AI Brakes, VMware Licensing Scrutiny and Microsoft Cloud Strain

Welcome to Techstrong gang on another Manic Monday. So just when you thought the news cycle would cool a bit as we approach the fall season, things are only heating up. I'm John Swartz in Silicon Valley.

I'm filling in for Alan Schimmel and Mike Zard, who are jetting at this very moment to Denver for the Splunk Conference. Today, my God, we have a pretty heavy-handed series of topics. So we have this heavyweight panel, which I know is going to be very opinionated about all these topics.

And we're going to start in no particular order with the great Steven Fossett. Yes. Hey, Steven.

How are you? You know, I'm doing really great, and I'm glad to be here in the place of Alan Schimmel today. You do that more often, and your impersonation's getting better and better.

I've never seen- I spent last week with Alan, and so I am Alan-ing. " And I'm like: "I don't know. " Alan-ing, it's a great turn of phrase.

Thanks, Steven. Also with us is Tom Hollingsworth, who has quite a few pointed opinions, I'm sure, about the first topic. Hi, Tom.

I'm here to cause confusion and delay, which is basically why you invite me on this podcast. Awesome. And last and certainly the best, we saved the best for last, and I say that with all due humility for everyone, because everybody here is great, I think.

It's Chris Laster in Canada. Hi, Chris. Hey, John, and Steven, right now it's interesting because we do as people, we talk like each other.

We spend a lot of time together, we start to talk together, which- Yeah, I know, right. We start to mirror each other when we're in conversations. Right.

And AI companions do the same thing. People are finding this out with just regular little chat accounts. " There's a lot to it.

Oh, wow. This is going to lead to some identity crises, I believe, but so be it. So we're going to start off with, I think the story that is permeating all news coverage now, and you're looking at it from every conceivable angle.

We usually, before our shows, do three topics that we list in a show rundown. And for this one, we had eight topics that were linked, eight stories that were linked. So I'm going to try to condense it down and make some sort of sense of all this.

So basically, and I'll just be blunt about it, all hell broke loose this weekend, and this essentially is a three-way scrum over AI development, especially frontier models. We have tech executives like Dario Amodei, Elon Musk, and Sam Altman who are pleading to tap the brakes on frontier model development and strengthen independent safety oversight, which I find to be very interesting timing, and I'm sure Tom does, too. At the same time, we had President Trump over the weekend and even today embracing no AI guardrails.

He's dismissing these warnings of AI killing humanity as very negative forces. He's saying damn the torpedoes, go full force ahead, whatever the consequences because we have to beat China in the AI arms race. And finally, we have the investment community, which is largely downplaying the immediate impact of the safety debate on the broader AI build-out.

I'm going to mention one of the analysts who's very gung-ho towards Silicon Valley, and it's particularly towards Anthropic and OpenAI. That's the famous, or some might say infamous, Dan Ives, who emphasizes that the ongoing discussions about safety are unlikely to derail the trillions of dollars in AI-related capital expenditures projected over the coming years. So Steven, what is going on here?

We have all these conflicting comments happening at the same time. What could possibly be, I don't know, at stake here? Well, it's interesting that I feel like this strange time that we live in where you have the three most prominent AI CEOs warning of AI extinction.

You have a high-profile, I don't know, defector is kind of what people are calling him out there on television and radio and basically every medium. Are you referring to Jacob Coxen? Of course, yeah.

Right. Is saying that basically that everybody in Silicon Valley believes that there's a greater than 10% chance that everything they're working on will kill us all. I think that people are probably very confused about this.

" You didn't hear Henry Ford warning about human-level extinction if we adopt moving assembly lines. Well, I guess you did hear Tesla saying that his stuff could kill us all, but that's a different thing. But the point is, how do you make sense of this?

And then Washington. Washington's trying to figure it out as well, along with literally every other politician. The problem is, I think that it's another example of where the Silicon Valley hive mind has gotten out ahead of or behind or off to the side or in an alternate dimension from everyone else on Earth.

And they all know what all this means. They all know that this idea that there could be an extinction level event, and therefore, we need to make sure that we're the ones that develop the good AI before the bad AI comes and kills us all. So they know that this messaging is pro-AI, not anti-AI.

But the rest of the world can be forgiven for saying, how could that possibly be a positive message? Because it sounds terrible. It sounds absolutely awful.

I was actually very pleased, by the way, that one of the prominent AI voices yesterday quoted my supervillain cosplay marketing in a tweet. Thank you very much. But the point is- So who was that, by the way?

I think that it was Zittron, but it might have been- Oh, Ed Zittron, right, who's been a frequent- Yeah, or Gary Marcus, one of those guys. Anyway, I was excited because I'm just a little nothing here. But that's what it is, right?

It's cosplay. " And that's the important thing, it's the turn, the unless, right? And that's where we're headed, right?

We need to get to the unless. And frankly, there was a big unless in the three AI CEOs saying we need to tap the brakes on frontier model development, and that is unless China gets out ahead of us, in which case all hell breaks loose. I think that's the whole game.

I think that's the whole message. It is the China-Washington messaging, not the we need to tap the brakes on this thing. But who can say?

I mean, it's all tea leaves, right? Right. Hey, Tom, I was going to ask you really quickly, too, because we were mentioning this briefly in the green room.

So these guys were full force ahead, the AI leaders. They were leapfrogging each other's models. They were hellbent on establishing some sort of edge, and all of a sudden now they want to be cautious and want to slow things.

What changed their minds? The fact that they realize that there's a very good possibility that one of them will succeed, and the other two don't like that. This is the standoff at the end of "The Good, the Bad and the Ugly," right?

They're all pointing guns at each other, and they know that the first one that blinks is the one that gets knocked out of the race. So how do they fix it? Somebody has to come in here and tell us all that we have to stop or slow down, right?

There was this great TikTok I saw yesterday where this guy plays different parts, and he goes: "Yes, we're going to slow down AI development. Yes, we totally are going to slow it down. Why are you guys winking at the camera like that?

" They don't want to have to stop, because if they take their foot off the gas for a second, their investors are going to bail on them, right? We've seen this over and over again. But if an external force requires them to slow down, "Hey, guys, there was nothing we could do about it.

" But they ultimately aren't going to stop. Kind of like what Steven said, this is about optics. Mm-hmm.

If the government tells us that we have to stop, and then another entity, China, pulls ahead, well, now you have to tell us to speed up again. And now we have the full blessing of the United States government to race to the inevitable end of our death cult. Because when you hear about this crap, and they keep talking about how AI is going to cure cancer and teach us how to raise babies, but there's also a chance it could blow up the world and torture us for all of eternity, that kind of sounds an awful lot like all of the stuff that I've been reading from Andreessen for years, where it's like, if you stand in the way of AI, and you prevent it from coming to fruition, you are a war criminal that must be hung from the tallest yardarm.

Or maybe I just want you idiots to put some guardrails in place on all of these models so that it doesn't erase the United States National Health database and cause people that need insulin to die. Like come on, guys. I'm not asking for you to bring about the basilisk and have it torture us for all eternity.

I just want you to put some common sense controls in place. It's interesting you mention that. That is a great irony.

We're all told that AI is going to help us live longer, yet at the same time, we're being told it's going to kill us too within the next few years. I guess the general public, their heads must be spinning. " I'm not going to watch it anytime soon, but they did have a segment that I read about AI and its great danger to all of us.

And I'm wondering, Chris, this is almost kind of a unifying theme throughout this show. Is this just a failure of institutions or confusion or the breakdown of institutions? Well, to that point, yes.

Infrastructure or advancements, technology, whatnot, are outrunning institutions, and that's across the board in lots and lots of ways. But I've got to say, I'm in that oldest Gen X-er crowd, but I grew up with older people and older kids, the last of the boomers, and God bless all your hearts, I have heard the end of the world coming every second of my life. Right?

I thought you were dead, Chris. Oh, Lord. Yeah.

Right. 54 billion people, and 100 million Americans will starve to death in the 1970s, and, and, and. But to the meta topic on this one, so it's interesting that, as you've all said, some senior people in the language model development space are saying all these things for whatever reason.

I would posit that they kind of don't understand how their own systems work, but that's a longer topic, right? But it's not likely to be the end of the world. If it is, then we've got the catalog.

We fire all the nuclear weapons at once or something. Right? But short of that, pragmatically, how are our systems really going to kill us all?

We've thought all through this, believe me, all my life, and it just doesn't work that way. Right? Will there be catastrophes and problems?

Maybe, yes. Everybody on this screen is old enough. We lived through COVID, right?

We can talk about all the friends I lost there, wars, friends of mine have died, and that's life. And we're all getting older and we'll die too. But that's not the end of the world.

I think that given the-- Look, I'll fill in for Alan to get some political outrage, but the US federal government has stated today from the chief individual that there will never be any AI guardrails. Right? So the US apparently, if the federal government has a voice, we'll do the experiment with no guardrails.

We'll see what happens. Does that cause power blackouts and other problems and death? And maybe.

That's not the end of the world. That's not exterminating the human race. Everybody needs to calm down a little bit and realize that life is full of consequences, and we're dealing with a bunch.

There's kind of this pivot towards security. I'm wondering, too, about the timing of it, especially the last couple of months. I'm wondering, beyond the IPO, is it also tied into the, I guess the disclosure that, yes, there are problems with RubyGems, Hugging Face, there are incidents that we can't ignore.

Is it acceleration of that, or do we blame AI agents? I'm just wondering, Tom, were there any other extenuating circumstances that have kind of accelerated this kind of fear-mongering? Well, I think the most realistic example for people out there is the fact that they created models that can effectively do exploit chaining and do discovery very, very fast.

So the public face of the AI paranoia is that AI can hack anything at any time. Right? Like the people use their phones or their laptops, and they kind of mess around with tools that don't really write well, or they don't know that you're not supposed to put Elmer's glue on pizza or whatever else.

Right? But when you hear, "Oh, well, now it can hack my bank records because it knows how to get into a bank system that's 12 years old," that's a problem. And they want that fixed.

They want to make sure that there's no way to do that. And so what I propose as a way to combat what Chris was talking about, how people say there should be no regulations. Cool.

I want you to go out and find the tax returns for all politicians for the last 20 years, and I want you to publish them publicly on the internet. I promise you we will have AI regulation tomorrow if that happens, because as we know, there are no guardrails unless it impacts me. There are no rules unless it's something that embarrasses me.

Like that guy who may or may not have been logging the SpaceX AI CEO's private jet flights. Boy, once he did that, now we've got a problem, and there needs to be regulation in place to prevent people from discovering things I don't want you to know. That's the genie in the bottle problem with AI.

" And that's how we get to that point. If you want this to be open to the world and you don't want any guardrails in place, then you have to submit to the AI, too. We need to know all of your investments.

We need to know all of the things that you've been doing. We want all of your private journals leaked onto the internet to know how you really feel about all of this, and that's the only way we can be sure. I want Steven to have the last word, but before you do, Steven, I would be remiss if I did not ask Tom.

" So in that three-way shootout, who's the Clint Eastwood character? Which gun would he- At this point, I don't even know. I think they're all to go because I want to shoot all of them.

I never thought I'd ever hear that word on this show. Steven, I'm going to give you the floor for the last word on this. Well, thank you.

And I would just say that, and this is going to completely fall on deaf ears because this is completely the wrong audience, but I wish that people would just take all of this with a bit more of a grain of salt and understand that they're being played by the companies, by the AI agents, by the politicians, by everyone. And I don't want to sound too much like, I don't know, a Democrat, but there's a central position here that isn't so crazy and extreme, right? We have to understand that, yes, there are potentially catastrophic outcomes, but they're fairly unlikely, at least in the short term.

We have to understand that, yes, theoretically, we can make advances in medicine and materials and computing and sociology and so on with AI, but those are also fairly far out and probably fairly unlikely to be achieved. And yes, when a CEO comes to you and says, "Hey, I'm developing this new thing that might, I don't know, cause a black hole to emerge in San Jose," maybe that's true, but maybe they're just saying it in order to get publicity. Now, it's really impractical to think that people in Washington are going to be smart enough to get that message because frankly, they haven't shown that otherwise, and again, not just this particular group of people.

The law tends to follow very slowly after reality and yet tends to aggressively try to address problems that they don't yet understand, and I think that's kind of what we're seeing here. And so they're being played, too. So we're all being played, and that's just how it is.

Just another manic Monday. Wish it was Sunday. That's my fun day.

Thank you, Mr. Bangels. Let's move on to the next topic.

We're going to talk about another big deal that has some cracks in it, and that's the VMware-Broadcom deal. So European Union antitrust regulators are intensifying their scrutiny of Broadcom following its controversial licensing changes made to VMware software. We talked about this before.

The regulators are looking into whether terms of the $61 billion merger may be squeezing enterprise customers and creating an unfair landscape for regional cloud vendors. The inquiry focuses on whether Broadcom is exploiting this vendor lock-in to impose terms that enterprise customers cannot realistically reject or navigate around. And Tom, I know you know this topic really well, and I'm wondering what you think of what's going on in Europe.

We covered this a little bit on the rundown last week. Al Cook was doing a great job of talking about it. I feel like this was the inevitable outcome of what we all thought was going to happen when this deal was announced, right?

We knew that the purpose behind selling Broadcom was so that they could squeeze more profit out of this. We knew that the way to do that was to focus on the largest, most cost-effective customers, the ones that are paying us a lot of money, and then focus on keeping them engaged with the product. And what happened?

Exactly what we knew was going to happen, and exactly, in some cases, the opposite of what we were told. We're not going to cut R&D. We're not going to raise prices.

Well, okay. I can see where that came about. The news that came out last week was that they were expiring support for VDDI, their disk backup solution.

Why? Well, some of the pundits out there think that the reason why is because a lot of companies were using that as a way to export virtual disks for migration off of the platform. " And that's their right as a company.

But it also means that now people have to scramble and fix things, and well, companies that we've worked with quite a bit, like Veeam and Commvault and Cohesity, that use those APIs and things to do backup, are going to have to restructure. But they probably were already looking at that anyway. " I'm curious to see what VMware's response to this is going to be.

Sorry, Broadcom's response to this, because VMware is not a thing anymore. It is VMware by Broadcom, and the B there is very important. Stephen, you said you have a contrarian opinion?

Is that true? I will take Tom's bait, and I will have a contrary opinion here. So my contrary opinion is that the European Union is barking up absolutely the wrong tree here, and they should just let Broadcom do what it wants to do.

I know that it seems anti-competitive for what they're doing with VMware. It seems anti-competitive to remove VDK, to raise prices, to squeeze customers. But it's actually very pro-competitive what they've been doing because they've been letting customers go, and letting customers go aggressively to other solutions.

And there are lots of other solutions out there. There are lots of competing solutions. I would've certainly said that VMware had a data center monopoly five, six, seven years ago.

Nowadays, uh-uh. I would say that this is an example of a fading monopoly. And frankly, I hate what it's done to the people that were involved in the VMware community and the people who were developing VMware and basically everything that was good about VMware as an environment.

But everything Broadcom has done has basically shaken loose a former monopoly and allowed it to no longer be a monopoly. In a way, I think that's positive for the industry and positive for technology overall. So rather than us being locked into Broadcom, people can now make a choice, and there are other options, very good options out there of where and how they can run enterprise workloads.

And frankly, Broadcom just keeps making it more and more interesting to explore those choices, more and more compelling. So I know that there's some pain short term, and again, it has really hurt some people and their careers because of what Broadcom did. But on the other hand, I think it's good for the industry and good for technology to let people move off.

One company that I know has made a lot of hay of this is Nutanix. They've made that almost their obsession or their one talking point. They've beaten it into the ground.

I'm not sure how successful it's been, but that's been a constant. This has been Nutanix's obsession since the day they were founded. Yeah.

Yeah. Yeah, that's true. I'm just wondering, Chris, do you see this action or perhaps as it goes deeper and deeper, do you think this leads to maybe it's low-hanging fruit in terms of the EU antitrust regulators?

Maybe they're looking at this as a test case and then decide to then expand and look at other companies. Is that potentially a possibility? I would entertain it as a possibility, but I don't honestly think so.

But that's not my domain. But I can't see the logic in doing that. Who in the EU policy structure would make that choice?

But what I like about this segment, though, is portability. And I will play the middle line again. Tom and Stephen both have good points.

You should be able to replace anything in your infrastructure if you can't. You should be able to replace everything in your infrastructure. If you have a dependency, we talk a lot on this show and in the world these days about Anthropic or Claude or OpenAI or whatnot, but everything.

Are you locked in? Is all your information in somebody else's stuff? If they go away, if they change their rules, can you continue to exist?

If not, fix that, which is way easier to do these days than it ever has been. For the same reasons we have all these topics we talk about on shows like this. You can look at your systems and say, "Where are my dependencies?

What do I do? " A vendor is not you. They're not your infrastructure.

If you can't live without them, think about that. One thing, the state of things with the European Commission right now is they've deepened this questionnaire. So they haven't launched a formal antitrust investigation, but if they were to and then found Broadcom at fault, it would carry penalties up to 10% of their global annual revenue, which I find kind of hard to believe.

But nonetheless, it's interesting the juxtaposition of this versus what's going on with what we just talked about in the previous segment. And this question isn't so much about Broadcom, but just the regulatory climate. So in Europe, they're looking at these things deeply.

They're effecting changes as they have with other vendors. Meanwhile, it's kind of like laissez-faire. It's all everyone for themselves in the US.

Does what's happening in Europe have any impact, and I'm being very charitable, any impact whatsoever in influencing US regulators? And maybe I'll throw that to Tom. I think it does.

I've known for a long time that the EU is kind of maybe more of the authoritarian grown-up in the room when it comes to these regulations. A lot of times things that would pass muster in other places won't pass muster in the EU, so it forces the companies to set the lowest denominator that they can, which is can I pass through the EU? You can't really play in both spaces.

A good example would be GDPR regulation. You cannot have two different policies, one for the US and one for Europe. You really have one policy for both because what happens is in the great legal tradition, if I can find even one little opening to go in and get through things, then I'm going to.

And to the point of that is something you brought up, John, that I think most people need to understand. In the US, we tend to fine people based on some archaic system of English common law from the 1200s of it's like half a day's ration of your oxen or something scaled up for where we're at right now. The EU made it real simple.

We're just going to fine you for 10% of your profits because that hurts. If your company made $450 billion last quarter and you fine them $300 million, that's quite literally a rounding error for them. They don't care.

But if you say, "We're going to fine you $60 billion" ... from the last quarter's profits, now suddenly they sit up and pay attention. " Because $60 billion in fines really makes the board of directors and the stockholders sit up and pay attention because they're like, "Wait a minute.

How did you guys blow this? " Right. Yeah, it's funny.

I think about the EU, was it the DMA, the direct market? Yeah. It's Digital Markets Act.

Digital Markets Act. I'm sorry. Thank you for correcting me.

And I think it's had influence. It had somewhat of an influence on Apple and the Apple Store. It- It made some modifications they would never make in the US And that's the thing, right?

Now granted, that's a slightly weird corner case because you can only access the DMA, the alternative app stores, if your location is in Europe. So they really kind of are playing by that rule. But it has forced them to change the way that they provide or they collect revenue from the App Store, right?

They've had to negotiate down for smaller creators, and they've had to create offerings that maybe they aren't going to take 30% of that. It has at least opened that conversation because if you'd have tried to do this five years ago, the answer would've been, "Ha ha, good luck. " Remember, yeah.

And that would be the end of it. Right. Remember the Epic Games trial that went on and on and on?

And although technically they won, on most of the accounts they really didn't. Yeah. Yeah, so it is kind of encouraging.

It does modify behavior in some way. It will be interesting to see where this goes because there have been so much chatter and complaints over this lock-in impact of this merger. So we're beginning to see some sort of traction or some sort of consequence.

So we'll see where it goes. We're actually moving around really fast. I'm going to save some time for the last segment, which I think is really interesting as well as the other two topics, and it has to do with Microsoft and under strain.

So Microsoft is planning a major expansion of its data center network. Its goal is tripling its computing capacity by 2032 to meet constraints in cloud and AI business shortages to overcome them. Their data center capacity is projected to grow from about 12 gigawatts today to more than 38 gigawatts in 2032.

That's a scale that would rival or exceed, I believe, the entire current power consumption of major US states. This is a very interesting plan because it includes facilities that Microsoft owns or leases, but not capacity purchased from CoreWeave and other specialized cloud providers. I'm wondering, given the debate that's going on about data centers, Chris, what did you think about this story and what Microsoft plans to do or hopes to do?

Well, GitHub's woven into this story and we use GitHub a lot, right? And it's a great example. So you have to remember that I and my pack of monkeys are hyperlocal folks, right now.

We build systems from the ground up, and like we were talking about in the last segment, it doesn't mean you can't use platforms is where I'm going with all this. GitHub's a great infrastructure. It's a great stratum.

Strata, right? We look at all the platforms out there as things anyone can use. But again, if you're dependent on that, if your whole memory lives in there and if that goes away, you're dead, then that's probably not a good idea.

And GitHub is a great example because, and a lot of our audience knows this, knows how it works, but it's a code repository, right? Now it's a big huge file system with decent provenance, and you can build structures in there and you can mount them on your own machines. So that's one small example of the way to do it.

If you're going to use GitHub as primary infrastructure, make sure all the important things are replicated somewhere else, so if it goes down as it has, part of this story, you're okay, right? And to the meta question , again centralization, whether it's going to be these individual massive model builders or individual data center providers, and we're going to use all the power on Earth to have massive central compute. I'm biased.

I don't think it works that way, but that doesn't mean we don't need it. So I expect the big players to stagger a little bit as they're trying to adjust populations of employees the size of medium-sized cities as well. And that takes time.

So, it's one of the characteristics of this area is things ramify, right? And Tom, I wanted to throw in "Shockwave Rider" as a random 1975 science fiction. John Brunner, look it up anybody.

The idea that we're going to take away all governments and everything's transparent, we're going to publish everybody's information. That book, I read that in 1975, but it's dystopian in the end. That's not a solution.

There will be large and small systems. AI governance, I know I'm soapboxing, sorry, isn't about not having cybersecurity rules and not having provenance and so forth. AI governance, I don't think that there's a political nation state global organization out there that can define the term.

But we can define privacy and security and all these individual things. We're actually really good at that. And, in as much as Microsoft and Google and OpenAI and these massive companies don't screw up too much, they'll get a slice of that.

But, and Microsoft, what can I say? They're Microsoft, they've been around forever. They'll be okay.

But I wouldn't count my national infrastructure or my company or anybody else's company I'm advising to put all their eggs in that basket for lots of obvious reasons. Yeah, they're going to make a significant investment. They spent $145 billion on capital expenditures in their most recent fiscal quarter, and they expect $50 billion in capital spending for the first quarter of fiscal 2027.

So it's projected that their CapEx for calendar 2026 is going to be $175 billion. And I'm starting to see more scrutiny among Wall Street analysts towards those numbers than their ROI obsession, which lasted up until a few months ago. Now they're looking more at how much companies are spending versus what they're bringing in.

Stephen, is that something that should worry Microsoft investors or do they just plow ahead? Worries me. My question, again and again we hear that the AI data center build-out is akin or analogous to the railroads or the telecom build-out or dot-com.

Yeah, that worries me a lot. First, it worries me because basically every early railroad company went bust in a spectacular explosion of burning money. So that's not great.

Second, it worries me because most of those early railroads proved to be entirely useless and ended up being torn up with only the major trunk lines actually having long-term impact and long-term financial benefits. So again, not really great. And third, because those railroad build-outs made a few people spectacularly wealthy to the point that society basically rose up against them.

Also, not maybe something that we're all hoping for. My question with a lot of the data center build-out is that following along on this same trajectory? Are these data centers actually going to be useful long-term, or is this money that's being set on fire in sort of a 2G, 3G world when it's going to be 5G, if you catch my metaphor, that's going to actually make the money?

And I think that that's actually more likely. So if I was a Microsoft investor or if I was an investor in any of these companies that are building out these data centers, I might have some serious questions about whether the incredible capital allocation that's being spent on hardware is going to lead to long-term market dominance or just plain usefulness, or if this is basically a lot of money that's going to be set on fire to build the groundwork for a future in which there's a profitable data center business. Right.

Now, if there is, maybe I put my money on Microsoft as being one of the companies that might be able to do a good job in that future business and outlast the spending requirements for this. But, it's a big question Tom, one of the things that always comes back to me when I see these numbers associated with data center expansion or any type of AI capital expenditures, I just find some of these numbers hard to believe. Like there's not just a circular financing, but sometimes it seems like they're just almost random numbers that will never be reached, and they're always these nice tidy round numbers.

And maybe it's just me being cynical, but what do you think? It's a target. All they want is for that number to be circulated in the industry so that it looks like they're doing a lot, because ultimately what will happen is they will fall short.

And we know that because unlimited exponential growth is good for a company, but for a cell in your body, it's really bad. And we have a term for that, and I feel like that term should be applied to what's going on here because that's how it feels to me. When you look at the fact that a company that just four years ago was hyper-focused on a very specific kind of growth and having cash flow and all that other stuff, and now the largest tech companies in the world have exhausted their free cash flow stores because we have to build $200 billion worth of data centers in the next six months or we'll be behind.

It's frightening to me how quickly that came up out of nowhere. The term arms race gets thrown around a lot for a lot of things, but this really is what it feels like to me. And any kid who lived through the '70s and the '80s knows exactly what this feels like.

This is Reagan announcing Star Wars. " Every new thing has to be bigger, better, faster, more expensive than the last one, because if it's not, we're falling behind. And then when the Soviet Union fell, as we're still seeing the ramifications of today, some 40 years later, 35 years later, we were so far ahead of where they were that it didn't make any sense.

Look at the development of the F-15 Eagle fighter jet. It was developed to counter a jet that we thought we knew what it did, and when we found out what it actually did, we were so far advanced beyond it that we had no idea. That's what these companies are doing.

They're building more data centers. They're building bigger footprints. They're building more cooling and using more electricity for what?

Because OpenAI needs to run 10% faster than Anthropic? No. At the end of the day, let's say you succeed.

Let's say that all these plans come to fruition and Oracle has their $500 billion build-out and Microsoft has their build-out and whatever. What are we left with? You're left with companies that are not the Windows company, that are not the database company, that are not the online bookstore company.

They are operating data centers for other people to use. Because at the end of the day, you providing all of the storage space, Microsoft isn't really benefiting from this. OpenAI is benefiting from this and whoever else is running their models on Azure.

And people don't think, "Oh, well, that's nice that Microsoft is doing all of this. " Microsoft won't have stuff to use because all of their money will be going into operating the infrastructure for the people who are doing the thing. Mm-hmm.

It's funny you mentioned Star Wars. So years ago, I went to do a story in Colorado Springs, and I ran into a guy, ex-defense department guy who worked on that project, and you know what he told me was that that whole concept was the ultimate head fake or deke. It was done to accelerate spending by the Soviets, and he said in a sense it was part of a kind of political gamesmanship.

There were semblances of an idea, but he said for the most part, it was geopolitics. And sometimes I wonder about, we talk about this, like Steven or Chris, we talk about these data centers. How long does it take to build these things?

By the time they're built, they're not obsolete, but I just wonder if they're ever going to be built in some of these cases. And this is in the whole data center build-out topic, it turns into a circular argument, right? And one of the things I like about this one right now is I'm a big space fan, and SpaceX, Windshot, well, brilliant.

And what we're seeing there is the infrastructure to maintain mass compute. And that infrastructure is not the same sort of physical. It's not real estate and so forth.

It is a system that is also doing all sorts of other things and launching satellites including... And I honestly think the million satellites all doing heavy inference is likely. Now, but the timing, we'll see.

Now compare that to buying a bunch of land, building a data center. Maybe if we're putting together a system for building and maintaining data centers on Earth, maybe that makes some sense. And maybe the big bets are right, or maybe, as you guys are saying, there's other reasons for saying that we're doing these things.

But one of the things I love about the SpaceX conversation is that most of my peers disagree with me. Most of my friends are just really cranky about the data center thing. They don't see any sense in doing it in orbit.

I think that's the only place that makes absolute sense. The rest of it, maybe. We'll see.

All right. Final word before we log out, Steven. Got a comment on data centers in space.

Oh, gosh. Well, it's as if Alan is here because every time I come on, he tries to poke me with data centers in space, and I explode with rage over the fact that none of it makes any sense. There needs to be a 100-fold reduction in the cost, even below Falcon 9, in order to make it make any kind of financial sense to send that kind of mass into space, and it just ain't going to happen.

I'm sorry. It's a nice idea, but no, it just won't happen. But that being said, I share your enthusiasm for space, Chris.

I won't yell at you this time, but- I'll buy you a burger. Three years. Okay.

Well, I will buy the veggie burgers there, if that's true. That being said, I also believe that ultimately a lot of this is going to kind of come out in the wash, and a lot of these companies are essentially positioning themselves, to your point, basically kind of moving the pieces around on the chessboard so that when it all does fall out, they're positioned for success. And I think that if you look at who has what on the board at this point, I would say that Microsoft and Google have a pretty good position to come out holding the bag on some of these AI assets and ultimately making a new industry out of it, where others who are heavily indebted and heavily invested in previous generation AI components are in much worse shape.

And furthermore, I would say that the frontier labs are essentially, they've got to be acquired. They've got to be somebody else's R&D eventually, and I think Microsoft's going to end up holding the bag on Anthropic, and somebody else is going to end up holding the bag on OpenAI, and you heard it here first, or second if it's called. Wow, that's great prediction.

Hey, I think we're out of time, but that went fast. You guys were great, as always. " We have a lot of great programming on Techstrong TV that's on the rest of today and the rest of the week.

I'm sure we're going to hear from Mike and Alan from the Splunk Conference in Denver. And thank you for channeling Alan, by the way, Steven. We appreciate that and- I did it poorly, but I tried.

No, it was good. It's getting better. Thanks again for watching, and we'll see you next time.

The agentic AI race is on. Which approach will move your team forward? Join DevOps Experience 2026 for practical insights into AI and DevOps.

September 24th, free and virtual. com.