All of it's unprecedented and historic, and the numbers are massive. But I agreed with your conclusion that one of the key results of the past few weeks is Anthropic's efforts to seek regulatory capture are in a better place than they were two weeks ago, specifically because the China hawks have stepped forward. And that's a key part of their message. And so, one key question I have is: is Nvidia capable of hiring a CEO that has the strategic chops to understand everything that you just put on the table? Because I do think it's imperative to continue with Neumetron and the efforts that Hugging Face and their team have been putting together. And it'd be very easy to hire someone who didn't understand that component and not invest in it, because it's not revenue.
But as I've been talking about, you know, it was really important because Jensen Huang had a business interest as well, right? Nvidia does better the more that it can spread out these bets, the more labs there are. Do you think, I mean, where do we go from here? Who can come in and sort of understand that and realize that the fight in Washington is just as important, especially after what just happened, sort of learning about all of the powerful voices in Washington and the China hawks that wanted a different policy?
Yeah, look, I've written a couple of long pieces on how using open source strategically in this way is what I consider to be a very high-level strategic execution. And in many ways, you're playing defense. Like you're trying to prohibit someone from establishing a monopoly position. And Google's done this in the past. Google did it with Kubernetes when they were worried about Amazon and AWS. They did it with Android when they were worried about Apple having a monopoly in phones. So there is, for companies, using these open source strategies to help prevent monopolies. But not everyone understands it. I've tried to pitch it to certain CEOs who just don't rock it. And so I don't know how many of the candidates people have put out there on the list are sophisticated enough to keep doing it. And I think it's a big concern.
Certainly where I am in San Francisco, and I talk to a lot of startup founders, they believe passionately in open source and democratizing AI and having it be open, particularly based on sort of past tech revolutions, right? You bring up Google, which is really interesting. They do have open source models. They have Android, which was able to, you know, get an American operating system all around the world. They even are creating their own chips, the TPUs, that are gaining ground. Do you think that they can be an effective voice in Washington to advocate on some of the things that Jensen was advocating for?
One thing I don't know is who inside of Google drove the Kubernetes decision, you know, and because if that exact mindset were brought to the AI, you know, strict fight, competitive fight, and if Google recognizes that Anthropic, you know, and OpenAI maybe are getting escape velocity away from them, it is the number one thing they should go do, you know, bar none. So I'd be hopeful that they do it. The problem is I don't have any data personally that would suggest they're not executing the way Nvidia was around Neumetron and Hugging Face. We're not seeing those actions from Google. So I think it would be a new gear that they'd be switching into if they were going to go do that. I'd love to hear that from them, but I haven't so far.
Yeah, I mean, hopefully we can hear that from them, right? If they've been producing and releasing open weight models. What about some of the startups in the space that have been gaining traction? We should note, thanks to Jensen, right? There's Reflection, which is an AI company that is building open source models, but they were backed by Nvidia under Jensen Huang. There's also Thinking Machines. Do you think that they can carry the mantle?
Well, one thing that's difficult is for anybody to do it on an island. I mean, Google may have done that with Android, but they had the weight of Google behind them and they convinced the telecom industry and a whole bunch of other players that this was in their best interest. So one thing I fear about just an independent player making it happen is they're fighting by themselves. If any of those players were to work with a partner like the Linux Foundation, which is exactly who Google worked with with Kubernetes, and they could get more people to pile in on top of their effort. There was a rumor that Meta, at one point when they were really pushing Llama, kind of went around and searched for partners and weren't able to get there. Now, we don't know if they were willing to give up control. When Google did Android, they kept control, and in these open source efforts, there's a continuum of how much control one might have depending on the terms of how it's structured. But with Kubernetes, they gave up all control. So I don't know if any of those players would be willing to, you know, take an effort like that. But if you could imagine Google, Meta, and both of those startups all working on the same platform, all sharing weight, all doing code development, and then the startup community could get behind it, that kind of thing would have a 10x higher chance of success than somebody fighting by themselves. And that's consistent with how other open source programs have worked, like Linux.
One of the points against open source is that it's difficult to monetize. And I mean, it's amazing to me to look at a company like Meta, which pioneered the space in America with Llama, and then just do this like hard turn away from it. Do you think that is sort of this monetization question, even though we've seen the Chinese labs able to monetize through things like APIs? Why has the industry sort of gone away from it? I wonder too if like you can draw that connection too between the AI industry moving aside from Nvidia and a few others that we've talked about moving away from open source, and just this hawkishness in Washington against this, which is so different than, as you point out, other technological cycles that have embraced open and led to America winning.
Yeah, I wrote a piece this past year where I argued that it's just the natural evolution of competition in software. The marginal cost of software is zero, and so if you believe in pure competition, you know, the marginal price is going to equal marginal cost, and that's where you end up. And admittedly, you're not going to make nearly as much money as you would if you were able to be equally successful with a proprietary product. But one of the reasons that China ended up in a place where everyone was competing with open is they're not the incumbent. They're the aggressor. They're the ones coming at it. And so it's the right move for someone that wants to be aggressive, or it's the right move for, as I said before, someone that wants to play defense. But if you're worried about the prosperity of the individual, of the startup ecosystem, of innovation, you shouldn't care about that point, and it certainly shouldn't be confused in the regulatory argument. Like, you shouldn't say, 'Oh, but no one can make money, so therefore it should be illegal.' That makes no sense whatsoever.
Right. And I mean, look at the adoption numbers, right? You've just seen the explosion of open weight models gaining adoption this year, particularly among developers, but also increasingly enterprise. Bill, what happens next if we do not have this voice in Washington, but also, you know, the quote-unquote bank of Nvidia? Jensen, you know, essentially created like the neo-cloud space. He spread out all of these bets. He decided to give them allocation and capital. What happens next, and do you think there's a path to him coming back?
I don't know the answer to that question. There's a whole bunch of data that I think none of us know, and there's a lot of armchair quarterbacks trying to figure it out. And you know, in the past four days, it's become quite clear that Washington D.C. cares a lot, and that Rubio cares, and Bessent cares. And there may just be a whole bunch of conversations that have happened between the company and the State Department, you know, potentially prior to the incident, and maybe after, that would seriously impact how the board, you know, acts. And I just don't have the data, so I don't know. The recent — there was someone posted something 30 minutes ago that said he's not coming back. I don't know.
The board says they don't want him to come back, right? But I guess you've been in that situation before, too.
Yeah, and we don't know — like, we don't know that if the State Department had had a conversation with the chief legal officer or the board, that would drastically change how someone would think about what to do in this situation. So the Wall Street Journal seems to want to imply that there's a tie between what's going down with Jensen and what people are calling the bank of Nvidia, which you just hinted at. And they published two or three articles in the past three days. This stuff's just coming at us so fast. That suggests that there was internal concern about how far, how aggressive the terms had become on the marginal deal, and that that was starting to come back. And you know, there was this thing in the Q2 numbers where the accounts receivable increased dramatically and the free cash flow fell, which would be consistent with this notion that the terms were getting more and more aggressive. So the earnings call tomorrow, you know, that's one thing people are really going to be looking at is to see if that continued or not, and to see if there's any commentary whatsoever on this money coming out of here. Now, there is this conundrum in that if these deals weren't circular and they were just, you know, equivalent to all the demand that was in the market, then there should be no concern about them going away. If we're concerned about them going away, then they reduce in revenue. Like, you can't have it both ways.
Where do you fall on that? Were these deals circular and therefore too risky for the markets, for Nvidia shareholders, for new management, or were they, you know, I'd say like the bull case, like necessary to build the American industry, to, you know, spread out all of these bets?
I know I don't have access to the exact precise terms, and there may be this scenario where, you know, you shouldn't be able to use your balance sheet to drive your income statement. That just shouldn't happen. And the way that you are supposedly clear or clean is if the two contracts are completely separate. But it'd be possible to skirt the technical legal boundary and still violate the intent of those rules that were put in place. And I would just say this, and I don't think it's Nvidia specific, because I think Google did it and Amazon did it and Microsoft did it. I suspect when this is all over, the accountants will come to the table and suggest they should have been more — aggressive's not the right word — more restrictive of what was happening. I do think it got too far.
Right, and now you've got a CFO leading the company. So, you know, if you take what you're saying, she's going to take a hard look at these deals. What does that mean for the AI trade as a whole? Is there real risk here in the public markets of, you know, the bank of Nvidia pulling back and companies like the neo-clouds and others having a hard time?
Once again, I would state that everything in the past, you know, four years is so unprecedented. These numbers are so gargantuan, it's hard to even fathom or put in perspective. The amount of money that the hyperscalers have spent on CapEx — you know, there's this great chart that's been going around showing Google's free cash flow over time, and it just went straight up for 30 years or whatever, 25 years, and then it went negative the past two years. And this happened to all of them, except for Apple, as you mentioned in your intro. And so, you know, we've seen hundreds and hundreds and hundreds of billions of dollars put into these CapEx investments around these things. And, you know, is there a chance that it could go too far? Well, we used up all the free cash flow of all the big players. We then started using debt. We know there have been off-balance-sheet commitments. There was one where Meta did. The debt rate, you know, supposedly on the neo-clouds has started to go up, and then the AR pushed out at Nvidia. Is there a chance that we went up to the edge? Sure. There's a lot of data that suggests we probably did. And the key question would be, is can it come unraveled in a way that's kind of fast or not? And I know what the bulls would say. They would say at the end of the day, the demand for this technology is real. And Anthropic's numbers are unbelievable. To have this growth rate at this size so fast in the history of this company is just miraculous. Like, I'm completely in awe. And I know that's what they would say. So maybe it's a settling down where you try and find the bottom and grow at a rate that's consistent with the IRR that's on the table.