NVIDIA’s AI Land Grab, Gates’ Warning, and Who Pays

Hey everyone. Happy Thursday, and it is Thursday. I started yesterday's show saying Happy Thursday, too, but it was only Wednesday.

So I'm glad I got the right day, I got the right show, and you're watching Techstrong Gang. Let me introduce you to our gang for today. Making her return, I haven't seen her around here in far too long, my friend, Anne Anne Ahola Ward Ward.

Anne, how are you? Good morning. I'm great.

How are you? I've been a busy lady, doing a lot more TV lately and- I saw that you were on the local Houston show as well, right? Yes.

Yes. That was my 14th on that station and 106th appearance overall. But who's counting?

Not me, but- I am. Are you that busy you can't put your holiday decorations up? This is Dolly.

Oh, okay. This is Dolly Parton, and it will stay up as I'm a huge Dolly... Or I will always be a huge Dolly fan.

I think we are all Dolly fans and- Yeah. I thought you'd have your Tim Curry stuff out, too. Come on.

You know, Tim- You are a Rocky Horror type of gal, actually. Well, the- I don't know how to take that. I've seen it many times, yes.

So I got to be honest- Will I be watching it again this year? I never saw it in the movie theaters. Wow.

Oh, you have to. No one does it like they used to, where they would get all dressed up. No, they dress up, and the rice and everything.

My wife has done it. I think I was just a little on that side of the divide age-wise for it. Right?

Yeah. I don't know. I think you were just too conservative.

Yeah, that's me. They were still playing it roughly back then. Young Conservatives of America.

Speaking, let's move on though. But Anne, it's a pleasure to have you here. Our friend Tom Hollingsworth is here.

I don't know if you guys know this- Hey, gang ... but Tom's an OU guy. I am.

Right here in Norman, Oklahoma. It's getting ready for football season, and you know how it is. We're doing it right down here.

All right. And then- I wanted to like you, Tom. I wanted to like you.

And I'm not getting a penalty for that either. All right. You two, go to your corners and we'll wait for the bell to ring.

But speaking of the bell to ring, we've got our man out west, the one and only Jon Swartz. John- Hello ... what does it say on your shirt?

Is that Georgia? That's what I thought it said. Bulldogs.

Yeah. So we've got the Oklahoma, Texas, and Georgia this weekend. It's all SEC, baby.

Oh. And you got Florida and your family. Yeah, Gators.

Yeah. And the man who has-- He has no real college, Gabe. Who's he going to vote for up in New York, the Connecticut Huskies?

I know. Wait. I did go to Boston University.

That's true, he's a BU Eagle. Yeah. Yeah.

We got those ferocious terriers, man. I- Oh, that's right. That's the BC Eagles.

You're BU. Gee. BU.

I forgot for a moment. Anyway, the one and only Mike Bazan. Guys, welcome.

Welcome to Techstrong Gang. Thanks for joining us. So Mike, look, NVIDIA just stole everybody's breath away and captured all- Yeah ...

captured the internet it seems. Literally. But there's a lot of angles to this story.

What are we covering today? There sure are. So NVIDIA and HuggingFace are, at least reportedly, I'm not sure if it's confirmed yet, but they're going to get hitched.

And at the end of this thing, all these open models are going to be distributed by NVIDIA, and that gives everybody cause for pause. But John, what are they saying about this thing? You got the story.

What's going on? Yeah. So you're right.

Both of you are right. There's so many angles to this, and I will kind of tease a couple of them, and then we can all go down our various paths. 9 billion, which I would argue, and this is not hyperbole, this is going to change the AI landscape significantly.

The deal basically gives NVIDIA direct control of the model hub that much of the open source AI ecosystem builds on top of. And in a sense, they are, when I say they, I mean NVIDIA, are building their empire beyond chip dominance into the software community layer. We've been talking about this, about the software as a moat strategy, and this is a classic application of that.

What's also interesting, and Tom and I just kind of mentioned this in the green room earlier, is that Microsoft is also looking at buying HuggingFace. So there was a lot of action going on, and NVIDIA kept this very, very quiet. They have reported their results, and this is the blockbuster.

So in a sense, NVIDIA's created this strategy, I think it's a formidable strategy at the software level. At the same time because we have, in the bigger picture, OpenAI, Google, and Amazon are designing proprietary custom chips to cut their reliance on NVIDIA hardware. One thing, and I will tease this, and I won't go deeper because I know we all have a lot to say about this, but what I'm interested in, among other things, is HuggingFace's neutrality as an open vendor-agnostic hub once it sits inside NVIDIA, and what's going to happen there.

But I steer the four of you all over there a lot. I guess, John, quick question to you. Is this an issue where HuggingFace was running out of cash, or did they get an offer they couldn't refuse from NVIDIA?

Yeah, I think they were doing fine. I think what's happening, and this is part of a bigger picture story that we should look at, is this accelerating wave of industry consolidation. So Stripe bought OpenRouter for, what, $7 billion?

I think we're going to see more of this. So I think it was getting offers, and NVIDIA Yeah, either outbid Microsoft or it outmaneuvered them. I actually think we're going to start seeing more of this because we've got OpenAI and Anthropic going public, which kind of exhausts the IPO market.

So some of these other companies are probably going to get some really attractive offers, because they're not going to go public, and they're going to have this opportunity as an exit strategy. So I think this is just a great offer they got. All right.

Tom, what's your take on this thing? We knew that this was coming because there were rumors that were floating around. The funny thing is that NVIDIA managed to keep their name off of those rumors.

" Because this is what-- It's about a three or 4X multiplier from their valuation from just a couple of years ago. And I think that, given the fact that Microsoft and NVIDIA were both chasing this, it says to me that the next frontier of expansion is going to be in getting the models attached to the hardware that you're selling. NVIDIA is the gorilla in this situation, right?

Everybody is following their lead, and I feel like they needed to do something that would differentiate what they were working on so that they weren't just seen as an infrastructure vendor. They have to have something run on top of that. And buying up the SourceForge, if you will, of models is a great way to do that.

" I guess, yeah. We'll see how that works out. Mm-hmm.

And there's some folks out there who are hardcore open source committed folks, and they're a little sad this morning. So what's your take on what's going on here with this thing, and is this a signal of more things to come in terms of big companies just rolling up these open source players? They already own an enormous amount of compute that sits underneath AI, so this just gives them another layer of the stack.

But I also wonder if it's a defensive move to try and keep Amazon, Anthropic, and others from, I don't know, making their own chips. NVIDIA doesn't want to own every model. They want to own the piping in the road that every model's developed on.

It's strategically very smart, but I have to wonder. They seem to be playing 3D chess here. I have to wonder what this is going to do to the developer ecosystem and what this is going to do to open source.

So look, NVIDIA is not an open source antagonistic company. They're, in fact, a real big supporter of open source. They recently invested some serious money, well, not NVIDIA serious money, serious money for us, a couple million bucks into CNCF and Linux Foundation and stuff to help with AI.

So, and CUDA and much of what they build on is built on Python through Anaconda open source. It's not the open source people that are sad here, it's the people who want open market. Right?

An open market. NVIDIA is an investor in OpenAI. They're an investor in Anthropic.

They're an investor in SpaceX. Are they going to favor those models versus the two million other models on HuggingFace? I don't know.

But that's the real game here. It's not that NVIDIA's anti-open source. And a couple of other things.

Look, the IPO markets aren't shut off. If anything, if these two, OpenAI, Anthropic, go well, you'll see others following. But you've got a feeding frenzy going on.

Right? You saw it with OpenRouter. Tom, you said three to four times the less valuation.

It's crazier than that. This is a $150 billion revenue company going for $13 billion. Think about that.

I was going to mention quickly that this was teased by-- What happened today, or what allegedly's going to happen, was teased by NVIDIA at one of their shows. There was a pre-conference talk show with Jensen, and they talked about this whole kind of software-focused- Yeah, but they had to keep it quiet- Yes ... while the earnings were announced But they've left breadcrumbs over the last several months.

Yeah. This is not a total surprise. But think about, look, this is a company at the top of the heap, on top of their game.

They waited for earnings to be announced, and they blew out their earnings. Yeah. Stock was up after earnings.

You never see that because people buy on the climb and sell on the-- Right? Mm-hmm. Stock was up after earnings.

Then they announced this, or rumors start with the HuggingFace deal. They announced an AWS deal with two million GPUs next year. Anthropic deal also, hundreds of billions of dollars.

Just crazy, all well-timed, coordinated like a ballet. Right? This is a company on top of their game.

And they are the beating heart of AI. One other thing we haven't mentioned. In their earnings, Jensen and the team talked about this idea of AI computing being fungible.

Hey, OpenAI can't use that Ohio data center? That's okay. We'll put Anthropic in or Oracle here, or SpaceX there, or what have you.

Fungible computing That's what HuggingFace gives them, and that's what made it worth $13 billion. No. They now can see.

You think not? Not. I think totally overpaid, and HuggingFace is like, basically it's a distribution play for a bunch of open models.

That's what it is today, but it's going to give them into what's going on. Wait, let me finish. Go ahead.

Here's the thing. I can go build another distribution play for open models- Good luck ... and the people who build those models will be like, "Great, let's go there, because I don't want to go through an Nvidia control channel.

" And I look at HuggingFace, and there's no moat here. This whole thing can be replicated probably in- Do you got scars on your back from Microsoft or something? Nope.

Because Nvidia's not Microsoft. A lot of players in the AI space are starting to look at Nvidia like they used to look at Microsoft, that's for sure. They are.

Yep. Well, right. If you're AMD, if you're AWS, you're asking do you want your ecosystem depending on Nvidia's architecture or infrastructure?

No. But I think buying the center of open source AI, which is essentially what they're doing here, is going to encourage a lot more fragmentation, because everybody now has a much larger incentive to build alternatives. So is that a bad thing?

I don't know. Right. " Well, exactly.

And Anthropic saying, "Hey, by the way, we're going to be EU AI policy compliant. " It took less than 24 hours for someone to build a thing to detect and remove it. Like this is the new world where we can do what we want.

I get that, but you guys- Yeah ... you're being disrespectful to HuggingFace. Yeah.

Maybe. Believe me, I've tried building things like this. It's not easy building something that becomes the de facto standard that they are.

Mm. Right? They are the only game in town with that.

Yeah, you can duplicate the technology, but I'm telling you, it's not going to be... What's the big split from Tectonic? OpenTofu.

This isn't OpenTofu part two here. No. Nothing unifies people like a common enemy.

Just ask the great country of Canada. Yeah, but you know what? I don't think you don't see Nvidia and Jensen getting the...

They don't treat them like they used to treat Bill Gates. Not yet, but they will. But I look at the comparison- Is that a segue?

I think more of a comparison with Google, actually. Google could do no wrong. They were the white hat for a while.

Well- I'm not saying it's going to happen to Nvidia, but we love in America rags to riches to rags. So- And Nvidia's 30 years in the making. They were rags for a long time.

They hit on the AI, resuscitated, expanded this company, and it also giveth, and it can also taketh away eventually. And even in the Trump era, when you have a company that is so highly vertically integrated, which was their overall strategy, it's going to invite an investigation for sure. Well, that is coming.

So look, this is- Yes ... this is "The Trap," my book. This is why- Indispensability at its best.

Yeah. We know. So I wonder if this acquisition eventually raises the specter of Nvidia being at the center of global debates over open source safety, regulatory scrutiny, national security.

All of the above. You know, we don't get into- Not under this administration- You just might, but there's also some- ... but maybe in the EU.

Maybe in the EU. But guys, hey, we're over time. Taking it home.

I got to keep you on the stick today, guys. All right. Here we go.

So move on. All right. Shifting gears.

I feel like there's an annual memo from Bill Gates explaining to us how we're all going to hell in a handbasket in some fashion or another, and this one just came out where it's about AI and big tech. Not surprising, but I don't know. And some people looked at this and said, "Yeah.

" Other folks said, "Well, it's Bill Gates accusing other people of something," and that's pretty much the pot and the kettle. So, what was your reaction to this thing? Because I think everybody kind of went, "Yeah, but what's the answer?

" Well, with someone with his wealth and access, you have to wonder what does he know that we don't, right? And he's essentially started warning world leaders they're dramatically unprepared for AI-driven job disruption. This was specifically about job disruption.

" So the other perspective is that Silicon Valley also loves saying AI's going to augment workers, but companies don't spend billions on automation because they want to spend the same on labor costs. So that's kind of naive. So I think what he's really saying is that the risk isn't mass unemployment overnight, but the slower disappearance of entry and mid-level jobs, which we are seeing.

It's harder and harder for people coming out of school to get entry-level jobs. And so if AI starts doing the work people are using to learn a profession, we could break the career ladder. And so I think that's really what he was saying, because the tech isn't waiting for education systems or policies to catch up.

So to just sort of recap, our AI conversation has moved beyond robots replacing factory workers, because it's going for the white collar tasks. So coding is seeing huge disruption, research, customer service, marketing admin, and other knowledge work. " And so I don't think that he was wrong to call it out, I just wonder why now, and I don't think we have a clear answer of that.

But, just to sum it up, I think we keep debating whether AI's going to take jobs when what we're really seeing is companies already redesigning jobs in and around it. Let me add more fuel to this particular fire. Over on Digital CxO, there's a story about what Boomi's up to, and within that story, Steve Lucas, their CEO, is talking about the real danger here is that OpenAI and these companies, wherever they are, as they continue to need to grow, will launch applications into various vertical industries.

You're already seeing some legal stuff and some healthcare stuff. And his point was, a lot of that, while those companies say that they are not training their models on your data, they are not, but they're collecting the metadata, they're observing your workflow, and they understand how your code was constructed. And it's only a matter of time, given their financial situation, before they move up the stack, which Alan has talked about numerous times.

And as that occurs, they will start to hollow out entire vertical industry segments because they'll basically wipe out a bunch of competitors who don't have the same amount of data and the ability to compete. Because, just for grins, some local bank or fintech company is not going to be able to keep pace with some app that has been developed by a proprietary AI platform. Alan, you have written about this as well in some of your stuff.

So what's your assessment of just how far does this all go? You know what? Couple of things.

First of all, harking back to our first segment, since when did Bill Gates become the nice guy, right? All of a sudden, Bill Gates grew a conscience. I appreciate the work he's done with vaccines and everything, and child care.

Not child care, but child health around the world, and I give him a lot of credit. But in terms of this, here's the fundamental question you have to ask yourself. Are you an AI believer?

And I'm not talking about a Monkees song. But are you an AI believer? John got that reference.

Neil Diamond wrote that, by the way, right? Absolutely. Yeah.

There we go. But are you an AI believer? If you believe AI could do the kinds of things that AI believers say it can do, it's going to do short term, then everything Bill is talking about, and the kind of warnings we're hearing from Sam Altman and Dario Amodei and other people are absolutely true.

We need to do something before the wheels are so far in motion that we can't get out from under it. So- But there are- Sorry. Go ahead, sorry.

Go ahead. No, I was going to say, there are other people who don't agree. Right.

Hope is not a method, magic is not a strategy. I'm going to quote my own O'Reilly book. But I think that faith is required to trust these tools to take on tasks within your organization.

But anyone who's actually using it, like I'm using AppSec Script to duplicate docs that I have to duplicate by hand every week. And is it perfect? No.

But am I willing to put in the work to ultimately not have to do that work? Yes. Is it changing how I work fundamentally?

Yes. Is it freeing up time on things I didn't want to do? Yes.

Am I scared of it taking- Stop right there ... from my job? No.

I got a question for you, because I've been asking a bunch of execs this over the last week. Is it changing the way you work? Yes.

Are you getting a lot of work done? Yes. Are you working harder than you used to work?

Yes. Mm-hmm. This is all true.

Yes. But I also will compare it, again, as I always do, to the Victorian Regency era, when people were exploding left and right because of gas lighting in the home. There is always a cost to be paid for innovation and furtherance of technology and advancement of society.

It doesn't mean it's worth it. It just means that we have always paid a cost for moving forward, and we're paying that cost now. We just haven't gotten the tab yet.

Hey, Tom? Oh, sorry. Go ahead, Tom.

Tom, the saying is, every action has an opposite and equal reaction. And one of the things that people are saying is that as these kind of stacks emerge and big tech becomes bigger, companies will try to build their own moat. And the moat will be made up of open weight models and small language models and world models, and they'll go and shift all that behind a big old firewall and then tell their employees that you're not allowed to upload data into these proprietary models because they're stealing our workflow and our intellectual property, essentially, and that will say that that is a condition of future employment.

What do you think? I think that this is going to turn out to be very similar to the situation we ran into with the cloud. Remember how we were not going to adopt cloud computing because it's weird and it's different.

It's not what we're used to. And why did you charge $8,000 to your corporate Amex last month? Did you get a cloud account without us knowing about it?

That's where these people are. You can create an internal system that works well enough for what you want, but eventually people are going to chafe against it, and they're going to go around your controls and do something different and create a pathway to a paradigm shift. For example, since Alan referenced The Monkees, did you know that Davy Jones isn't the most famous Monkee?

It's actually Mike Nesmith Why? Didn't know that. His mom invented Wite-Out.

Mm-hmm. You know, correction fluid. The thing we use to edit typing.

That changed the way people do typing, but you probably don't know that because nobody taught it in school. Tom, here's my Mike Nesmith knowledge. He also pioneered music videos.

Yeah. They're- So he totally upturned that business, too ... they're brilliant people.

And now, do you guys watch MTV? Do you know whether they play music videos on MTV? Do people even release music videos anymore?

" The question isn't whether or not the industry's going to move on. We know it always does. The question is, what parts of AI will it take with it when it moves on?

Mm-hmm. And I know that's true because nobody uses an office phone anymore either because we all have iPhones. And I remember when that was the huge titanic shift- Mm-hmm ...

in the way things were. That people were not sure about whether or not that this little glass and metal thing that runs web AppSec was going to be a difference maker in the way that offices run. Hey Mike, so we both had these personal interactions, good and bad, with Gates.

I don't know about you, but my first initial reaction when I saw this 6,000-word manifesto was, what's his ulterior motive? But I also want to give him some credit because I gleaned out of it some interesting ideas he has, and I want to give him credit for at least bringing these up. This idea of imposing taxes on AI software and robotics that's comparable to a payroll tax.

He wants the government to establish national regulatory bodies to oversee AI across finance, security, healthcare, and employment. Now I'm not saying it's going to happen in the next three years or next two years, but it bears thought. These are things we have to eventually get in front of.

We can't let AI just continue to run roughshod and not think about some sort of governance or some sort of scrutiny before it destroys or upends most of our lives. I don't know. I- We seem to have forgotten we're in control.

Yeah. To be fair to Bill, he's a smart guy. I've talked to him, interviewed him.

But he doesn't suffer fools gladly. Yes. Especially when he was younger.

And so, he tends to be very poignant and will look at people and say, "Go fix this," without necessarily telling them how. And sometimes he surrounds himself with people who are not exactly- Syco fence? Yeah.

Let's just say it's not that they're not smart, it's just that they're willing to do anything to get to the end without considering all the methods and the means that were used. Let's put it that way. I used to see that.

I remember he was with a guy named, remember Craig Mundie? I remember interviewing both of them at the same time, and the topic of spam came up. And this was, God, this was like 15 years ago?

16 years ago? " And the look on Mundie's face, his jaw dropped and he went pale white, and then we moved on to the next topic. There you go.

Yeah, we're still working on that, by the way. But let's not get lost in the personality here. Mm-hmm.

It's the message. I don't think we've seen true AI disruption just yet. I think what we're seeing here is the tremors before whatever.

Yeah. But when stuff gets real, we're going to need government and private industry to work together because there are going to be casualties of this. There are.

Mm-hmm. " I don't know if- Because that's how revolutions happen So with this group of leaders and luminaries, what do we think we should be doing to help? What is the call to arms here?

The call- Why- ... to arms for both parties, frankly, when it comes to tech and business and AI, are idiots. Yes.

None of them know what they're talking about. So what needs to be created is something that feels like an independent bipartisan commission that's going to talk about how are we reinventing the American economy and then we'll- You forgot the word blue ribbon. Right.

Something. Just to start the conversation. I think that would be good.

The Warren Commission of sorts. But some- They don't have a choice right now, though. This is a grassroots movement.

This is happening in all these midterm elections. This has become a big deal. It's not just data centers, it's big tech.

It's going to be this meta judgment, settlement, excuse me. It's going to be these massive IPOs and acquisitions. The politicians are going to have to learn fast.

And I think they're trying, as incapable as many of them are around tech. It's- So- To Alan's earlier points in some of these conversations, though, there is an opportunity for somebody to emerge who kind of looks and smells like Teddy Roosevelt and maybe takes this bull by the horns. Yeah.

If you look at, and again, I'll go back to the Indispensability Trap, which will be out next month. Yeah, September 9th, hopefully we're done with editing. But what's happened in the past, you look at electricity, you look at phones, you look at the railroads.

When it became so indispensable and such ingrained in there, the government did step in and regulate. They regulated it, and for good or bad, it worked for a long time. In the case of AT&T and the phone service, for about 70 years, we had the best phone service in the world here in the US.

Just Ma Bell And then eventually they did break it up and now it's reassembled into two companies. I think this is going to be litigated heavily. We just saw Meta get slapped with $17 billion in child safety- They didn't get slapped.

They acquiesced. They agreed to that, I think because they thought the exposure was far beyond that. Right.

And so this is how I think we're going to end up regulating this is through tons and tons of lawsuits. Yeah, but what the-- I don't know, Anne, you're much younger than me, but- I am ... remember the IBM settlement, the AT&T settlement, Justice Harold Green.

These were all brought as court cases by the government to-- And even the Meta one is 47 states, the District of Columbia, and a couple of territories. That's pretty much damn near everyone. Meta got to the point it got because Zuckerberg was always preaching self-regulation.

"We can handle it. We'll take care of it. " We've seen that for years.

But that's what the tech industry always says. They all say that as they push it full speed ahead on all this recklessly. " We've heard it before.

I think it was in Cat- Tell us, Tom ... where he told me self-regulation will make you go blind, right? Something like that.

Yeah. And hair on your palms. Let's move on to the next thing here.

Wow. Block three. Woo.

All right. Moving on to the next block. We got more funny money floating around.

So Broadcom is trying to negotiate some sort of deal here involving how they're going to finance some of their investments in AI, and Alan's been writing about this stuff aggressively over the last few, I guess two months or so now. But Tom, what do you think is going on here? It seems like all this money now is being funded by bonds, and the bond market is shaky, and so is this whole thing starting to feel like maybe it's on a house of financial cards?

What do you say? I think that's why they're trying to do this with these bond-like structures is because they realize that there is a limit to what they can do, and you can't just make money out of nowhere. Unless you're the government, in which case you still can't really do it, but you can.

And that's the problem is there's only so much money to go around. It feels like the same $100 billion has been traded around eight different companies at this point. Like all you got to do is mark those bills and find out where they're going.

But Broadcom is very quickly following along from what Nvidia wants to do with this idea of creating a security that's backed by these financial instruments that's based around GPUs because then it puts it into a different category. This is no longer just buying assets and putting them in and all this other stuff. It's now a financial instrument which can be traded, which can be resold, and more importantly- Like mortgages?

You had to say the M word, Alan. But yeah, that's exactly what it is. Because unlike a house, GPUs are not going to be worth anything in two to three years.

And we've already seen that with Nvidia, where people are saying, "Well, hold on, you're saying that people are keeping the GPUs longer because they're more useful, but you're also telling us in the same breath that people are upgrading faster for new capabilities? " Well, it doesn't matter at that point if what you're doing is kind of putting it into a financial instrument because then people don't care about what's underneath, they just care about the financial instrument itself. And Broadcom may be one of the only companies left out there that has the, I don't know, objectivity to be able to challenge this.

Because before you say AMD, I think AMD has kind of drawn their battle lines about where they're going, and the answer is two steps behind Nvidia the whole way. So if you're not willing to go down that road, and it sounds like Anthropic is looking for other alternatives, Broadcom is one of the few left, other than maybe just going directly to the source with TSMC, that you can work with. So the key difference between Broadcom and Nvidia, Nvidia is using its balance sheet as a moat, as a weapon.

They've weaponized their balance sheet. Broadcom's tried to do that, but a lot of people on Wall Street are losing their appetite because what Broadcom has done, Broadcom has vouched for their customers with their own balance sheet, right? But when you look at their OpenAI deal, for instance, when you look at Broadcom's book deals that they're counting on in terms of revenue, an ungodly amount comes from OpenAI alone.

If OpenAI doesn't buy the hundreds of billions of dollars they've committed to with Broadcom, something like 60% of Broadcom's future revenue is pegged to that. Wow. Broadcom is not Nvidia.

But here's the thing about this whole big piece. Again, indispensability trap. In years, in other grid build-outs, in other technology waves, one of the reasons why there was a bubble burst in them was that it was done on other people's money, it was done on borrowed money.

What was supposed to be different this time is that this build-out was being built on the ridiculous free cash flows of Meta, of Amazon, of Google, of Microsoft, even Apple perhaps, right? These companies have free cash flow like nothing we've seen in history. But this AI build-out has gotten so God darn big that even the cash flows of a Google or a Microsoft or an Amazon are not enough.

Google was cash flow negative this quarter. Because they spent so much on this build-out. Microsoft was close, but Microsoft's a machine.

Amazon, it's the same thing. Meta, same thing. The amount of money they're putting into this build-out is even beyond their ridiculous amount of free cash flow.

And so the giant hordes that they had built up of cash reserves are going down, and they're out here now looking for financing. And some more creative than others. So Tom- Oracle, too.

Yeah. Tom, I can't remember if it was Lenin or Marx who said something to the effect of, give a capitalist enough rope and they'll hang themselves. It was Marx.

And the question becomes, as you look over at what's happening in China with these open weight models, but they're also signaling that they're going to build advanced processors, and they expect to have those in the next two or three years. And they will probably sell those for a dollar for what NVIDIA's charging tens of thousands of dollars for. " You would think that, unless the invisible hand of the market falls in and gives you a middle finger, and we're already starting to see that, right?

" Or, as we know for a fact, the lithography machines that are available in the Netherlands can't be purchased by the Chinese, so they're kind of two steps behind unless they can figure out that technology- But they're building those, too, Tom. They are. This huge push.

But the question is, can they catch up fast enough before that next generation comes out? They just got to be good enough. They don't got to catch up.

Which is true, but that then causes the rest of this thing to be a discussion of can all of this irrational consumption outlast everybody else when you think about, well, if we just spend a little bit more money, if we take out a little bit more debt this quarter, then we'll win every quarter ad infinitum forever. Because that's basically what you're saying. You either have to keep doing this until you win or outlast, what, everybody but NVIDIA?

Because if NVIDIA is using their balance sheet as a moat in this case, can you swim across that moat before you drown? No. But so what- At what point does this become sunk cost fallacy?

At what point, right? Yeah, but here's the whole thing. This whole house of cards, maybe house of cards is the wrong word, but this whole thing is built on the scarcity of compute.

And Mike, that's what you're talking about with the Chinese here. If you can remove the scarcity of compute from the equation, the whole thing starts collapsing on itself. But and Anne, when you look at the sovereignty movement, primarily in the EU, do you think at this point in the game they care whether they're buying their GPUs from China or America?

Nope. Some of them like China a lot better these days. So when scarcity ceases becoming the deciding factor here.

Right now, Jensen can sell more GPUs than he could make. But when there's more GPUs available than we could use, the whole equation flips on its head. But aren't we there right now, Alan?

No. Because it's not a question of how many GPUs Jensen has shipped, it's what they're doing right now. And what I understand- No, Jensen said yesterday, if they had more supplies, they'd be able to ship more.

That's their gating factor, is they can't make any more GPUs than they're making. So are people just buying these things and stockpiling them because they can't build data centers? Amazon just bought out the future of their 200 million or 2 million GPUs for next year and the year after.

Because that's what the scarcity of this. Jensen can't sell more than he's selling. But- Alan, I've got some tulips back here for next season that I think you should really look at.

That's exactly. I don't know if you know about the tulip, if you've ever been to Amsterdam and gone to the tulips- Of course ... and everything.

Right. The only- This is what happened ... the only industry- But are they eating their own seed corn?

Yeah. The only industry I've ever seen that was able to enforce a scarcity model for business purposes long term was De Beers in the diamond industry. But other than that, I've never seen- And that's changing with lab-based diamonds.

They are falling apart because it turns out that when you can make something and you remove the inherent value of that thing- Scarcity ... then it goes away. It's all scarcity.

" Read it when it comes out. But we've got to bank. We're done here, guys.

I'm sorry. We've had a great discussion, but I got to pull the plug. I wish we had another hour to jump into this, but.

Tom, always a pleasure having you here. Anne, don't be away so long. I know you're busy with the TV stuff, but we need you.

I'll be around, Al. I'll make time for you, Alan. All right.

John and Mike, always a pleasure having you on. I hope you've enjoyed today's "Techstrong" gang. tv, Techstrong TV YouTube channel, the OTT channel for Roku, Amazon, Apple, Android, iOS.

If you've got a screen, you can watch us. But until tomorrow, thanks for joining us here on "Techstrong" gang. We're out.