some of these other rulings that are in place. Um, as well as, as you start to hear and listen to the people peel back some of these pieces, it's not necessarily everything.

So you hear a 34% number, that's not necessarily 34% across the board on everything that's walking out the door. There's also a timing issue, um, in terms of when these things fall into place about how orders are placed, when the orders are placed, when they're shipped, and when they're coming across. I personally, because of something that I have ordered that's coming from overseas, and that's fairly pricey, and we've, you know, gone back and said, okay, so what's the situation on this?

How do we handle this? And what's going on? Saying, this is not impacted from what we understand, this is coming over.

So, so I think there's a bunch of things that we need to peel back and also understand about the tariffs. At the same time we hear that, you know, the craziness, not the craziness, the, the, the realistic kind of understanding about what the long-term impact on this. Um, and, and something I'll pass over to you guys.

I mean, yesterday as I was watching the market, and I think a lot of people that have got investments are watching this market and seeing what's happening there. Um, not that I was taking any action or anything, but like Intel, Intel was the few companies that closed yesterday up. Now, today is a different situation, so whatever is going going on, but this is part of TSMC, so why, why isn't, why was Intel up yesterday?

Well, who, what were people thinking about that? Why isn't Intel getting slam, like everybody else's? Um, against, you know, somebody knows something that I don't about what are the pieces here?

So I'll pass it over to somebody to comment on that. Yeah, so the TFS TSMC is pretty interesting. Taiwan got it polluted in semiconductors.

So Intel, uh, manufacturers overseas in Taiwan. So obviously what they make semiconductors aren't impacted. So you saw winners, you know, in the retail space, we saw RH on as the, the CEO of Rh, uh, for those of us remember RH as Restoration Hardware.

Mm-hmm. You can tell that. Uh, I have a wife that likes furniture.

The, uh, the impact was quite great, and he ended up, uh, having an explicit as he's having his analyst session with, uh, on the phone. The impact is uncertain, is, I think the, the problem is the impact is uncertain. And businesses don't like uncertainty.

CIOs, CTOs don't like uncertainty and the complexity. I think one of the things that most of us were hoping for in this administration was, uh, less regulation, simple, uh, cutting through the red tape to, uh, be able to accelerate our business concerns. And this is not cutting through the red tape.

This is adding more red tape. This is adding more complexity, and we're seeing the market reaction. We're seeing CIOs react.

We're seeing the price of, uh, hardware equipment react to that price, to that, uh, to these concerns. And I think, uh, as enterprise customers, for people who care about infrastructure, whether you're building clouds, whether you're trying to plan for your next ai, uh, innovation and waring off born in ai, uh, companies and their disruption to your industry, this is yet another, uh, unnecessary, in my opinion, unnecessary distraction from trying to actually get work done. I, I don't know what my talent pool is gonna be for ai, uh, and data scientists.

I don't know what, uh, Nvidia, the impact of the next NVIDIA chip will do. Now, on top of that, when you com combine those pieces, the network gear from Cisco, who doesn't manufacture in the us, the, uh, Dell hardware, who got hammered, uh, uh, recently, HPE, all of these, uh, the big cooling systems, the liquid cooling systems to keep AI systems cool. My hard drives from companies like solid dyne, like, okay, semiconductors aren't impacted, but the things that you put them in are, so this uncertainty, this lack of, uh, clarity is I think one of the biggest problems that most, uh, most infrastructure leaders are reacting to.

And one of the things that I, and I don't, I don't wanna get into, if I'm getting too far into politics telling you to shut up, and it, 'cause I'm not, I'm trying to stay away from that. The, the depart head of the department of, of, uh, commerce is Howard Lutnick. He was the CEO of Cantor.

Fitzgerald has been CEO since the middle of the nineties. I mean, we're not like talking a guy that doesn't understand markets. He has a little bit of market going on.

You know, our department, our head of US treasury is Bessner, you know, what was he New York Stock Exchange? Is that what, where he came from? I mean, you're talking about some really heavy duty guys that understand market dynamics in market marketplace.

And somewhere along the lines, I believe in this, you know, explosion of what we're going through, there are com these, these guys aren't, these guys know how to weather through storms. And they knew that this was going to be a storm. Um, you know, Howard was on the news all day yesterday on every channel that there possibly was to talk about what, what was going on and, and what was happening.

So I have a, a faith that, we'll, you know, we will get to a level, level space here, um, or belief that somewhere along the will level out, um, we will have higher prices over the long haul. I am sure we will. So we have to plan for that, and particularly into the IT organization, um, because that's so much of what's coming from overseas.

But anyway, if I've gotten, if I've gotten too much into the government, I don't believe so. I mean, the, what's what's going part of the intent is, and this is all just my, our, our, you know, understanding from the outside, you don't actually know what the, the real intent here is. But one of the stated intent is to bring manufacturing back to the United States.

Um, and so if, if we, you know, manufacturing, especially in high tech devices, has been hollowed out in the United States. And so the argument is, um, you know, it, it, it's good. It could be good for us to bring it back.

And there has been a lot of, you know, know since, uh, Trump took off. There's been a lot of talk about, about, uh, reshoring, a lot of, um, uh, uh, previously, uh, overseas activity from, you know, from the big chip makers, including Intel and building foundries. Uh, it's, I'm not clear the bureaucracy's been cleared.

It, it takes 10 years to build a fab that it takes two years to do in Taiwan here, uh, if you use, use existing processes. So that's the red tape that Keith was talking about, that needs to be taken away from these tariffs to actually, uh, have the effect that, that they need to have, uh, uh, in order for, to bring prices down because you bring some of that manufacturing back. The argument is that, uh, you know, it might, might have much less price impact, uh, and a stronger industry given that things in the world might get a little, a little crazier if anything happens.

If there's a, a blockhead on Taiwan, you know, China has a big exercise surrounding the island again this week, um, and, uh, doing live artillery exercises right off the shore of Taiwan. Uh, and by all accounts, they, they want, they want it back hopefully without, without actual war, they can just blockade them into submission. Um, but all that's potentially coming too.

And so, uh, the, the tariffs, uh, you know, our preemptive strike to bring some of that back before we lose Taiwan. Right? So I think there's some other maneuvers.

There's o other long-term plays, although I think the, the temperament of our current administration, they like to change things a lot. And as Keith said, we, it planners, um, it want predictability in the business cycle. Private clouds, uh, had a lift recently because you can predict your prices out for, for up, up to four to five years.

'cause that's how long you can cut your assets, um, and you know that it's winning because you, they know what their cost is gonna be for four or five years, and that allows them to do long-term strategic business planning. Right now, nobody can do that with things that are happening. And that's the big challenge.

Yeah. And I think the, the one last point on this is the uncertainty part. The Trump administration, I hope, will only be in power for another four years.

That's the, that I think that I said it was stay outta politics. The, the, the, I don't mind, I don't personally mind diving into politics, but this, this is the point of the uncertainty. The Trump administration is only going to be in power for another three and a half years, three to four years.

The, so Trump can only directly, uh, control terrorists for another three to four years. These long-term investments, if I'm, whether I'm a IT infrastructure manager trying to figure out my investment for the next five to six years, or if I'm a manufacturer trying to determine whether or not to bring onshoring back, I'm looking at the 10, you know, these 10 year timeframes, and the first three to four years is uncertain. So am I going to make a hundred million a billion dollar or $10 billion bet over the next 10 years based on, uh, the temperaments of the next three to four years?

So this is the uncertainty that, uh, that, that's been injected into the reality of managing infrastructure. Do, do you, how do you assess risk? This is for folks much smarter than me.

Do we need risk assessors to help understand what is my true cost for continuing to my existing procurement strategy, changing that, or changing my, uh, overall delivery methodology, uh, as we get more data from folks like Diane, CIO network, et cetera, we'll get more insights and we'll be able to share that. So Diane, you published a playbook. Um, is this generally available to folks or, uh, On, it's, uh, yeah.

We'll, we'll probably publish it as more formal research with additional details, but we wanted to get something quickly into it. Leader's hands give, give them an idea of what they can start to do to assess where, what their most tariff impacted items are going to be, so they can find those in their budget and, and start mitigating, uh, what those are. Um, so we've got basically eight steps that they can take, uh, to, uh, create more dynamically manage their IT budget, uh, understand their risk, uh, manage the risk.

And this is actually an opportunity for the CIO to really service the rest of the business because since it already has to build a capability, um, to manage, to monitor their tariff impact. So as tariffs change, which they'll change on a daily basis as countries negotiate mm-hmm. Uh, deals with, uh, uh, with the United States, um, things will change.

And then you'll, you'll want to change your purchasing. Uh, and so it's gonna be, it's gonna be a huge headache. It's a, it's a very much unneeded step, but, uh, the CIO can provide this, this, this tariff management dashboard to the rest of the business.

'cause the business purchases all sorts of things. Mm-hmm. Uh, not just it, uh, uh, to run their business.

And, and many of those are for overseas, and they're gonna have to manage their tariffs as well. So, uh, the CIO can create kind of this o operation center for tariff management, um, using the, the analytics tools and AI capabilities they've, they've built, connected to ERP and start to really, you know, get, get their, get their arms around this and maybe even do some predictive analysis. Yeah.

Awesome. So let's go on to another topic, which a little bit related, um, Intel, um, Intel held a event this week called Intel Vision. The new CEO spoke there.

Um, uh, Daniel Newman, I believe was in attendance as well as some other people. But, um, I think the big news there, some of the big news there was the IBM Cloud and GOUTY three, some substantial numbers that were put up there. Um, and I'm gonna, again, I'm gonna tout my, my good buddies over there at Signal six five who did a quite a bit of testing over here.

And, uh, but, um, you were gonna kind of Keith, uh, expound on that, what was going on. Yeah. So let's, let's give, uh, our good friends over at Signal 65.

They're due because they were quoted on stage, essentially GDI three. You're looking at performance that is similar to, if not surpasses, H one hundreds. Well, you might be thinking, okay, Reen has been announced, Blackwell will be, will be shipping shortly.

And we, on this very pro program, talked about the significance or the insignificance of H one hundreds moving forward. The basic, uh, the, the basic reality is that people are doing work, uh, uh, that's significant. The gov, the US government shared this stage with, uh, Intel talking about how they're using gaudy three.

Uh, and our good friends at kaa, Luke Norris was on the stage demoing, uh, much of what the government has been doing with weather data and being able to correlate and build applications that make changes and deliver value to, uh, the mission of the US government around weather. So I think this was a pretty big win for Intel and IBM because the announcement was that Gotti three will be available in IBM Cloud, which mitigates some of this investment risk. And I'm not just talking about tariffs, but around AI infrastructure and the complexity of managing AI infrastructure.

We're outsourcing a lot of that to the cloud providers. And the bus big announcement was IBM was, uh, their, their go-to-market partner for cloud. Now, I, I have a question for you, Keith.

Um, you know, so we see these trips like Gaudi and they have these impressive numbers, but unless they are dramatically cheaper, um, or dramatically faster, uh, how are they gonna prime market share away from Nvidia? Who's, who's the, you know, the 90% market share elephant in the room, uh, 'cause of Cuda, uh, and you know, when developers wanna build on whether, yeah, So there's two, there's two attacks. One, uh, slightly cheaper.

They are, I think they are, uh, at least from a list price perspective, cheaper than H one hundreds and availability you can get them. So the, uh, this ability to actually get the infrastructure to actually do work is one of the biggest challenges. And this is why the go to market announcement with IBM is so important.

Uh, IBM, uh, I don't think it's by mistake. Uh, A-A-W-S-G-C-P and Azure are big intel partners, but IBM caters to the enterprise. And Diane, you know, how important it is to have a cloud provider that you can say, you know what, stop, don't update my, don't update my, uh, public cloud department because of my change control and because of my version of SAP that I will refuse to move from.

We're, we're just, we can't move at that speed. So I think, uh, Intel has found probably the right partner next to OCI, maybe to, to help cater to the market that they're Target. Well, I've done a lot of research for years about what's switching costs.

Uh, this is what really holds it back. Uh, and so this, uh, this, there's a lot of research that shows that, uh, there's this thing called the nine x rule that says that to switch a wholeheartedly to a, you know, a whole new way of doing something or a whole new technology and replace, replace what you're doing now with that, uh, it has to be nine x, um, a cheaper or faster or better or something in order, in order to, to justify the switching costs. You have to rewrite your code.

You have to, uh, buy, buy all that infrastructure. The switching costs are usually high in, in, in technology. Uh, and so, uh, I, but I think one of the switching costs here that are in, in favor of Gaudi is that they're actually available.

You can get your hands on them where you can't get your hands. You want, you wanna get black. Well, good luck.

So I think this is, this is, but the real angle I think that, that, uh, Intel can use, uh, to get some AI on ramp. Yeah, and the other thing we heard is that there's going to bes, you know, no, no surprise here about the investments going into the gouty space. Um, and, uh, we had seen not enough investment over there on that side of the house with Intel, but with this new, um, new focus and where they're going, um, this is going to, you know, they're, they're putting in a lot of engineering on the development and growth, and I believe also on the software side of the house.

So I think we can see some advancements there. Alrighty. So let me go on to the next one.

Um, wait, next one is, um, there's a new data management company that you've, um, Diane, you've been watching Hydro Hydraulics, h hydraulic Hydraulics, and that pelts like hydraulics that we know hydraulics, it's spelled H-Y-D-R-O-L-I-S. That's why it's remarkable against it. Anyway, so, so data management is the hot topic because, uh, unless you've been getting your data story, uh, uh, in shape and inside your organization, you aren't doing ai, uh, on your own data, if you can't bring your data together with, with your other data, uh, you can't access it, um, uh, you know, in a consistent, uh, you know, way, you, you can't use it in all of these systems that are providing, you know, so much potential value.

Uh, and so hydraulics, uh, provides a data management solution for structured and unstructured data, uh, and they do it in a time series way, so you can see what the data was over, over time, which is very valuable for spotting trends, doing predictive anything, whatever is, is to understand not just what data you have today, but what your data was over, over a period of time. Um, and there, there are big claim to fame, however, is be able to do all of this with petabytes of data and very at, very high performance. So, uh, they just received $80 million.

Uh, so one of the biggest investments ever in a data management startup because, uh, that's the gateway to analytics and ai. Uh, you can't do it unless you get your data story. And, and hydraulics can take everything, all your structured data and all your unstructured data and put it on the, put it on the, the same playing field so that you can actually feed it to your foundation models, lms, action models, whatever it's you're gonna be doing.

And, uh, so, So this is more of a metadata analysis kind of, um, data management function that you then can, um, that's, that's, that, um, scans the entire and then indexes it or however they do it, um, from a data management standpoint, Yes. So, uh, it will, and well as it be both, uh, use both operationally and uh, uh, analytically. So that's been the, one of the, the most difficult things in, uh, if you, if you, if you worked in databases, you know, you use the same database for transactions and the, and the same database for analytics is very, very difficult.

Um, and so they help, they help a bit with that problem. Okay, great. Yeah, that, that's, that has been the promise of platforms like SAP Hana in the past, right?

But not at petabytes. Uh, the limitation is obviously the amount of RAM that you can have in the system, and the ability, I think, again, is the, I think in practice transacting and analytics on the same platform at that scale is really, really difficult. Mm-hmm.

Yeah. Very difficult. Very, very difficult.

Well, that is great. Um, I also was on a call this week with, um, WCA that we were, um, which is a parallel file system. Um, and, um, they've had with a couple of my peers, both, uh, Russ Fellows and Brad Shiman from the, the intelligence team, and kind of parsing apart what they are doing, because what we've seen recently is some new architectures coming out or some additions to the architectures coming from Vast, coming from NetApp, coming from, um, pure in terms of how they're designing the global file systems in order to speed up the, um, training and the inferencing with the engine.

So what they have done is going beyond what we were seeing with this, this caching KV caching capability, um, and actually speeding up, this is with within the memory kind of capability in a new layer of memory. Um, some of you guys know, um, some of the people that are over there, um, at Wicca, um, um, the, uh, one of the founders of NetApp is on the board. Um, Val is on, is actually now working for them, and he's one of the c former CTO at NetApp.

So these guys really understand file systems and understand where this is going. So it's quite exciting to see in my space where I've been playing, um, you know, all this activity that's, you know, going into the data management side of the house, which is more of the management of the metadata and the analysis of the data as well as how, what we're doing to speed up the processes that are over here. So, um, it's great to see the investment also in that market because it a very dynamic market.

Yeah, I think it's important to note that Val has been, he spent the past maybe three years in AI and, uh, focusing specifically on this, uh, type of challenge within the data space. So I'm, I'm excited to see what comes of it. Yeah, yeah.

He's now actually taken a position with them as opposed to an advisory, which is what he's been doing. So that was, it was great to see, um, you know, Russ fellows who, who understands the guts of the engineering part of the file system and him go after it and say, okay, the, these numbers are extremely, um, in interesting in terms of the, the proof points of what they're able to do, um, at, with the huge amount of tokens in the system. So, Alrighty, guys, anything else?

We got to bring up this scene, this af this morning, this afternoon, whatever time you're listening to this thing. Just one LA uh, quick piece. Uh, on previous episodes I've talked about my theory about we're heading towards super intelligence or AI models that are smarter than any human.

5 has PA convincingly passed the Turing test in a, uh, in a, in a, in a, in a widely, uh, um, followed test. So it's, it's the three party Turing test, uh, at a, uh, it passes at 73% of the time with humans. Um, that is the highest, uh, any any computer has ever gotten.

Uh, and, and that, and so now we, you know, basically AI can, can pass as a human to most people most of the time. So that's very, so What is de Turing test testing? It, it, it is a test to see if you're, you talk to an artificial intelligence, if you can tell if it's an AI or a human, you get, uh, a certain period of time, uh, you know, uh, depends on the test, right?

So five minutes, 15 minutes, and then at the end they're asked, are you talking to a human or are you talking to an ai? 5 can, uh, uh, passes as a human 73% of the time and a new research study. 5.

The, the I'm getting closer and closer to being replaced. Yeah, exactly. We are, Is that the, you know, but interesting data point, uh, this, I'm, I'm launching whole few one next week or so, a new a hundred days of AI is I'm going to teach myself how to code I that.

Yeah. You saw that on online, for those of you that haven't Yeah. Teach you.

So I teach myself how to code with, uh, AI as my pair coder slash teacher. So I'm gonna Do, are you using Cursor or that's that's, that's my next, uh, thing to try out. I I will, you know what?

It's interesting. I did a experince man, my buddy has a 10-year-old application that was written in England js, and I said, you know, what would it take to modernize this app because JS is, is, uh, in of life. And, uh, I took it, uh, I asked Cursor and it told me it would take me, with my level of knowledge, it would take me 12 to 16 months to rewrite this application.

And it made me realize that, uh, we're so far from using this tool to replace actual workers. It does speed us up, but it, uh, so I want to find out exactly, I wanna measure this for CIOs and CTOs. What does it really mean to augment your staff with AI capabilities?

So it's gonna be an interesting Experiment. We'll have to watch that. Tha that's gonna be, so if It's passing at 73%, does that mean it can lie as well as we can at Yes.

Well, yeah, because it has to pretend you have to prompt it to pretend to be human. It's lying the whole time. Actually, if you, you know, technically if you wanna argue that, That, oh, you only have to be 50%, right?

So maybe we could translate that to an age, you know, like the 3-year-old fibbing, the 5-year-old fibbing, the 15-year-old fibbing, the 18-year-old. Yes. A new test.

Alright guys, thank you very much for tuning in. We will post a link to the CIO playbook in the notes. So go and grab that.

Um, and then don't forget the star C, the star the pound, CIO insight, um, Twitter or X mm-hmm. That ion is running. Is there a topic this week that you're gonna pop Up there?

Yeah, so CIO chat's gonna be, uh, on, um, uh, what, you know, I, we picked it before tariff, so it's not gonna be about, uh, uh, about tariff. Uh, we're gonna talk about, um, uh, the ideal, uh, it reorg now that we're in a new, a new age of ai. How would you reorganize it today?

So that's the next topic. Okay. And what time do you think we'll see that pop up and start to, uh, when will that be?

That's, it's, it's, uh, every Thursday at 2:00 PM Eastern time. Great. So hope to see you guys there.

Alright. Come in and join in the chatter with, uh, Diane and um, and then listen to us next week as well. We will look forward to talking to you.

Thank you.