Apple Pushes AI, OpenAI Files Quietly and SpaceX Teams With Google

Hey everyone, happy Tuesday. Welcome to the Techstrong gang. I'm not even going to talk about basketball last night, except that the NBA obviously wants this series to go seven games and instructed the refs as such, and I'll leave it at that.

We've got a great gang today. There's so much interesting stuff to talk about, it's hard to pick out, but we've got three great segments for our great gang. Let me introduce you to who we've got on today.

We've got Sid Nag joining us. Sid, good to see you. See you, Alan.

Chris Lask, sporting a scarf. Always happy to see Chris- God ... in his scarf.

One and only, Kate Scarcella, in the Elton John glasses, kind of. Oh. In Vegas.

Our friend Jon Swartz. One in Vegas, yes. One in Vegas, exactly.

Look, what happens there, you don't have to tell us, Jon. And then, the man in the high castle, probably still licking his wounds a little bit from last night as well, Mike Guzzardi. Gang, welcome.

It's a great Tuesday, except for basketball, obviously. But before we get into it, is there anything you want to say about the basketball game? Well, I'm glad the circus is done that came to town to watch the game, and we can- Yeah ...

get back to normal. Well, and that was the jinx, too. But let's not- Yeah ...

get into it. I'm not going to get into that aspect of it, but I'll just say flat out, I don't think the Spurs won that game as much as the Knicks lost it with a little help from the referees, and we'll call it even. Let's move on.

So, this is just too much of a cliche and corny, but Apple got serious- There you go ... yesterday at WWDC. Tim Cook's last one, so happy for Tim.

But Mike, too little too late. I'm going to lean heavily on Jon for this, but I don't know. I was reading this stuff from afar, and I felt like this was the developer conference.

So they got up and pretty much said what they've said in the past before, and it was one of these, "When I get off this couch, I'm going to kick some a*s, and I hope you guys write some code because we'll be ready someday soon. " But Jon, I kind of got a sense of deja vu reading this stuff. What's your take?

Yeah, it was deja vu all over again, as Yogi used to say. I remember going to these events back when I was a newspaper reporter in 2016 and 2018, that timeframe, and Apple was talking about AR and their AR toolkit. Oh my gosh, I heard about that for a couple of years.

Nothing really came of it. Eventually, something did come of it. And I kind of get the same sense from Apple Intelligence and their whole AI push, and I use the word push really loosely.

This was basically an opportunity, maybe a passing of the torch from Tim Cook, who's leaving in September, to John Ternus, who's a hardware engineer, who I am positive will be making some efforts and making some progress in this area. And maybe in a sense, this is an attempt to kind of airbrush the company's image as an AI laggard. But in a sense, what they did was they made a couple of announcements.

I'll be really quick with the announcements because I know everyone wants to talk about this because it's Apple. But in practice, they were integrating AI across the entire product ecosystem, specifically Siri. So Siri, in a sense, will become more conversational, capable of handling multi-step requests, conscious of onscreen context, is going to receive its own dedicated app to track past interactions.

The Photos app's going to gain advanced generative AI editing tools. iOS 27, surprisingly, is going to support devices dating back to the iPhone 11. There is a bit of a however element here.

The rollout's going to face some global hurdles, especially in Europe. So Apple announced that its AI features will be restricted in the EU and China, which had an effect on the stock. So if you're in the EU, you can access the new Siri on your Mac, but the phone and watch capabilities are blocked.

I think my general takeaway from this, and I think it's more messaging and marketing than it is product, is that at WWDC was where Apple tends to be leaning into its traditional strengths rather than trying to win a model race. So while its big tech rivals are spending hundreds of billions of dollars annually on infrastructure, Apple is positioning itself as a consumer-facing AI distributor rather than a model developer. And maybe in the long term, that works.

I think Alan thinks it's going to work. I think it will too. But in the interim, it makes for a kind of a somewhat boring, muted story, and a kind of kick-the-can-down-the-road moment.

Yeah. Focusing in on the Siri side of it, my attitude was welcome to 2022. The whole world is talking about agentic AI, and they're touting what their generative chatbot's going to be like.

And it's not even their generative chatbot, it's Google's chatbot. Yeah. But we've got the Apple security, and that's what people are going to pay for, right?

Because we're going to make it more secure than those guys are because we're Apple, and we never get hacked. But beyond thatWhat was in for 2022, 2023 here? Well, that was the interesting thing in my mind, and I don't know, John, you're closer to this, but they were going out of their way to position Siri as almost the broader, more important digital assistant that would have AI capabilities versus relying on an AI agent or something that you might run locally on the device, and especially if that AI agent happens to come from a third party.

But it seems like they are trying to go to some lens to say, "Hey, Siri is still the center of that Apple universe," and- Yes, it is. Yeah. They're wedded to it.

They're betting on it. They're going to stick with the same horse, and they're on, I guess, in a sense, what they want us to believe is they're onto the next phase in this kind of weird journey, and with AI as Alan said, a trusted consumer experience layered across its ecosystem where you emphasize privacy, device integration, distribution as your key differentiators. But I think it appears that Apple decided two years-- That being two years late with their own model is worse than being on time with Google.

I also think that they're trying to extend Siri from being an interactive chatbot to something that would do sort of complex multi-step interactions across different applications. So, I think that's an interesting point. But the one thing that jumped out at me is the focus on agent-centric than app-centric.

So are we now going to have an agent store as opposed to an app store, which is going to be very interesting from an Apple ecosystem perspective. And then I think they also are working with other model builders beyond what they call the Apple Intelligence framework. So it's not just contained to that framework and that AI model, but they're going to be co-opting with other model builders.

So I think that's an interesting observation that I made. But I think it's clearly a shift away from users manually navigating apps towards AI systems that coordinate workflows on their behalf, more of a complex multi-app task management capability within the Apple device, if you will. With Siri being the center of all of that.

The part that I don't get is, so this is a developer conference, and they're not exactly being clear about where the functionality of Siri is going to end and where I, as a developer, might build something that I would actually put in that agent store of theirs. And so if you're a developer, you're kind of sitting on your hands frozen a little bit, or the hell with it, you're just going to go full speed ahead and hope for the best. John, is there a little- Yeah, that's totally spot on.

So there's a couple of guys who I've talked to who used to work at Apple. They worked on the iPhone development. They worked on iPad.

They are now AI developers, and they told me the one company they have more problems working with is Apple, because they have no clearly defined delineation of where one thing starts and one thing ends, and it's this unending frustration game for them. So there is, there's that disconnect, and they almost feel as if they're kind of hanging in the balance. And this one guy in particular who I've talked to off and on for the last few years, worked there for 20 years.

He says it's maddening. They don't get a lot of support, whether financially or document-wise, or even just meetings-wise. He has a hard time getting through to them.

And I also, I don't know. When I see these events now with AI in particular, it kind of reminds me a little bit of some of the iPhone stuff, where it's so incremental to the point where it's like, aren't we repeating ourselves? Isn't there a lot of overlap what we just talked about two or three years ago?

And I know they're being deliberate, and I know they're being methodical, but there kind of reaches a point in this technology where you can't keep doing that. No, it's starting to sound stale and old. Chris, you wanted to say something?

So everything in today's show looks like the unbundling to me. So Apple wants to keep the user interface, that's the obvious thing. But honestly, this is the same Apple.

Remember December, when everyone using Apple on Earth couldn't use their devices anymore because they changed the entire user interface all at once. For a billion people, the stories were endless. The 87-year-old neighbor who had to show me how to use the new calendar.

And that's the centralization problem. And we add to that, what we're talking about is the world's largest deployment of agentic AI. A billion Apple users are all going to get agentic AI on their phones.

And Alan, you remember the twins. My solar powered boats, I have Alexa built into them, and just with a single Alexa account, trying to get enough awareness of my specific environment, with a handful of IoT devices, it's not possible. So we're going to, I'll just say it, we're going to inflict agentic AI, which, I think Sid, you said it.

This is exactly right. It's where they should be going, but from a company that couldn't get the UI right six months ago. Look, when we say agentic AI, they've called Siri an agent from day one, but I don't think that's the agent you're looking for.

That's not the agents we're talking about today- Mm-hmm ... in terms of the open clause and the codexes and so forth. It's still a chatbot.

Well, not necessarily. It could graduate into a meta agent of agents. So if you build- If the queen had you know what, she'd be the king, too, but as they're describing it right now, it doesn't have those appendages I think Apple's not going to let go of their control of the Apple ecosystem.

But we'll have to recognize that there are others who are going to go build agents, and their users are going to want to use them, so either they do it or they're going to move to the Android platform. Right? So Apple really wants to welcome their users to the world of agentic and agentic apps.

So those apps have to sit somewhere. So maybe there's, like I said, could be an agent store, like an app store where you download agents onto your device. But they don't want to have chaos.

So they want Siri to be the super agent controlling these agents. That's the way I see it. Okay.

First, I'm going to come in here, and I have to say that as a person who is 60, I thought I've heard everything other than this word came out of Helen's mouth. I have never heard that expression before in my entire life. You've never heard that?

No. So thank you for that. You know what?

Who did I hear say that recently? I've never heard it. This is the first time, so thank you for it.

King in the King? Yes. So thank you for that.

No. Someone famous actually said that recently, and I'm trying to remember who now, and I can't remember it, obviously. It's all right.

I'm going to give it to you. So there you go. Okay.

I'll take credit for it. You own that one. But I'm going to be the little Pollyanna here with Apple, and it has to be because I just remember starting from the Sinclair and the TRS-80 and having the five and a quarter disc, and then going to Apple and having the three and a half inch disc to bring up the new face.

I don't know, folks. I'm really feeling very, very positive about Apple. I think if anybody's going to be able to do this, it's actually going to be Apple, where they put intelligence that feels very, very natural in our personal hands.

So I'll be the only one here. And so Mike, when you go back and you say, "Kate predicted this. " Apple will be- And you know, Samsung...

I don't think Samsung will get it together, huh? You have to admit, we've come a long way from writing basic code on the ZX81 Sinclair platform. I wrote that back in the day, so There are always ones that just sort of seem to...

So I think it's good that... So this is what I think. I think it's really good that they actually are not putting out a bunch of new features, that they've actually stopped.

I think it's great. The new CEO who's taken over, I think his background is great. I think from a software perspective, people not understanding, hey, what is software doing?

I think everybody needs to say, "Hold up on the features. Hold up on the software. " I think Apple's doing exactly what needs to be done.

So yes, Mike, you kind of- So my concern, though, is we wind up in this situation again where it's like the video conferencing stuff and the software only runs on their platform, and then I got to go get WhatsApp so I can talk to my kids who are using some other platform. And with agents, I kind of don't want to have that replay out. I kind of want to have an agent that talks- Yeah ...

natively to everybody else's agent and not through a bunch of gateways because of whatever protocols and proprietary systems that are in place. So I wonder if this strategy's going to work in an era where more people are looking at agents maybe as an app rather than something that's an extension of an OS. Does that make sense?

It does, and you mentioned protocols, and I think that there's some good company, a good company out there today that's sort of looking at that, and I think actually, interestingly enough, came from Apple. And I think that we will have this platform that will directly work with Apple with one protocol, one sort of workflow type of agent that will be the middle broker in some way. So I think we have no idea how great the future looks for us and our digital technology.

So I'm really excited for where things are going. So everyone hang onto our seats because we are about to see some really cool things in the next year. All right.

Mike- We've got to jump. Well, wait. There's a lot of people in New York who are reconsidering their commitment to Samsung because, well, Samsung announced that they're moving to Texas in the middle of the Knicks-Spurs series.

" There you go. Well, look, I think the takeaway here is let's just mark it down that Kate is a big fan of the Apple announcement- ... and their strategy going forward here on AI.

And we'll revisit that in a few months- Yeah ... as we're all using Huawei phones. But let's jump- ...

Mike, to our next segment here. OpenAI, man, they're playing catch up lately. It seems like the- I know ...

the Anthropic people have beat them every step of the way. They have joined what is apparently a rather exclusive, confidential IPO club, and I look at this since they did it and Anthropic did it, and to me, it just screams alarm bells. It's kind of like, let me get this straight.

Every other company in the world has filed an IPO and it's been a pretty straightforward affair, and yet you guys got to have confidential IPOs. Plus, I will add to this bit, Standard & Poor's rejected Anthropic, OpenAI, and SpaceX because generally speaking, I think that they think that maybe they don't have a good financial underpinning of this whole model and what's going on here. So Sid, should we be concerned here?

Because it feels like there's a lot of sleight of hand going on. Yeah, we talked about this last week, right? With the Anthropic IPO.

Clearly, the whole AI business is shifting from a venture-funded industry to a public market asset class, right? So all these companies want to raise a ton of cash. They need the cash, right, to grow their business.

And many of them have revenue, but nowhere near the... 22 for every $1 they were earning. So you can do the math right there.

So their valuations are probably going to get impacted. So they don't want to raise more VC capital or any sort of internal private equity, et cetera. So they're going to the public markets, and they're expecting us to buy their shares and fund their business.

So it's going to be identical to the Anthropic story and the SpaceX story to some degree. And I think the other interesting observation that one needs to think about is as they go IPO and they become a publicly traded company, now they have to file all their SEC quarterly filings and disclose all the nitty-gritty details. So all of that is going to come out in the wash.

So I'm going to be very carefully watching what are the internal workings of these companies as a consequence of going public, which will result in all the filings that they have to disclose, including how much they're spending on infrastructure, how much they're making on their model, on their prompt engineering tools, and all of that. So I think the breakdown is going to be very interesting to watch. I think they- Hey, Mike, can I mention- Go ahead ...

one thing? To your point about this creating some unease, imagine this statement. This statement is from OpenAI about their IPO.

Anthropic does an SEC filing that's pretty specific, lays out what they're going to do, what their projections are. This is what OpenAI said, and I'm going to just quote part of it. "We expect the IPO to leak, so we're just announcing it.

We have not decided on timing yet. " That should really give investors a lot of confidence. Well, they do have to disclose pitfalls.

But guys, here's the deal. First of all, Anthropic and OpenAI at this stage are two different pictures. From what I've read, Anthropic was actually profitable this last quarter.

And they are reducing their cost of compute per dollar of revenue earned or generated. And they're generating more revenue. So I think Anthropic has clearly overtaken OpenAI from a business point of view in terms of financials.

However, we've seen a real cleavage or a real diversion between them where Anthropic is clearly going after the enterprise market, developers and so forth, and OpenAI has become the consumer brand, right? They've got millions of users, mostly consumers paying $20 a month, where Anthropic has less people, but paying a lot more money. But you're right, neither one of them, if they were a regular IPO with a market cap of $4 billion, you can make a great argument.

But these companies are going public for north of a trillion dollars. A trillion. Rarified.

You've got a half a dozen to a dozen companies that have trillion-dollar market caps. These babies are coming out IPO over a trillion. SpaceX closer to $2 trillion.

So if you're looking as a wise investor, they may be good companies with bright futures. Are they $2 trillion companies? 9% of the S&P 500?

Sure. I don't think so. So maybe the public market doesn't reward them.

Maybe Wall Street gets hesitant. Where are they going to go for money? The BC market, you said, is worse than the IPO market.

There's only one place left to go, and they're waiting on it. The government. The US government is going to take stakes in all these companies.

They're going to give them the money, some billions of dollars that they want because the government, yeah, what's $10 billion, $20 billion here and there? We spend that on compensating people who tried to overthrow our government. And that's who's going to pay for this.

Kate, you're laughing. Kate, but this is who's going to pay for it. The government is going to take stakes in these companies because it's for the public good, and as we spoke about yesterday, Mike, it's the beginning of the basic, what is it?

Basic income? Yeah. BIU.

Right? Yeah. It's the beginning of them being able to take this and set it up for the disruption that these AI tools are going to cause.

I think it's time for these guys to open up the kimono. They really don't have a choice, right? They have to figure out how to convince their investors that they are a viable business, right?

I mean, you said- Not if they can convince the government that they're going to be so disruptive to everyday life that the government has to get involved Yeah And what Alan is saying is actually true. It's unfortunate, but it's a true statement right now, and it's nauseating. Also, Anthropic's profitability is a very convoluted number because they're not accounting for stock-based compensation, which is a huge, huge value if you look at the dollar.

Absolutely. So they're playing around from-- S**t. Financial matters.

The phrase you're looking for is financial engineering. It's very technical. Pretty much, yeah.

Creative accounting. They've got an audience of one right now. Mm-hmm.

Yeah. I will generally leave the financial analytics to you folks, but I was at Cisco in 2000 during the dot com crash, right? And it was just a fascinating thing to see from the inside, because there's billions of dollars flowing.

But, you get this chain of things. The service provider line of business was the hot part of the company at the time. Right at the time, all our big customers who had big payables, the service providers, as we recall, a lot of them went away, right?

A lot of really popular names weren't there anymore. And I'm not making any predictions about this sort of thing. As you say, it's very complicated.

There's a lot of value floating around. But again, we're getting back to centralization risks. And if it goes wrong in one spot, then there isn't other ways to soften that blow.

All that dark light gets lit up, right? So here's the thing I want to keep track of or asking people to keep in mind. There's no connection between rational thought and what happens on Wall Street, so I believe that these stocks will go public, and they will be very popular, and demand for the shares will go up.

However, you want to keep a close eye on just how many of the insiders are selling, how much of their stock- Yeah ... or what and how quickly, because that's going to tell you a lot about which direction this thing is heading. It's going to be pretty sobering, I think, with OpenAI too eventually, because for the first time, they're going to be forced to raise the curtain on its true revenue, profit margins, massive infrastructure losses for the first time, and that's going to throw a bucket of cold water on people.

But you're right, Mike, initially, it will do well, but- Mm-hmm ... what the big stockholders do is maybe a little bit more interesting than what the general stock does. Mm-hmm.

All right. Sounds like we talked this one out. Maybe we should move along, move along.

All right. The AI super highway. What do we got next, Mike?

Well, this is just mind-boggling, but let's start with the beginning. So Google revealed that it is using GPUs from SpaceX in what it describes as a, quote-unquote, "temporary arrangement," and basically it is a second source agreement, as far as I can tell. It's to the tune of millions and millions of dollars of GPU consumption.

Millions? I'm guessing billions. Yeah.

Which SpaceX apparently has a lot of excess of these GPUs out there for some reason. So I guess they were using them quite literally for rocket science purposes, but maybe the utilization rates for that aren't particularly high, so they have a lot of this excess capacity going on. But as you look at all this stuff, I'll throw this to Chris, but what is your take on what's going on here?

Well, I think the AI world is discovering fleet economics, right? And I'm a big space fan, right? Apollo-era kid, been watching it all my life.

And one of my favorite little name drops is Gwynne Shotwell, right? COO running SpaceX, and six years ago, I got to spend an hour with her and spend a lot of time talking about these things. And what she was excited about at the time, so Starlink was just coming along, and Starlink is profitable.

SpaceX is profitable. But she was, before all this, really excited about a company she had started called Orbitz Edge, and that's what we see now in the million AI constellation licensing with the FAA and so forth that's going on right now. And this looks to me to be profitable, too.

Right? Having massive compute in orbit permanently with a cost logistics supply chain where it's predictable and, again, not in this ephemeral OpenAI, for example, monetization, but a physical infrastructure that pays for itself and provides that compute. Yeah.

If I was Google, would I be building a lot of data centers, or would I partner with Gwynne? Yep, I would probably partner with Gwynne. Why not?

I don't think that's the GPUs they're getting. I know, but it's a step in that direction, right? Because- Well, again- But that's Grok ...

there's Grok on the way to, yeah. It's a step in that direction. Well, SpaceX and xAI are the same thing now.

And this is all one big- And that's the point. Don't be fooled by the word space in the title. This is not about SpaceX or Gwynne's thing.

So here's my take. It's about Grok. Well, this is about Grok failing.

Right. Well, if they're quote-unquote- So here's the thing, because this, you can't just look at the Google deal. Last month or the month before, SpaceX announced a similar deal with Anthropic.

Right. The Anthropic deal calls for them to lease to Anthropic up to 220,000 GPUs to Anthropic. That's twice what this Google deal is, 110,000 GPUs.

Together, that's 330,000 GPUs that they're leasing out. I think this is not an admission, but a silent admission that they're not going to be using that for xAI or Grok or whatever you call it. Maybe they're getting out of the model game.

And this is the article that I wrote. This is more parallel to the airline industry, where we're seeing stratification Of the different players, right? Nvidia could have developed their own model.

Nvidia could have been an AI provider in that regard. They had the software that everyone's using. They make the GPUs.

But no, they chose their lane to stay in. They make the GPUs. Hold on.

The key word is leasing. Okay? So first of all- But that's what the airline industry works on, leasing ...

Right. Correct. So if you know, I'm sure you do, SpaceX or xAI is going to redo Grok from scratch, right?

So they're going to need, and with Cursor as its harness and so on and so forth, you can get into all the details of that. We don't want to right now. But there's plenty of work that's going to require that SpaceX or xAI, depending on what you would call it for this matter, is going to have on its hands to utilize those GPUs.

But they don't have it done yet, so they're going to utilize it in the future. What do they do with that capacity lying around? They're going to lease it, to your point, 220,000 to Anthropic, 110,000 to Google, right?

And Google, I think, is also realizing that-- Google is heavily invested in TPUs, not GPUs, right? So while they go and build out the TPU technology with the announcements they made at Next with 8i and 8R and all of that, well, while they build up the capacity for their needs, they need something in the interim, and that's what's going on here. Or maybe, back to Chris's point, SpaceX wants to be a provider of compute, of GPU, not of a model per se.

They're going to be a provider of compute. Now, if you're Google or Anthropic, though, this is a great hedge. Look, I've got people protesting about the data centers I'm building.

They're not coming online as fast as I thought. We got problems with fluctuating energy prices. This takes a lot of that risk off the table.

The same way an airline leases their airplanes, they make long-term fuel agreements, so they fix in what the cost of fuel is. It's the same model. So Alan, are you saying that Elon's game of selling futures like FSD is not working anymore, so now he has to get into the GPU- No, I think he- ...

physical business, the Core Weave? Is that what he's doing? No, I think he realizes that the SpaceX model that Chris is talking about, putting stuff in low orbit and everything else, that may...

Look, you know what? The guys at Nvidia did pretty good with this model, right? " Well, now wait a minute.

So here's the part that I'm scratching my head about because last time I checked, coming up with an alternative strategy to make up for your first screw-up is not something to applaud. And the first screw-up seems to me is that they bought a boatload of GPUs that they didn't have a whole lot of use for, and they're sitting around in a factory or in a data center or wherever the hell they are, in a warehouse. " And by the way, I'm pretty sure that somebody wrote a check to Nvidia for those GPUs.

So who's winning here again? Nvidia. Well.

But I think, Alan, you're right. My expectation is that in the not too distant future, all the major horsepower compute is in orbit, which means that by definition, you have to have orbital capabilities to even put it there. So yes, SpaceX becomes that hardware layer.

Now, monopolies, there's the Skynet risk right there. I'm not a fan of the CEO. I love the COO, but it's more practical than that.

SpaceX makes money. They launch hardware. It's a very specific, practical thing that works.

And if that's going to work at all, they could be the global hegemon of orbital AI, but that doesn't make any sense. There's going to be Meta and everybody else's models will run there as well. And now I have to say, I'm a fan of Grok.

I think Grok has a great architecture. Of the three major model vendors out there, system vendors, I think for some of the same reasons I'm not a big fan of their CEO, their AI architecture I actually approve of slightly more in certain ways than the others. But I can't see them fighting that out.

They're not going to die on that hill. Grok, xAI will exist, and will Gemini and the rest, and they'll all run on hardware, a lot of it in space, and SpaceX will own that. Yeah.

No. I think Google is making a substantial bet that Gemini and Gemini Enterprise is going to grow so fast, and that's going to need a lot of compute capacity. So they're really focused on the growth of Gemini, to your point, Chris, and Gemini Enterprise, which they announced recently.

They're already getting a lot of demand for that, and that needs a ton of compute. I think it's a smart bet. Yeah, totally.

I think they are well-situated to be one of the leaders in bringing AI because they got Google behind them. Yeah. But guys, the 8 trillion or more data center build-out that we've been talking about is running into trouble.

People don't like it. Yeah, I think. And it's costing more than we thought, and it's taking longer than we thought, and this is a- I love your- ...

good way of putting it on someone else's problem. There was a community just, what was it, last week in Southern California that Voted on permanent ban on data centers. Yeah.

So, this stuff about data centers in space, maybe someday, and we'll get the latency. Maybe someday, but you know what's going to happen first? There's going to be a moon base, and then there's going to be people living on Mars by the time that happens.

So I don't think that- Yeah. Three years. Three years.

No, Mike, I'm with Mike. I'm with Mike on this one. Yeah, for sure.

I think we are building- I'm voting Mike, too. All right. I think we're building for, again, what I've talked about a lot, we're building for today, what we don't even understand for tomorrow.

I think context and memory context, even within our own personal devices, will become stronger at the memory and how it's computing within our own environment. So I'm very uncomfortable with who owns the largest compute farms, but I think the future is not in bigger warehouse full of GPU. Well, and you know I agree with you on that 1,000%.

But at the same time, I look around the world, and people are spending trillions of dollars on data centers. SpaceX is going to be spamming low Earth orbit with GPUs. And that's what's going to happen.

So the question is, how does that play out? Right? And I think on the infrastructure part, as we said, I think just last week, right, the problem with this infrastructure spend is we're spending it on stuff that doesn't last, right?

Then in five years we're going to throw it away. So what does last? The physical buildings and the power grids and everything else?

I don't know. Because I think that over the time frames that is calculated in five and 10 and in decades, no, low Earth orbit is just cheaper. If in fact we're going to throw the GPUs away in a couple of years, let them burn up in the atmosphere, that kind of thing.

At least that's where the trajectory is going over the next three to five years. And sorry, Chris. And in terms of a blueprint of taking a company public, you look at what Altman, you look at the statement that OpenAI makes.

" Then you look at what SpaceX is doing, where they make these announcements with Google and Anthropic, and if you take the combined potential revenue, it's $26 billion annually from these two contracts. That's what Wall Street wants to hear before you go public. Like, here we are, we're this juggernaut AI cloud infrastructure behemoth.

This is how we're using xAI and Grok. This is the messaging. I mean, whether it- And here's what I- You can't use the word cloud in that sentence, because this is- No ...

this is the antithesis of cloud. Because cloud is all about elastic capacity, and this is dedicated capacity. It almost seems to me we're going back- But let me talk about- ...

to the bad old days. So what happens- Sid, what makes you think they have 330,000 GPUs available in an already built-up data center? Because this isn't about waiting for space to come online.

I don't know. I don't know if they have it. But regardless, it's dedicated capacity, guys.

This is not elastic capacity. Well, but- It's not cloud. Okay?

But should it look- So- Call me old fashioned. Call me old fashioned, but if I contract to sell you something, shouldn't I have it to sell you? What a concept.

You should. What a concept. You should.

But SpaceX is not in the cloud business. They're in the dedicated compute- But they're contracting 330,000 GPUs. And not just GPUs, running GPUs in data centers.

Yeah. That's why Google said it's temporary. Data centers, not cloud hyperscaler.

So- No ... here's a million-dollar chip. Is Google going through a, what's it called?

Dr. Jekyll- And Hyde ... and Hyde situation?

Are they- Dr. Jekyll and Mr. Hyde.

I don't know. Their cloud computing- No, I think they realized that let someone else have the pain of building out the data centers ... their cloud computing business is growing at double digits, right?

And so that business is still very strong. And here they're going and investing in dedicated compute capacity. And then what happens to the Amazons and the Microsofts of the world who are running a huge, huge, huge hyperscaler cloud business, right?

So we're seeing a massive industry shift. But they're all building- All that, right? They're all building AI factories.

Yeah. I'm going to go with Kate on this one in the sense that we don't know what's going on and there will be a lot of changes, but I would also point out that the one thing that is clear from all three of these little segments is people are making s**t up as they go along. Welcome to business in 2026.

Oh. Hey, on that note, we got to wrap it up. What an interesting- That's funny ...

continually, day in and day out, we are just seeing how AI is wrenching the threads that hold our civilization together at some level. And our whole business. Someone I see put out on here, "Throwing DCs in space is stupid.

The GPU compute for inference is going to be in our phones. " Thank you for the comment, and we could discuss it on the next one. But for now, Kate, Sid, John, where's the glasses?

Oh, sorry. Sorry. I was admiring- Hey, you know how to end it like that ...

sorry. Like a rockstar. Chris, Mike, thanks for joining.

Thank you for watching. Hey, remember, we do this every day at noon, so stay tuned for that. If you miss it, you can always check it out on Techstrong TV or our Techstrong OTT app on just about any screen you like to watch this kind of stuff on.

But until tomorrow, this is Alan Shimel. You've just watched Techstrong Gang.