OpenAI vs Anthropic IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts

Welcome back. Number one podcast in the world. It's July, All-In episode 280. Friedberg is on a little vacation; we'll leave it at that. Bestie Brad is here. How are you doing, Brad?

I'm doing great. I'm on vacation in maybe Idaho or somewhere, Jason.

I know. Who knows? Who knows? It could be anywhere. He could be anywhere. I mean, there are lots of places he could be.

I'm in my flag room. Very patriotic room here.

Very nice.

You know where I work on the East Coast in the summertime, and I spent some time in D.C. this week. It's been a great week celebrating America 250.

Great. And you're going to be out of there by the second week of August?

Yes. I have it from August 10th through the 30th. I'm good. Exactly. Jason B&B.

Oh, absolutely. You don't know the half of it, man. I am on a summer bender. I'm like, "Where's your vacation?"

By the way, where are you right now?

I am in Paris. I did about 8 interviews at the RAISE conference. They'll be coming out in the All-In feed. And, of course, Chamath from the factory. Look at him. You're working in the factory, Chamath Palihapitiya. It's going to be a hot software summer for Chamath. How's your hot software summer gone?

It's good. Selling enterprise software is hard, but it's good.

Chamath's like, "Man, I was such a dick to all my CEOs in the SaaS period." And now, you know—

I went to Geneva. Shout-out to Marc Benioff. He works out of Europe, sees all his European customers, and he had a dinner in Geneva, which I joined. Then I, Jensen, Brad Smith, Anthony Tan from Grab, and a bunch of other folks were put on this U.N. Commission for AI that Marc is the co-chairman of.

Holy.

When you see Marc Benioff in action, this guy is a master. He is the empresario of empresarios. Yeah.

You see how he's built such a ginormous business. It's impressive. It's impressive. What is a commission by the United Nations for AI? What is their calling?

Open source.

No, but I mean, do the United Nations actually do anything? Do they actually? It's so funny. Anthropic was there, too. One of the co-founders, Tom Brown, I think was his name.

Yeah.

That's good. Was the Anthropic guy running around saying, "It's the end of the world. It's the end of the world"?

No, no, no. He was very awesome. Tom's awesome.

In fairness to him, he wears it on his sleeve. It's like, "Hey, we really believe we're doing the right thing, right? Just trust us." And I think the future is open source for all these countries.

Well, we're going to get into that. That's on the docket for sure. But let's start with the IPO update. There's a trillion-dollar IPO rush to the exits, and this was a big topic of discussion, Brad, at the liquidity summit last month. We'd never seen a trillion-dollar IPO. We had one this year already.

SpaceX is trading right about where it went public, so it was priced, I guess, to perfection. We're theoretically going to see 2 more. OpenAI and Anthropic are slated to go out. Brad has the inside information, so I'll try to get it out of him.

Let's quickly go over what happened with SpaceX. It ran up to $200 a share, and it's been down a bit. It's at $150 a share. As I said, that's right at the IPO price, so it's trading at that $2 trillion market cap. Currently, it's the seventh-largest company in the world.

Anthropic confidentially filed on June 1. I don't know why they call this a confidential filing when it immediately comes out, but I guess the information is confidential. Polymarket says there's a 65% chance Anthropic's IPO will happen this year, on light volume of $360,000.

Two weeks ago, Gavin Baker, another bestie, said he thinks they're going to end 2026 with over $100 billion in revenue and be very profitable. He said on a couple of episodes of the program that he thinks it would trade at $3 trillion right now if it went public.

Chamath, you made a great call on the pod. You said, "Hey, good idea for Elon to get out first." What are the chances that these other 2 get out this year, or maybe in 9 months, in the first quarter of next year? We'll start there.

I think these are all great businesses. The question is, what is the market-clearing price? That's more a function of how much appetite the markets have to absorb new issues and at what scale. That's number 1.

I think that's mostly determined by price. Anthropic and OpenAI are probably in 2 different places. The last time we heard from OpenAI, their cash burn was still quite high, just because of the diffuse nature of their business and their greater reliance on consumers than enterprises. I think Brad mentioned on one of the pods that Anthropic may actually be accidentally profitable. I think he said something like that.

Let me tell you something really interesting. I sat down with my CTO today and I said, "How are we doing on token spend?" He said the most incredible thing: "Right now, our token costs are doubling every 45 days."

I was like, "Gah." And he said, "Yeah." I said, "What is the downstream productivity?" He said, "Maybe 5% max."

I said, "Okay, so my costs are doubling every 45 days, and my upside is essentially flat." He said, "Basically." I said, "Explain why that is." He said, "Honestly, what we're finding out is that you need to use a lot more tokens to get to this next iteration of improvement because we've effectively already asymptoted."

I said, "So what should we do?" He said, "Honestly, we have to figure this out." We're going to take a step back and try to figure out what to do. I don't know how many other companies will actually go through this reckoning now, but the point is that everybody, in the next 3 or 4 years, will go through it for sure.

I suspect that if you can get out now, you should get out now before all of that starts to seep into the water table. I think that's probably what allows you to get out at a huge price and raise a huge amount of money.

All right, Brad, you are well invested in and well known for being invested in these 2 next IPOs, so you probably have some good insights, since you talk to them on a regular basis. What are the chances they get out in the next 6 to 9 months? Both of them—would you say there's a 100% chance, unless there's some outside event, like a blockade of Taiwan or some black swan event that we're not anticipating? What do you think the chances are they're public when we're sitting here and I'm skiing in Hokkaido?

I think it's very high. But let me first say, the SpaceX IPO, where we were also investors and also bought in the IPO, was textbook. It was a hugely successful IPO. They raised $75 billion at a $1.75 trillion valuation.

It went out below where we are today. It's up 25%. Let's call it on $35 billion of forward revenue. If you think about that revenue multiple, it's trading at $2 trillion on roughly $35 billion of forward revenue. It's an incredible achievement.

I think it was textbook. Anthropic and OpenAI were watching very closely because, frankly, we had not had an IPO of that size. To Elon's credit, and to the team's credit—Bret and Gwynne—they really pioneered some smart and interesting things as part of that IPO.

You heard from Gavin that Anthropic is rumored to be trending over $100 billion in revenue, compared to the $35 billion. If they exit the year at $100 billion, that means their GAAP revenue next year could be well over $100 billion. Based on the SpaceX success, I think it would be a blockbuster IPO.

I think SpaceX has shown them the way on things like the total raise, pricing, liquidity, inclusion into the indexes, and how to do the lockup. I think they've gone to school.

It was a staged release in terms of getting out of the lockup. It has to hit certain milestones, and some of those are time-based.

Early inclusion in the index, raising $75 billion—you know.

The early inclusion in the index. Let me have you unpack that for a second, because people said, "Hey, maybe this feels unfair that they should be forced to buy it." What's your take on that? Is that just haters going to hate, or is there something to that?

I think there was a legitimate concern. The concern is that a company that had not been through the process of being vetted post-IPO has a lot of volatility. You've seen that chart, Jason: the peak-to-trough drawdown in the 6 months post-IPO is 50%. We've seen a pretty big drawdown here from peak to trough as well.

You don't want to jam it into an index at the peak and then have a 30% drawdown on top of people, which often happens in IPOs because people get excited and it runs ahead of itself. But they didn't do that here. There was fear that it was going to happen.

Both the exchanges and the indexes looked at this and made some modifications, because the other side of the argument is that it's so damn big and important that it needs to be part of the index.

Right. The reason the rules had previously existed is because most companies coming public were younger, earlier, less tested, had less revenue, were less profitable—all the things that made them less important in the overall scheme of things.

So I think that they pioneered some really smart things. It's worked well, it's traded well, and so I think that provides a bit of a blueprint for Anthropic. But in terms of the enthusiasm, is Altimeter, as a fiduciary, an enthusiastic buyer of Anthropic based upon the things we know today around profitability, model improvement, and revenue growth? Yes, everybody would be piling in. Everybody would be trying to get into the top of the book.

The last I heard—again, rumored—they would like to get out this year on OpenAI. Everybody knows that Anthropic kind of passed OpenAI on a revenue trajectory, but I will tell you, OpenAI's kind of got its swagger and mojo back. It's coming out today with a whole new set of models. We know GPT-6—there's a lot of talk of that coming out within the next 30 days—a whole new generation of models. I think their revenue has really ticked back up.

The most recent rumors I see on Twitter are around $70 billion of revenue this year. Just as a reminder, $70 billion may not be over the $100 billion that's rumored for Anthropic, but it's still twice where the revenue of SpaceX is. So can they get out at over $1 trillion on that type of revenue growth, being one of the 2 frontier premier labs? I think the answer to that is yes.

I'm not sure there's a huge race between the 2 of them to get out first. I think they'll both go out when it's time. I think OpenAI has a little bit more complexity associated with the corporate restructuring that they have to go through, so I would be surprised if they go out before Anthropic. But the fact of the matter is, I don't know. Today, as I sit here today, Altimeter would be a buyer at scale and at size in both of those IPOs.

At $3 trillion, are you a buyer, or are you, “Hey, it's obviously going to trade up and down, and there's no rush”? I think you were the one who said on the pod—or it might have been at Liquidity Live—when I asked you point-blank, “Hey, should retail get involved in SpaceX? What are your thoughts?” And you were like, “Hey, listen, it's a 4% float, 5% float. It's going to trade up and down, but a year from now, it might be trading at the same basic price. It's going to be priced not to perfection, which it seems to have been, but I think your position was it's going to be priced reasonably. There'll be plenty of time to get in. You don't have to panic about getting your shares.”

Yeah. Once a company is valued at over $1 trillion, the get-rich-quick schemes are over, right? You and I share a deep passion, Jason: We’ve got to get retail investors, the citizens of the United States, in on these value-creating opportunities earlier. The accredited-investor laws we have in this country are insane and keep people from participating in these things, but it is what it is, right?

So they're coming public at over $1 trillion. I still think there's a lot of meat on the bone on SpaceX, on Anthropic, and on OpenAI, but you're not going to have things that are—I don't expect that they're going to be priced in a way where you're going to get a 50% to 100% durable bounce out of the IPOs. If so, that would mean they were probably mispriced right into the IPO. But I do think that these things can be compounders. They're going to compound at the rate they compound revenue, and I think all of these companies are going to compound revenue at well over 30% for the next many years.

And 30% a year—for people to understand, this is high growth in public markets on very large revenue numbers already. Growing 30% when you have $100 million in revenue is one thing. Growing 30% when you have $10 billion or $100 billion, this becomes a different task.

So let's talk a little bit about these 2 companies, Chamath, and what the public's going to perceive them as. ChatGPT seemed to be the public brand, the consumer brand, for large language models. It's the AI for people who are doing their homework, or mom and dad are trying to fix the dishwasher, whatever. Then Claude took the lane of, “Hey, we're going to be the one for corporate.”

It did seem like OpenAI got very distracted with Sora and the Disney relationship. “We're going to make a puck with Jony Ive”—everything consumer. Then they realized, “Oh, wow, the revenue seems to be enterprise-first.” Is that going to wind up being the big mistake when we look at it? They kind of gave the Google position, the high-growth position, to Claude and Anthropic, and they took the Yahoo position? Or do you think they'll catch up on the enterprise? Maybe they should just go back to trying to be the consumer version. How are these going to be positioned a year from now? How's the public going to look at them?

The problem with enterprise revenue is, at some point, the person that's spending it has to see an ROI. I asked Fable 5, Anthropic's new model, what the lift of the S&P 500 earnings-per-share growth since 2024 from AI was. It answered, “Oh, it's 50%.”

Then I looked through it and said, “Well, no, you're including the money that Nvidia makes from selling chips to Amazon.” So I asked a different question: What was the EPS growth of the S&P 493? The answer was 9%. I said, “Okay, well, that's different.” I said, “Unpack that.” The overwhelming majority of that was pricing power sitting on top of inflation, and the other 3% was from buybacks. So the answer, as far as all publicly available data, was that the actual ROI was somewhere between 0% and 2%.

So I don't know. I think that enterprise looks really good. The problem is that very smart investors like Brad and Gavin and others, at some point, will start asking companies, “What's your ROI? What's the actual EPS lift?” And if the answer is, “Well, I don't really know,” or, “I'm not sure,” and you don't necessarily have the pricing power to continue to raise prices, enterprise is probably a little bit more brittle because there are fewer buyers and they're more demanding.

Consumer, on the other hand, all of a sudden becomes an incredible safe harbor because you have tens of millions of buyers, and having those 2 orders of magnitude more buyers at a much smaller price point inoculates you from the vicissitudes of an ROI discussion.

So it all really depends on what the actual ROI is of this money being spent. I think that we're in the phase of just being astonished, as Brad said, about the scale of the revenue growth.

Yeah.

But at some point, you'd have to be an idiot not to ask, “Who is paying you this? And can they sustain paying it to you?” I just don't know what the answer to that question is. At some point, it may not be now, but people will have to answer that question. You're spending $1 million a year on tokens, and that $1 million-a-year spend is doubling, tripling, and quadrupling. At some point, you're going to have to show an ROI that's—

—above the risk-free rate of return. Otherwise, you're going to have some angry investors on your hands.

Our discussion here for the last couple of weeks on the pod has centered around that, and the industry has responded in the place where all the CTOs, CEOs, and capital allocators hang out, which is X.com, formerly known as Twitter. Here's Prashanth, the CTO of Uber.

When you ask, “How are they getting the ROI out of this?” people are now bringing that conversation front and center, and they're explaining it on X. Remember, Uber was also the one that ran through all their tokens in the first quarter. On the other side of the business—which is legal, operations, marketing, customer support, HR, and procurement—he lists here: “Today, 99% of our engineers use AI tools.” Okay, great, right? Everybody's doing vibe coding and has coding assistance.

“More than 70% of pull requests are attributed to local or cloud agents. Our engineers have built 200 agentic skills.” So how are they bringing agentic AI beyond engineering? What they've decided to do is essentially—he talks about these agentic pods—and this, to me, seems directionally how this should be done. You find engineers and, as we talked about, forward-deployed engineers—a fancy way of saying, put an engineer into departments—and have them work with the department heads who understand systems thinking and how their process is done.

The long and short of it is, they're making massive, massive progress on the operational side of the business. So, Brad, you're pretty familiar with Uber and have been a long supporter of that. This is a company that knows how to deploy technology pretty well, and they're an operations machine run by an operations machine.

Brad, they should report the EPS gains attributable to AI.

They should report the EPS gains attributable to AI.

Yeah. Well, I mean, this first step seems like they're really being thoughtful about this. First, “Hey, this token spend got out of control with the developers. We're going to need to pause this and look at it.” And then, second, “Here's how it's going to lower cost and create more efficiency.”

So, Brad, let's talk about that side of it—not just token-maxing with the developers hitting the slot machine of, “Okay, let's see if this pull request produces the right code or not”—but these departments in a more strategic way. It's not just the person who works in HR using Claude Code or Perplexity or whatever and trying to vibe-code something. This is, “Hey, we're sending engineering in to work with your top systems architect, and we're going to try to find that ROI.”

Yes. I would say first that Chamath is right. The only question is on what time frame. There's no doubt that there's a lot of money being spent today that's in the experimental bucket, where I think there probably isn't direct ROI, Chamath, to your point, but I think we're so early that nobody cares.

I think we're so early in terms of enterprise adoption. Remember, the total addressable market here is every single small, medium, and large company on the planet. We've never seen revenue growth like this because we've never seen a TAM like this. If you look at the distribution of revenues across these businesses, it's not concentrated with 4 or 5 customers. There are millions of customers independently making the economic decision that's rational for them every day: that it makes sense, like Pine at Uber. Of course, they're trying to find things on both sides right now, mostly on the cost side—cost takeouts—to justify the investments they're making in tokens.

But I think we're on the verge of breakthroughs in intelligence that are going to dramatically change the revenue side of the equation for a lot of these businesses: breakthroughs in life sciences, breakthroughs in product innovation, et cetera, where they could not divorce themselves from this even if they wanted to.

For example, Jensen Huang has talked many times about how all of his design work—all of his design work now at NVIDIA—is using AI to design the next-generation chip. The machine is building the machine, so you can't get rid of that even if you wanted to. Tiny intelligence advantages at the frontier, where he sits, are required. I don't think Jensen is going to use anything but the best models he can to build out those capabilities.

I just think that we're not going to see that in the next few years. You're going to see it under the hood, of course, but that occurred at Snowflake. There was tons of optimization at Snowflake, but the revenue continued unabated. Their revenue growth continued unabated because they further penetrated use cases and further penetrated the enterprise.

Let me be provocative here: If these guys end the year over $100 billion, I think they're on a revenue trajectory where they could 3 to 5 times again next year. We've never seen anything like this. Never.

You're saying $100 billion to $300 billion? And Jason, you and I have talked about this. Our minds were blown. If a company could go from $100 million to $300 million, we're talking about going from $100 billion to $300 billion. $200 billion of incremental revenue is incomprehensible in the history of Silicon Valley.

In the history of the world. The fact that we're even talking anywhere close to this tells us something different is going on here. I think the thing that's different is that intelligence is the largest TAM we've ever seen in the history of the world. These guys are penetrating it.

So yes, the super-sophisticated companies, the 80/90[?] and Chamath, are helping optimize their token spend as early adopters. 100% that's occurring, but it's not really changing the trajectory that the frontier labs are on.

One of the interesting things about this technology that's really unique is that we talk about intelligence on demand. When you made a piece of software or had some technological innovation, it would typically accrue to 1 group of people in an organization, maybe 2 groups of people, right? Excel comes out, and the accounting department's having a field day with it, but it's not really affecting human resources or marketing. Maybe it trickles down eventually.

Every single person in every single organization is playing with these tools. If everybody's playing with them, everybody's trying to apply them all at the same time. It's kind of like you've got a 1,000-person organization. People are spending $200 a month. Chamath, they double it every X number of months. Now they're spending $400 a month per person. They're spending $5,000. If the average salary is $100,000 to $150,000 at this organization, it's only an incremental 3% or 4% on top of their salary. The way I look at it is, did it make that person 3, 4, or 5 times more effective at their job? I think the answer is yes.

That's why there's so much token maxing going on. It's also a bottom-up type of product. You can just get into this product for $20 a month, and no CIO or CTO is going to say, "You can't spend $20 a month on your corporate card for this technology." When a bottom-up technology hits everybody at the same time, that's what would explain this revenue ramp that we're all having a hard time adjusting to.

Who isn't impacted by the technology? That's my question to you, Chamath. In what organizations you're working with at 80/90[?] is there a department that says, "The intelligence on demand is not for us. We don't need it"?

Well, it's less about being dismissive that way. It's more that regulators and other people won't necessarily allow you to use it the way you want.

Okay, so finance, HIPAA—there's HR data you're not allowed to put to work just yet. What I'm finding is, once you start using this and getting some gains, it's very addictive. We were sitting here, Brad—I don't know, maybe in January—and I got that OpenClaw bug.

Then I started playing with this Hermes agent, which is not a French company, by the way. They just use French names. It's Nous Research, or whatever it is. I started playing with that. It's a very peculiar piece of software, but it's very open source, so I went to OpenRouter, got my own keys, and I've been playing with GLM-5.2.

Then I talked a little bit about Bittensor on the program, known as TAO, dollar sign TAO. It's a crypto project. Somebody is creating a subnet that is putting GLM-5.2 and other models out at really cheap prices. So I all of a sudden experienced, because they gave me an API key, having my token cost go down 95%.

When you have unlimited tokens as an exercise—which is going to come to everybody—eventually everybody's going to learn how to drop the price by 95%. It's going to happen as well because people like Groq with inference. This is all inference, right? This is what people are using. They're using inference to do this.

Inference is being impacted in 3 or 4 different ways. The software is getting better, and open source is improving at the same time. You're going to have distributed networks like BitSensor, and you're going to have better chipsets from Groq and Cerebras, et cetera. All of that's happening at the same time.

Once I got down to 95% cheaper, I started setting my agents—instead of having them do daily runs—to do hourly runs. Then I took my agents from doing 1 task and broke them up into 3 agents, having them do 3 different things on the hour. When you start doing hourly tasks, then you wake up in the morning and 14 jobs have been done, you're like, "Wait a second. This is completely different."

As one example, I have all the All-In episodes and all the This Week in Startups episodes, and we set these cron jobs to go find what the new trends are in technology. I have a trend-spotting agent running every hour, informing me of the top 3 or 4 trends, and I just give it words. It really does change your thinking.

When costs go down, what do you think the tokens are going to cost, Brad, next year?

We've seen 90% reductions in the price of tokens for each of the last 2.5 years. We've talked a lot about Jevons' paradox, which I think you're referencing here: you're going to use a hell of a lot more when it happens.

I think the central debate right now in AI is the one that Chamath keeps pointing us back in the direction of. For 18 months, since the DeepSeek moment—when the DeepSeek moment happened, the markets fell 40%—there was a reason for that. Many started arguing that the frontier models were screwed, that open source was going to kill them, that they were closing the intelligence gap, and that model routing was going to make it easier to route these tasks to cheap tokens.

But despite all of those arguments, now that we're 18 months into this—and I had this back-and-forth with Gurley a lot—the facts in the field are just the opposite. I love open source. I want all the competition in the world. Let's be very clear.

Despite all of those arguments, the share of economic value—the share of wallet—is actually increasing to the frontier labs, while the share of tokens, these commodity tokens, is obviously going up to the other guys. There was a tweet this week from Jesse Zhang that we ought to pull up here. The economic value, the share of wallet, is actually increasing to the frontier labs.

I had a little back-and-forth this week with Nikesh on this, trying to suss out why that's the case. What people would have thought is, "Cheaper, pretty damn good. 90% is good enough to do all these tasks that you're talking about, Jason, so nobody's going to use the Anthropics and OpenAIs of the world." But despite that, it looks like their share of wallet has gone up.

I think it's not that. I think it's more that when the iPhone was a novelty, everybody would keep upgrading because you expected that the new price was worth it. At some point, there's a moment—and you can debate when it happened—where people said, "You know what? I'm just going to keep the old phone because it's good enough, and I just don't see the difference."

I think there's going to be a moment like that. When I use Fable 5, the problem is that it's nerfed on a bunch of things that I would normally research.

You know, I was with somebody this weekend, and he was telling me about some health issue. I put it into Fable and said, “It won’t answer you.” I thought, “Okay.” I think everybody will get to a point, at different times, where they just say, “You know what? It shouldn’t really matter what model I’m using. If I get an answer that I think is reasonable, I can go about my day.”

I think when the corporate CFO gets involved, it’ll be an entirely different conversation. What I can tell you after this UN commission that I joined with Marc Benioff, Jensen Huang, and Brad Smith is that there is not a single country in the world that is not trying to figure out its own sovereign AI strategy. I don’t think they believe using a closed-source American model is the answer.

Mhm.

I think we have to keep in mind that there are trends. One is the geographic penetration of humans, and there are still many, many, many more people who don’t use it than do. That’s an upside and an opportunity for everybody.

The second is that the experimentation, as you said, needs to transition to ongoing, repeatable usage. The third is that all of this then needs to plug into the existing regulatory infrastructure that we use as societies to run the world. When you put all of these things together, it’s not clear to me who wins, except that you’re going to have a lot of diversity of choice.

Certain countries—I can tell you after this week—have no desire to subjugate themselves to any technical risk, and so they’re willing to spend the money to have their own. We can argue and debate whether that country has any chance, but they would rather take an open-source model like NVIDIA’s and stand up their own stack, soup to nuts, for their own people and their own companies inside their own country. If the models are 99% as good, or 95% as good, there’s going to be a claim that some countries make, which is, “It’s just good enough.”

That’s the question. That’s the question.

Separately, there are companies that will not have the earnings growth to justify this without going through some long, protracted carve-out of costs. Most companies—you know this—just don’t do it. They don’t have the nerve to do it. They’re not capable of it.

You wrote that famous essay to Zuck, and he was pressured into finally doing it. Aside from a very few companies, most people just allow the problems to compound. I just don’t see a world where, when you get clobbered over the head, you don’t look at other ways of displacing cost. If it’s a Coke-Pepsi kind of thing and Pepsi is one one-thousandth the cost of Coke, I don’t know. I just think it’s a risk that has to be managed in the perception of the market participants and the underwriters.

To add to that, Brad, open source is very hard to implement when compared with just firing up Claude and having Claude already approved in your organization. The number of steps it took me to configure my new setup and get onto this Bittensor network to get OpenRouter going took hours, and I’m pretty familiar with technology.

To your point, Chamath, it dynamically routes now. I’m dynamically routing and using GLM 5.2, too. Then, if I fall back to Claude, which Claude am I going to fall back to?

Here’s another piece of evidence to your point, Chamath: There are some organizations that just aren’t capable of this. They don’t have the team that does this naturally. We just talked about the CTO of Uber. Now let’s talk about another CTO. Andy Fang is the CTO of DoorDash. Shout-out to Stanley.

As you can see here in this tweet, a lot of people listening to All-In over the last couple of weeks are coming out and, as I said, explaining what they’re doing to address this exact issue. He says, “With our internal coding benchmarks, we’re able to confidently introduce open-weight models into our AI code review without degrading code quality. Have the frontier model Fable from Anthropic do the hardest work, delegate lower-level work to Kimi K2.6, and they are now releasing their benchmark.”

Another group is releasing its benchmarks and saying, “We know this is an issue.” The CTO has been charged to your point, Chamath. The CFO says, “Make sure this is profitable. We get the ROI.” They put that on the CTO. Here’s another CTO from a leading tech organization that knows how to implement this.

The really interesting thing we have to forecast right now is what happens in an earnings miss, and what happens in a moment where, for whatever reason—maybe there’s just an externality—there are a series of earnings misses. Where are people going to look? I think people find it very difficult to lay off other people. It’s much easier to cut other costs.

The more successful these companies get in a very quick amount of time without really proving the ROI, the bigger the risk is. It’s complicated. The game on the field when you’re working with these enterprises and just trying to explain it to them is that they’re getting smarter quickly.

I would just say Chamath’s absolutely right about the sovereign stacks that are going to get built around the world. This is not either-or. We are going to have open source, and we are going to have frontier intelligence.

The preponderance of the tokens today is already shifting toward cheaper, lower-tier models out of OpenAI or Anthropic, or the other frontier labs that are out there. Obviously, we talk about those two. SpaceX has released an incredible model in the last 2 days. Meta’s out with a terrific model today, and Gemini is still in the hunt. So there’s lots of choice.

The Meta thing was really intense because I thought, “Okay, we talked about the game theory, which was that Mark should scorch the earth with open source.” I think they flubbed that play. But then I think he has now said he’s going to create a price war. If you look at the tweet or the quote—there was a post, Jason. I don’t know if you can find it—he was basically announcing, “Hey, guys, I’m going to give you the same quality at one one-hundredth of the cost.”

There’s a lot between here and there. There’s a lot of enterprise distribution that’s required, and you know this: There have been a couple of misfires before. But I thought it was interesting that the vector of challenge was on cost.

I got it.

Yeah. Here’s the Mark Zuckerberg tweet. Sorry, I queued it up for you there, Brad. He's at D—that was his old handle back when he was in college—and he’s done more tweets today over this Muse Spark announcement than he’s done in his history. So he’s getting into the X conversation.

Quote

“Today we’re releasing new Spark 1.1, a strong agentic coding model at a very low price. It’s available through our new Meta Model API and in Meta AI.” He’s coming out saying, “Hey, we’ve got the strongest agentic tool here. Please come use it.”

He also wants to have his own, essentially—he wants to jump into not the hosting space, but he wants to provide tokens as well. So again, I think we’re going to have a tremendous amount of selection.

The competition is great for America, but if you look at the things people are doing, let me give you an example. The premium workload—you talked about summarizing a document—may take 20,000 cheap tokens to do. Of course, shoot that to a lower-tier model or an open-source model. But if you’re talking about replacing a software engineer for 2 hours, that may take 2 million expensive tokens.

The consequence of using something that’s 95% as good is really high, because you have a long-running task. If the task breaks early, in the middle, or at the end, there’s a huge cost to that.

You still burn the tokens, right? You’re pulling the slot machine, and you lose.

And the time and the compute. If an AI agent is replacing a $200-an-hour consultant, take that as an example. Three consulting firms are competing, and they need the smartest consultant. If they’re charging $200 an hour, the difference between spending $3 on a cheap model or $15 on an expensive model to replace a $200-an-hour consultant is irrelevant. That inference-cost difference is irrelevant if you’re getting something that’s bulletproof for $15.

I think that’s what we’re seeing play out. The best evidence for all of this is revenue growth, right? We can sit here and speculate all day long as to revenue—not from Anthropic and OpenAI, but from their customers.

No, I’m talking about what Anthropic’s revenue growth is compared with OpenAI’s and compared with the open-source models. Millions of independent actors are choosing every single day. The open-source companies are growing, but they’re growing by selling something that is really, really cheap.

There’s room in every single market for premium products, mid-tier products, and commodity products. I think we see a lot of this token growth. People are speculating that the intelligence gap between that commodity stuff and the frontier stuff is going to collapse to the point that people won’t pay for the frontier stuff. There is no evidence of that on the field today. It may develop over the course of the next couple of years, but it’s not on the field today.

Yeah, yeah. To give people an idea—we keep mentioning what sovereigns are doing—let me give you the specifics. The UAE, very famously, has its own Abu Dhabi Technology Innovation Institute shipping Falcon.

You probably have heard about that. The Saudis have Humane, and they’re doing their own models that are Arabic LLMs. And then this week, Japan is investing $6 billion in a consortium. It’s called the Neoterra consortium, and they’re doing that and skipping ahead to physical AI, i.e., robotics.

Okay, joining the conversation here, the one, the only Saxy Pooh Sacks, bringing you into the discussion. Talking a little bit here about the debate that we started on the podcast: getting ROI from tokens. Where are the tokens going to accrue—to open source versus the frontier models? A bunch of CTOs chiming in on X this week, in the last couple of days, are in fact talking about how they’re managing intelligent routing: first to open-source models, then falling back to Fable and the frontier models.

How do you think this is playing out? And if you’re an investor in the space, how do you think about the frontier models and their growth when you have CFOs coming in and saying, “Hey, justify this cost. Do you have a cheaper solution, and what is that cheaper solution?”

Well, look, I think that enterprise CTOs would like to shift their token consumption to cheaper models for the obvious reason that that would be more efficient, and they’re seeing their compute cost, or their token cost, skyrocketing right now. So everyone’s trying to figure out, how do we put the brakes on this, or at least control it and make sure we’re getting ROI?

You also have the AI sovereignty issue that we discussed last week, that Alex Karp talked about, where they’re worried about giving up the secret sauce or the alpha in their business to a frontier lab that may one day be competing with them. So there’s no question that enterprises would like to diversify. They would like to get off of these frontier models when they can.

The problem is, I think in most cases, they don’t have the technical ability to do it. I mean, Coinbase figured out how to do it. DoorDash figured out how to do it, which is to say they built a token-routing system—a layer of middleware that allows them to send frontier tasks to frontier models and non-frontier tasks to more mundane models. But I don’t think your average enterprise has the technical capability to do that.

So I think this is a case of the spirit being willing but the flesh being weak. They are willing, and they would like to diversify off of these closed models, but they’re unable to do it. So this is why the share of wallet of closed models actually increased. I think open source went from 19% last year to 11% this year. So open source as a share of enterprise spending is actually decreasing.

Now, I don’t think that means usage is decreasing. I think usage is skyrocketing in both of these categories. It also may be the case that, because the whole point of using an open model is that you just pay for the compute cost and don’t have to pay a lab, it’s hard to measure that usage in terms of spend. But nonetheless, anyone who’s saying that these closed models are going to lose, or are somehow losing, you’re just not seeing it in the data. Like Brad’s saying, the revenue is skyrocketing.

I think the most you can say is that enterprises that are technically capable would like to gravitate toward hybrid architectures. At the same time, it takes technical expertise, and it is just phenomenally convenient, whether you’re a developer or an enterprise, to go with the frontier labs. That’s why their revenue is skyrocketing.

It is the easiest choice. It’s the most refined product.

And there’s one other thing here as well. This was discussed in a really interesting blog post by the founder of Decagon, which is enabling AI-powered customer support for enterprises. What the founder said is, “Look, open models are great when you know exactly what you’re trying to do. They’re smaller, cheaper models, but you have to do post-training, you have to have the data set, and you have to know exactly what you’re going to use them for.”

Totally.

But if you don’t know exactly what you’re going to use them for, you want the most powerful general intelligence that you can get. Right? So what he said is that for mature use cases, yeah, you want to go open, but for immature use cases, which are all the new things people are discovering right now, you’re just going to want to use the most capable general model that you can.

And then, once you figure out what the workflow is, what the workload is going to be, and exactly what you’re trying to accomplish, then you can use a small, highly trained model. And I think he said, “Optimize in the prompt to get the gain” for customer support. Your model doesn’t need to know physics, for example, and so you don’t need that capability. But enterprises are still trying to figure out exactly what all these workflows are going to do.

So I think that’s another factor, which is to say that it depends on the use case and how mature that use case is. You really want the most powerful frontier models that you can at the discovery of all the potential for the technology.

Yeah.

And then let me just wrap up. There’s one other interesting post that I saw by Nesh Aurora, who also said that, yeah, enterprises would like to diversify. They would like what he called model fungibility. They would love to commoditize these models, right, and just hot-swap them.

Yeah. Headless is the term being used, right?

Yeah. That’d be ideal for enterprises: you sort of swap out the model for the cheapest one that gets your task done. But then what do you do about memory? What do you do about context? What do you do about history? What he said is that no one’s really figured out a way to abstract that stuff away from the model yet.

Again, this goes to the technical challenge of creating this middleware layer that would do the most efficient token routing. It doesn’t work unless you can make all of that context, memory, and history fully portable to the cheaper model that you want to hot-swap to.

Which means you have to have some technical ability, Sacks. The people with technical ability—the tip of the spear, the 1% of people deploying this technology—are starting to figure that out. Here’s another proof point and some more evidence. This is Ali, the founder of Databricks, I think a company you’re very familiar with, Brad.

What he realized when you start taking apart the harness and start looking at the skills, the memory, and all this accoutrement that you put around your tasks, he said, “We find that, using the same model—not open source versus OpenAI or Claude—the choice of harness can significantly save costs, by about 2x.”

So they found with GLM-5.2 that this performs extremely well, and that their tasks are literally getting cut in half using the same model but with a different harness. And that rings true to me. Once you’ve built one of these agents—and I was talking about one earlier; I’m running it every hour on the hour to find trends—I asked it to start optimizing. When I optimized it, it was 80% less token usage.

Now, in these apps, you can go to your analytics stacks and actually see your token use by hour, by job, and across which models you’re using. This is really sophisticated and hard for a consumer to do with the technology, but it’s definitely a trend.

So with that harness that he was using, is that something they built in-house?

Yes, I think it’s—

Yeah.

Okay, got it. So they basically created a layer in front of these, and it can multiplex different harnesses and models for different tasks. So he’s not only routing to the right LLM; he’s routing to the right harness.

People don’t even know what skills are. People don’t even know what the memory is at this point. That’s all abstracted into the Claude product or the Perplexity product, et cetera.

Yeah, there’ll be a massive business. There already is. All these inference clouds—the Basetens of the world, the Fireworks of the world—every single hyperscaler in the world is going to do this. They’re all going to provide tools that allow you to achieve some level of model fungibility.

The big question is, at the end of the day, we’re going to have—it’s going to be very heterogeneous—but what is the mix between these two? I again think the TAM is so damn big here that you’re going to have huge open-source use cases, sovereign use cases, et cetera. You’re going to have plenty of room for the frontier labs.

Let me throw something out. I’d like to get your opinion on this. To a certain extent, there’s this implied assumption in the world that there’s going to be this convergence of intelligence, right? And if you look at the benchmarks today, it seems like everybody on the benchmarks is converging. Yet, if you look at the revenue distribution, it’s not converging at all.

One of the questions I have is: will the model router itself be smart enough to overcome, David, the inherent intelligence advantages of the general-purpose frontier labs? The non-consensus argument might be that intelligence is not converging at all—that superintelligence becomes fully self-recursive, and as it becomes recursive, you actually extend the lead.

Because the smarter your model gets, the more revenue you get; the more compute you can buy, the better the model is that you can build. So I think there’s a chance that, over the course of the next 2 to 3 years, as we take on much more complex agentic tasks, the distance between the frontier and everybody else doesn’t converge. It actually extends.

We shall see. But I think there’s this implicit assumption in all the arguments today that everything’s converging.

I'm not sure that we've really run that to ground. Another piece of evidence to put into this mix: I interviewed Anton, the CEO of Lovable. Lovable is an app or service that allows you to vibe-code different pieces of software. They've got a really interesting take on that.

They went from $100 million to $600 million in revenue over the last 2 years. They went from zero to $350 million in the first 2 years of the company. The product has been out for roughly 30 months.

Then I also spoke to the CEO of ElevenLabs, Mati. I asked both of them point-blank, "Are you guys major customers of the frontier models?" Yes. They're spending tens of millions of dollars with those frontier models.

I asked them, "Are you concerned about data leakage and them releasing competing products? ElevenLabs is doing voice, and obviously Claude Code is an obvious competitor to Lovable. Are you going to make your own models?" Both of them said that they're working, essentially, on their own models.

Those are major customers of the frontier labs who want to get off the frontier models and have their own proprietary model. The ability to create verticalized models is getting easier and easier every 6 months or so. That's going to be another trend to look for: verticalized models for voice and verticalized models for building code.

We're going to see people stop using the frontier labs. These are major, major 8- and 9-figure customers. I think they're going to just run for the hills and only use the frontier models.

But Jason, the counterpoint there is ElevenLabs. I love Mati.

Yeah.

Do you really think he's going to use an inferior model? He's got to have the best voice agent in the world. And if the best voice agent in the world is given to him by using the frontier labs, can he afford, in a competitive marketplace, to say, "I'm going to use the cheaper version, the thing that I built for myself, even though it's not as good as the other thing"?

If he builds something better, I totally agree with you, which is what he believes he's doing. He believes he's making a better version. So that's the question that I just put on the table: whether or not you're going to see this convergence, whether it is, in fact, that easy. I don't think it's that easy, but we shall see.

Okay.

It may come down to how discrete and predictable the use is. Decagon, the customer support AI company, said that 90% of its usage is now being sent to open models, but those are open models that they've had the opportunity to post-train on and customize extensively based on all of their learnings and the data that they've gotten.

Again, it comes back to the maturity of the use case. If you know exactly what you're trying to do, it's probably easier. But I don't think that's most enterprises, though. Maybe there's going to be a pattern where all the immature use cases—all the things you're figuring out—you're just going to want to use the most powerful model possible.

Once it gets really well-defined, maybe you start moving some of those workloads to post-trained open models.

Purpose-built models. Yeah.

David Friedberg

Yeah, purpose-built. It could be something like that.

That's kind of what's happening with the people who are the tip of the spear. They're working on routing the jobs. They're working on the harness, and they want to be independent and have that AI sovereignty we talked about last week.

Well, but just to take on Brad's point for a second, I actually agree with what I think you're saying, Brad, which is that the market today seems to be pushing toward a duopoly, or it has become a duopoly, certainly measured in terms of revenue.

If you look at market share based on token revenue, there are only 2 companies making meaningful revenue: Anthropic at $60-some billion ARR and OpenAI at $40-some billion ARR. I don't know whether anybody else even registers. It may be the case that the more tokens Anthropic and OpenAI produce—I mean, we've got to remember that every token they're serving up is on behalf of a use case, right?—they themselves are learning from that and getting better at providing whatever that offering is. So who knows? The gap may be growing.

A year ago, it seemed like we had 5 major labs. Now it seems like there's a top 2 and then everybody else. I could see AI easily becoming another tech market that becomes a duopoly.

Which, by the way, is the trend. The historical trend is monopoly or duopoly in most tech categories, for better or worse.

David Friedberg

Yeah. In this case, though, we're using revenue as the metric to determine the winner. Keep in mind, when you're doing open source, those are dark tokens. Those don't show up as revenue.

We don't know the utilization that's occurring at DoorDash when they're using an open-source model. We do know their OpenAI and Anthropic spend, right? And because we see that in Anthropic's revenue ramp, the more they deploy these things on their own hardware, using commoditized hardware and the neoclouds, you don't see that. It doesn't come up as revenue. It comes up as free.

The only thing you're paying for there is the hosting cost, and that will come up on NVIDIA's balance sheet. So the gains you'll see there will be Cerebras, neoclouds, Crusoe Cloud, and so on. Keep that in mind when we're having this discussion.

Let's talk a little bit about sovereignty here. The CCP said that they might—or there's a report out, according to Reuters. Reuters generally does a good job of this. They dropped a couple of anonymously sourced reports about AI in China, published about 15 minutes apart.

The big scoop is that Chinese Communist Party officials are reportedly considering restricting overseas access to China's top models. Two Chinese regulators met with Alibaba, ByteDance, and Z.ai, which is doing GLM-5.2, the model we keep referencing. They're discussing limiting access to the top open- and closed-source models outside of China.

Why are they doing this? They're making any theft or leak of AI research a national-security offense, and they want to control who can fund Chinese AI labs. We saw this with Manus, which was a Chinese company that tried to go to Singapore. The CCP pulled those employees from Singapore back to China.

Their main concern, according to the report, is model misuse. Chinese authorities are deeply worried about the potential for models to exploit software vulnerabilities and that Washington might deploy a model against Chinese interests.

Sacks, last week I proposed the reverse to you in your previous position as AI czar. Do you think the United States should be banning those models? Now we have the opposite: China saying, potentially, according to these reports, that they might restrict them. So explain the game on the field here. If you're going to look into what China's thinking, why would they want us not to have those open-source models, and how is this chessboard developing?

Well, last week I explained why it would be harmful for the US to ban open models. So if you're China and you want to harm the US, maybe you would want to. I mean, it does kind of make sense.

Because our companies are benefiting a lot from all this R&D that they're doing.

Now, at the end of the day, I think the story is probably a little bit overstated. I think there are a few Chinese models that were open source that have gone closed source, but I would be surprised if they all went closed.

For example, the number-one model in China, as I understand it, is ByteDance's model, which is already closed. That's kind of like their ChatGPT equivalent, and it's always been closed.

Then you've got Alibaba's Quen, which was open and now I think is going closed, and Z.ai, which has GLM-5.2, which we've talked about a couple of weeks ago because it seemed to be catching up to what was then commercially available as the American frontier on certain tasks. I think they're going closed too after having been open.

The tactic is that you stay open until you catch the frontier or get close to it, and then there's a really compelling incentive to go closed because you want to capture all the value for yourself.

Which, by the way, is exactly what Sam Altman did famously at OpenAI. Not only did they go from a nonprofit to a for-profit, they went from open models to closed models. So it's exactly paralleling what Sam realized 3 years ago.

I mean, in a way, that was what I think Meta's original strategy was: Llama was going to be open, but then they actually sort of backed away from open a little bit.

This is kind of an obvious strategy, right? If you want to catch up, you go open. You're not going to make any meaningful revenue on closed anyway because you're not close enough to the frontier. So why would anyone buy your product? But if you go open, you get the developer community on your side.

David Friedberg

And you get utilization. More people use it, which in AI gives you reinforcement learning. Yeah, Sacks.

Well, having spent some time in DC this week and talking with both the White House and Treasury on this topic, what I can tell you is that while there may be some debates about regulation of US models, the one thing there's absolute agreement on is doing everything to stay ahead of China.

The president is very interested in how far ahead we are of China and what the things are that we need to do to stay ahead of China. It is a unifying force in Washington.

And the idea that we were going to take our frontier labs off the playing field while letting Chinese open-source models run free and, on top of that, distilling our models—I will tell you, GLM-5.2 has watermarks from Mythos all over it, right? So we know they were distilling, et cetera. And I think the U.S. government is going to take steps against distillation, which they should do.

So I think China doing this, in some ways, doesn’t hurt the United States. The United States can spin up open-source models. We’ve got Reflection AI spinning one up. Obviously, we’ve got good work going on at NVIDIA with their open-source models. I’ve talked to a couple of the frontier labs about open-source models, and I said, “Why aren’t you guys making open-source models?” They’re like, “There’s not a lot of demand for it. If there was a lot of demand for it, we’d make it.”

So I think the U.S. is in a good position. I think this is probably more chess-playing by China than an actual threat, because it would hurt them a lot more than it would hurt us.

Yeah. And then, to just back up your point, Sacks, about when you’re behind, go open, and then once you catch up, start tightening things up—that’s exactly what they did with Android, right? Google released Android. At a certain point, they were like, in order to use Android in the license, you have to include Google Search, you have to use Google Drive, and you have to use Chrome. And they started tightening it up, so it’s not really an open-source project at this point.

By the way, I think the absolute best thing that could happen for America, in terms of winning the AI race against China, is if China somehow sprouted its own doomer community.

Yes. We need a Chinese yud over there. We’ve got to get their doom up. We’ve got to get their doom up.

Exactly. We need a lot more people over there freaking out about job loss or RSI or whatever.

Yeah.

That would be the best thing that could ever happen to us: if they started cracking down on their labs in the same way that the doomers want to do over here.

Yeah.

Brad, let me just say one comment. Look, I agree with you that, from the president on down, everyone wants to win the AI race. In fact, that was the whole thrust of the big AI policy speech the president gave about 1 year ago: declaring that we were in an AI race and America had to win it.

I think the big risk is more that—and this would not be at the top level. I think if the president could make every single decision, it would be perfect. The issue is, at a lower level in the bureaucracy, do people somehow do things that are counterproductive? Maybe they think it’s going to help us in the race against China, but they end up doing something ham-fisted. They just ban something without really truly understanding all the implications of it.

So I think there’s no question that the administration wants to win the AI race. The president definitely does, and at the top levels they will all make smart decisions. The question is whether, at lower levels of the bureaucracy, you can get mistakes being made. And then you have the influence of Congress, whatever they want to do. Those guys are more responsive, I think, in a way, to the doomer community that’s creating a lot of political pressure right now.

Well, the throttle, paradoxically, to all of this might not be the software, and it might not be the chips—it might be energy. When you look at your data center projects and the other ones that are going out there, if we need more tokens, if people need more inference, we have a gating factor in the United States, which is energy.

There’s an analysis that my team put together, which I think is quite staggering. If you just look at the load growth that’s expected between now and 2050, we are about 3 entire Californias’ worth of energy short. And that’s just assuming regular consumption of devices, cars, fridges, televisions, and computers. So, yeah, we have an enormous problem in the United States with respect to electrons.

Yeah. And if you put Taiwan into the mix here, where the chips are coming out of, I had a really big wake-up call. There was a Wall Street Journal article about this. The amount of LNG, which is what Taiwan runs on, is like—they have 2 or 3 weeks of it. If China decides to blockade Taiwan, they’re going to run out of energy immediately.

So this is energy both in China and Taiwan and in the United States. It’s all dependent on that. We have to get nuclear running, more solar running, more batteries, more of everything. And that is obviously a regulatory challenge here in the United States.

All right, let’s talk about your time in D.C. Brad Gerstner went to D.C., everybody, and huge, huge congrats, Brad. You’ve been harping on these accounts, now called Trump Accounts—the Invest America accounts. Tell us what happened in D.C. this week, because I think you finally have the number 1 app in the world. Trump Accounts is the number 1 app in the world. Congratulations, and a bunch of announcements. So what are the contours of the announcement, and maybe you could take us behind the scenes?

This has been a 4-year mission in the making, thanks to you guys. You were early supporters and backers. We talked about it on here. Founders are crazy, and you guys probably looked at what I was working on and thought, “You’re nuts. You’re wasting your time on this.”

When it got signed into law last year, that was a huge moment in a founder’s journey. It’s like getting your first round of funding: “Okay, we actually know this thing is going to happen.” But on July 4 of this year, the app went live, right? That means millions of accounts got created, and the accounts got funded.

To celebrate that and really take the next step forward, we designed a joint bell-ringing—the first in history—between the NYSE and Nasdaq from the Oval Office. That was incredible. We had hundreds of CEOs there, kids there, and families that were impacted.

The president really laid out that this is much bigger than just a program to give a few people some accounts. This is really about making every child a capitalist. In fact, the president suggested that we’re going to auto-create accounts for all 50 million kids, or upwards of 70 million kids, under the age of 18. So he called on us to get the accounts open faster, for more people to have more impact, and to make sure no child is left behind.

Brad, just slow down, because a lot of people don’t even know what a Trump Account is. Just explain what it is, and then you should contrast it with a 529 account and some of these other things.

Great. So, as you guys know, the idea was very simple: $1,000 for every child at birth that could compound over their life in a privately owned investment account. So you’re born, you get a Social Security number, and you get an investment account. If you do that, you start with $1,000, somebody matches that, and you save $10 a week, that’s $50,000 at age 18.

And that’s invested in the S&P 500.

S&P 500. So when these accounts are created, all that money goes into the S&P 500. There’s no cost. It’s a free account for the lifetime of the recipient. And that was packaged into the Invest America Act, which was passed into law 1 year ago as part of the reconciliation bill.

So that’s what actually occurred on July 4 of this year. All those accounts were created for all of these kids. That’s the reason the Trump Account app is number 1 in the app store, because parents started hearing about this and saying, “Whoa, I need to go download it and get this set up for my child.”

We had over 1.5 million accounts created in the first 24 hours after the launch. We had over $1 billion in deposits. I was contributing money into the accounts of my nieces, my nephews, my kids, and friends’ kids. Every account has a QR code. Jason, somebody can just send you the QR code for your kid. You double-click Apple Pay on your phone, and you send them $25 or $50. So that is the most essential part of it.

But we also had a bunch of announcements around philanthropy. When you’re 18, 19, or 20 years old, you can start putting it toward school, or you can roll it into your IRA—

Roth IRA, I guess—your retirement account.

Obviously, Michael and Susan Dell were the anchors here: over $6 billion, $250 for each of 25 million children, primarily lower- and middle-income kids. SpaceX’s president, Gwynne Shotwell, joined the party and put $350 million in her SpaceX shares for children of lower-income communities.

So with this, is there a way to do it so you can target specific communities by geography or by, I guess, their net worth?

No, it’s just ZIP code and age.

ZIP code and age. Okay. And then Micron put in $250 million, up to $1,000 per employee. So that seems to be a really interesting way to do this. You can do an employee—I’m sorry, an employer—contribution.

And Brad did it for all kids in Indiana, I think. Right?

Correct. All kids.

Brad, this is a big number here. This is a big announcement. Brad is the guy who complains when we make him buy in for $10,000 after 10 p.m. at the poker game, and who rage-quits the game when he loses $6,000. Somehow Brad dropped $100 million. I mean, oh my God. Let’s get a round of applause and a golf clap. Brad, I’ve never heard of you doing any philanthropy.

Sometimes you show up with a bottle of wine to the game, but this is a big number. This is a big decision for you, huh?

I think this will become the largest direct philanthropic platform in the history of the country. We told the president we think we can raise $100 billion in the first 12 months. The scale of the philanthropy, and the nature of the philanthropy, is directly to America's kids without a charitable middleman directing who gets what and how it's distributed.

So you think about the people who are worth $10 billion or $100 billion. How do they give that money away at scale effectively? Now we have a platform they can do that with. It goes directly into the accounts. The money can't be taken out until the kids are 18.

There was a bunch of noise about how some people won't do it because it's called Trump Accounts. Some people said this is going to create this weird class divide, with people who had TDS refusing to give to their kids because of the name. I think your website or something said something like $13 million by the time they're 50.

I tweeted something to the effect of, “That is an irresponsible amount of money not to give a kid because you don't like the fact that it's called a Trump Account.” It's patently insane.

If you go on Bluesky, which is like the open-source, lib-TDS social network, people are like, “Just put one. You all trust those Trump Accounts? I sure as hell don't.” And so there's a bunch of people.

Speak to that. Speak to that for one second, Brad.

I would say this: The enabling legislation is the Invest America Act. A lot of Democrats call them Invest America accounts. They're officially Trump Accounts, and the facts on the ground are that parents aren't listening to that noise.

The parents who are signing up for this are across the income spectrum, across the economic spectrum, and across the political spectrum. They know and understand that their first responsibility is making sure their kids have a connection to the American dream and have savings for their life.

Yes, they are called Trump Accounts, and I've read some of that blowback. But the president himself—let me just make this case very strongly—there's nobody I've talked to about this over the last 2 years who cares more about every child getting an account than the president himself. In fact, that occupied a lot of our conversation over lunch.

The president is pushing us very hard, and the Treasury Secretary gave you instructions. He gave you an order that he wants you to auto-create the accounts. This is a brilliant move. We know the Social Security numbers of people who are under 18. He told you, “Get to work and automatically create the accounts.”

Are you going to do what President Trump has commanded you to do, Brad? Are you going to auto-create them, Brad, or are you going to disobey the president?

Our intention is to get all 50 to 70 million accounts created over the course of the next 90 days using all of this data. We have to work through Treasury, the White House, Social Security, et cetera.

You also have to get through Elizabeth Warren, Bernie Sanders, and Ro Khanna, who are going to try to stop you. Are they going to try to stop you from doing this? Are they giving you blowback because they don't want to give Trump the win, which is totally—

No, listen. I'll give credit where credit is due. Cory Booker has come out and supported these, as have Gavin Newsom, Governor Wes Moore, and John Fetterman, senator from Pennsylvania. There are plenty of Democrats who are able to get over that hurdle.

You bring up a good point, and I said this on CNBC yesterday. On the one hand, you have Bernie and Mamdani. They want to tax all these corporations. They want to control all that money in Washington and decide who gets it. It's a model that's very dependent on Washington.

On the other side, you have the president, this administration, and frankly a lot of Democrats who are closer to the center, who say, “No, let's set up a private account for every kid in America. Let's fund them. Let's not make them dependent. Let's make them independent of the government to build wealth on their own, develop financial literacy on their own, and be more likely to graduate from high school, start a business, or buy a home.”

Those are 2 very different worldviews for America. I think the antidote to more socialism is more capitalism. As I told the president, this is more capitalism.

Sacks, if this succeeds, and Brad does as he's been instructed by the president, we're going to go from 50% of people owning equities in the country to as many as 70%, maybe even 75%, of the country having access and, for the first time, being part of equity nation. What's your thought on this, Sacks?

I think that's a great thing, and I think this is a tremendous new philanthropic platform. That's really important, especially in this time of growing anger, backlash, and populism against billionaires, with people questioning whether the system is rigged, whether they can be successful in America, and whether they'll be able to be part of it.

This is a really important antidote to that. But I almost think the philanthropic aspect has gotten too much attention, because people are naturally attracted to the freebies. The part that hasn't gotten enough attention is all the comments I saw on CPA Twitter, where accountants were talking about what an unbelievable estate-planning strategy this is, or basically—

Less advantaged. Yeah.

Like a wealth-management technique, whatever you want to call it—planning for the future. There's never been anything like this before. They were basically saying this is in the top 3. There are certain things you just have to do, like if your employer offers a matching 401(k), you have to do it, because otherwise you're losing out on free money.

If you don't do a health savings account or you don't max out your Roth IRA, there are certain things you have to do because they're so tax-advantaged or you're getting free money. In this case, you're getting both. There is the opportunity for, frankly, free money for your kids, but also the tax advantage is huge.

Let's just go through this, and Brad, correct me if I get any of this wrong. You can donate up to $5,000 a year to your kid as long as they're under 18. It's not just you; any friends and family or others can contribute as well, which is new. Then they get tax-free compounding until they're 18.

Your employer can contribute up to $2,500 tax-free. At a minimum, you should go to your employer and say, “Sign up for this. If you have to take $2,500 out of my salary and make it a donation to my kid's Trump Account, do that, because that's a huge tax savings.” Neither side has to pay tax on it.

Like Brad said, this is basically like an IRA. You get tax-free compounding. When the kid turns 18, they can get access to it, and they can do a rollover into an IRA or into a Roth IRA, which is even better, because when the Roth IRA matures, you don't pay tax on the money distributed out of it.

With a traditional IRA, all the taxes get deferred until the end, whereas with an IRA-to-Roth IRA conversion, you're supposed to pay taxes at that point. I saw one really clever CPA say the best way to do this is to wait until your kid is no longer a dependent. Maybe they're in college or they just graduated from college and they're in the 0% tax bracket because they're not making any money. Then do the conversion.

You'll be able to convert the Trump Account very cheaply into a Roth IRA, and now they're potentially going to have $200,000 to $300,000 in that account that they can use for tax-free investing for the rest of their life, or they could potentially start a company with that.

Other things you can do with an IRA are use part of the money for a down payment on a first home purchase or, if you get into a health emergency, use the money for that. There are all these things. You're allowed to distribute money out of an IRA without incurring a penalty, but generally speaking, the point of an IRA is to save for retirement.

This is where I think it gets really amazing: If you start with $200,000 to $300,000 at age 18, you'll be at $10 million-plus by age 60 if you just let it compound. There are ranges.

Nepo babies we're creating here. We're going to have a lot of rich kids with trust funds.

Yeah. So this is all you have to do to make sure that your kid is protected for retirement: You, your family, your friends, and your employer can just contribute to their Trump Accounts.

That's way better than Social Security.

Brad, I have an idea. I have to go sell some enterprise software, so I have to leave. But I'm really proud of you. I think this is incredible. You should convince OpenAI and Anthropic to give the equity of those companies—if this is going to be as big as you say, $100 billion, $300 billion, zillion, trillion—and put it into the accounts of every kid.

Can you just explain how that works against the $5,000 limit?

I got to go. Love you guys.

See you later. All right. Good luck on the sales call.

Yeah. Always be closing. Chamath. Always be Chamath closing. ABC.

Chamath always be closing. Again, David and Jason, as we build out the platform at scale, imagine now that you have 50 million accounts opened. I've said on CNBC, and I've obviously talked with Dario, Sam, Elon, and others about making those donations.

I don't like the idea of shaking down our companies, taking their shares, and then putting them in some government slush fund that perhaps Bernie or AOC or somebody is going to control in the future. I've said it has to be voluntary, and, number 2, it should go into citizen accounts—privately held citizen accounts that compound for life.

You asked the question, David: How does it happen given the limits that you have—the $5,000 per child? That's why you have to have 50 million accounts opened, because then you can take dollars in at scale. We can also set up a pooled account, David, where it can be distributed over time. You distribute it to all the kids subject to the limits that you have today, and then any remainder you can distribute to the 3.5 million kids who are going to be born next year, or the 3.5 million kids born the year after that, or the 3.5 million kids born after that.

We are on a trajectory now where we're going to have over 100 million of these accounts set up over the next decade. We could have 70 million today, and then you're going to add 3.7 million a year. So you're going to be at 100 million private individual accounts that are compounding for people's lives, that anybody can donate money into, that people themselves own.

I think one of the things that gets lost is that moms and dads, or somebody working their summer job, can put in $10 or $100 into these accounts. They get to see it on their phone. When we started this, my sons and I designed this app, and it's basically what we ended up with.

Shout-out to Vlad at Robinhood, who helped you with it.

Vlad and Joe have implemented a more elegant version of this, but every kid owns a little bit of Nvidia, a little bit of Microsoft, and a little bit of Apple. Imagine opening up that account, David, in middle school or high school, going to the money page, and now you're getting excited because you're seeing, “Oh, man, I'm in the game. I have ownership.”

While you referenced what it could be for families who can contribute $5,000, obviously Michael Dell, Susan Dell, Gwynne, and everybody else are focused on the 50% of Americans who feel left out and left behind, who would otherwise have zero. If you do the math on this over the course of the next 15 years, you could have somewhere between $2 trillion and $4 trillion added to the accounts of families and kids who would otherwise have had zero.

We talk a lot. Go ahead.

We talk a lot.

David Friedberg

See, the philanthropy piece—does it basically work in that other philanthropists can contribute toward that $5,000 per kid? Is that right?

So, when Gwynne Shotwell contributes 2 million shares of SpaceX to 2 million kids, each kid is getting a share of stock—

David Friedberg

$150. So now—

That's counting against their $5,000 limit. That's what makes it compelling. Every family that can afford to do the $5,000 should, because it's so compelling from a tax and savings standpoint. But even for families who can't, there are going to be beneficiaries of philanthropists who just want to give this type of direct giving.

It seems to me this is so much more efficient and so much better than the whole NGO industrial complex, where they take a percentage for their offices and their salaries.

Absolutely.

Yeah, they're just drifting.

Yeah.

David Friedberg

One of the numbers I saw was kind of amazing. Again, it just goes back to the power of compounding. If a Trump account had been maxed out and you had the standard market rate of return that we've had for, say, the past 30 years, then by age 28, that kid would be a millionaire.

Incredible.

That's right. All the numbers you hear me quote—the $50,000 and the $200,000—don't assume maxing out. That just assumes people are adding $50 a month, because Michael and others and I have really been focused on the families who don't have the capacity to save today. We're getting all of them into the game.

The president directed us. He said, “Listen, we have 529 accounts that already help the top 10%.” That's not who we're focused on. This is about the Main Street agenda. This is about all the families he ran for who feel left out and left behind, and we're reconnecting them to the American dream through universal ownership.

They all have their own account. They all have a private account on their phone. It's a game changer for the country. It's the largest change to our social contract since 1935 and Social Security. Importantly, I think it couldn't come at a better time.

Every employer should be signed up as an employer that can contribute, because you could take that $2,500 and, hopefully, it's additive—not just a substitute. But even if it's just a substitute, JCal, your employer takes $2,500 out of your salary and puts it in your kids' Trump Account. You're reducing your taxable income, so it's a no-brainer.

If you're a profitable company, your employees are going to love you.

Yeah, but I think every employee is going to want it, and every employer should do it because the tax savings are for both, right?

The tax savings here are huge.

David Friedberg

Yeah. So, just off the mechanics of it, I want to level up here for a second.

Yeah. I have been often critical. I call balls and strikes, and I can tell you in detail the things that I have a problem with in this administration and the actions they've taken. I have done it here on the pod.

I want to address the people who are negging this, specifically because it has the name Trump Accounts on it—which, I told you at the poker game, call them Trump Accounts. I don't know if that was an obvious thing or if I was the person who told you to do it. I'm not taking any credit here, but I remember that conversation where I was like, “Just call them Trump Accounts.”

Whatever criticism you have of Trump, however valid you may feel it is, this has nothing to do with Donald Trump and how you feel about him. Put your TDS on the side. Put your valid criticisms on the side.

In this country, we have a K-shaped recovery going on. We have immense tension between the haves and the have-nots, to the point at which people actually believe that socialism and communism are better operating systems than the best operating system humanity has ever created, which is called democracy plus capitalism.

Kids love capitalism. They love building businesses. But we are in an existential moment right now. If these kids—young kids, and there are a couple of generations of them right now—do not believe in America anymore, while we're sitting here on the 250th anniversary of this amazing experiment known as America, this is the most American thing you can do.

Put aside your TDS, put aside your valid criticisms, and embrace this. Give the flowers to Brad and to the people donating, like Michael and Susan Dell and Gwynne. This is beautiful. This is the most beautiful gift I've ever seen to a country, and this could be a unifying principle that brings us back together as a country—that everybody gets to participate in capitalism.

This is the number 1 way to do it: Let kids on their smartphones, instead of saying, “You know what? Mamdani is right. I should get a free bus ride. I should get free pizza. We should take Ken Griffin's wealth and seize his penthouse and pay out of there.” All that, you know, while we have CEOs getting shot and their homes firebombed.

Well, you know what? If you're one of those CEOs, you've done incredibly well. There was something called the Giving Pledge, where they pushed affluent people at the TED conference for decades—Bill Gates and everybody, Warren Buffett, everybody was pushing for this.

This is the perfect version of the Giving Pledge, because you're not just saying, “I'm giving away my wealth by the time I die.” You're very strategically saying, “Every single person in America gets to be part of the best part of America, which is entrepreneurship.” Everybody will be part of the equity nation.

My final point is that one of the happiest countries in the world is Australia. If you've ever gone to Australia, everybody feels safe. We have a large number of people in this country who do not feel safe, and the reason they don't feel safe is because they don't think their kids are safe—to the point at which people do not want to have kids in this country because they feel the system is just too hard.

This could change that. If people feel, “Hey, kids have a shot, and I don't have to worry about my kids,” that's meaningful. I worry about my kids, and I'm affluent. I can't imagine being a single parent and what anxiety you must have as a single mother or father when you're making minimum wage and you're behind the 8-ball for your entire life—and now your kids are set.

That's all people want. The only thing a parent wants is to make sure their kids have a better future. That was the promise of this country, and somehow it went off the rails for the last 2 generations. This puts it back on the rails. This is superannuation funds in Australia.

In Australia, people are extremely happy. The reason they're happy is they're forced to put $14,000 a year, or whatever it is—12% or 14%, I think—of their income into essentially a 401(k) that they get to direct to a certain extent. It's forced savings. This does the same thing at a very basic level. This could replace Social Security. This replaces the Giving Pledge.

And I just want to say, Brad, a lot of my friends got involved in politics—some of them on this very program, a lot of people. It's very divisive. You threaded the needle here. It was a master class in balancing these 2 crazy parties and the divisiveness in this country. I just want to give you, as your friend, your flowers. This is absolutely outstanding what you did, and you have been incredibly humble in your approach to this. This would not have happened without you, Brad.

This is your legacy of everything you've done in your life. Lots of success, and I've seen it up close and personal. This is a million times X everything you've done in your whole life. You'll be remembered for this. This is architecture.

It should be as big as Social Security. I mean, it's a new platform.

It's not just philanthropy; it's also retirement savings. I mean, right? And everything in between.

And this is not static. This is dynamic.

We can add to this. There could be other features. I'm seeing a lot of people in the comments say, "Why stop at age 18? Why can't you—I mean, you have the Trump Account roll over into an IRA or Roth IRA after age 18, but why can't you keep it going and then people can keep that $5,000 contribution going? And, you know, it doesn't mean we take away—

Exactly.

—retirement benefits that are owed to current Social Security recipients, but sunset it. At a certain point, you could just say, "Hey, the next generation is going to be on this platform rather than the old one," and it would be a lot better. A lot more efficient than the government running it, right, Sacks?

Yeah.

We don't want the government running this. Let me ask you a question about this, Brad, because I do see one thing that people say, which is, what if your kid turns 18 and then they just want to blow the money? How do you trust that they're going to put it to good use, as opposed to—I don't know—YOLO and whatever?

As you know, these things are always political trade-offs and balances. I wanted them to have to compound until they were 30, right? Because I figured by 30, you were a little bit more mentally developed on financial issues. But the argument ultimately came, "You're old enough to vote, you're old enough to fight a war. If, by 18, we don't allow you to have control of your own money..." So that's where a political consensus was built, David.

But the other thing, remember, is they can only take up to 25% out to buy a home, start a business, or go to college. The rest rolls into an IRA, and there are built-in penalties for early withdrawal in an IRA, so there are disincentives for people to pull out. But let's be clear: we have to do a much better job in our education, our public education system, leveraging this as the platform.

If a kid doesn't have any money, it's hard to get excited about learning about money. But if a kid's in the game and has $12,000 in the 7th grade, now you've got my attention. I own a little bit of Nike. I own a little bit of Apple. I own a little bit of Nvidia. Let's talk about that. How did it get there? How did it compound? If I added $50 a month, what does it turn into? All of these things will now be present on every child's phone in America.

37 states require financial literacy. Every state should build this into the curriculum. We're working with a lot of states on that. We haven't talked about that. We have about 25 states that are going to add money into the accounts of the kids in their states. So, you've got—

States are going to do it.

States. State action. Oklahoma, West Virginia, Indiana, et cetera. So that's also sweeping the country, where the states are looking at programs they already have, where they're spending money for kids that are ineffective, and they're saying, "Instead of continuing to spend money on these things that aren't working, why not just block-grant the money directly to the kids?" Because if we give the kids the money, they're more likely to graduate, more likely to buy a home, start a business, et cetera.

I think that we're, as I said in the Oval, this is day 1, and I am fully committed to the next decade, as is my partner in crime on this, Michael and Susan Dell. I appreciate your guys' comments on the legacy of this. It has been the most profound work and kind of honor of my life.

Standing in the Oval Office with my 2 sons, who are really my 2 co-founders on this—we made the sketch of this at our kitchen table in the fall of 2020. That's where the conversation started. Lincoln's been with me in every single meeting with every congressman, senator, president, former president, et cetera. The president shouted him out again when we were there. That journey as a father with my kids—the payoff has been really extraordinary.

The other thing that's notable here—and again, you've got to call balls and strikes and give credit where credit is due—is that Joe Gebbia joined this administration. A lot of people in the tech industry are like, "Oh, you joined the Trump administration," whatever. Okay, there's some criticism. He's an incredible world-class designer.

I was talking to producer Nick, our producer here, who also happens to share the same last name as me. He signed up for this, right? He's doing well, but his wife signed up for it. The software is fantastic.

Let's pause for a second. The American government has made exceptional software, and this all got done in Trump's first 18 months. Immense credit for this. Of all the challenges this presidency has had—the Iran war and other issues—we made great software. The American government, because of Joe Gebbia, makes kickass software. Also, a major shout-out to him. He could be doing whatever he wants. He's done incredibly well as the co-founder of Airbnb, and he's doing this. That's a real patriotic thing to do. I'm just over the moon with this. I think it's fantastic.

The dream team we had: Michael Dell, myself, Vlad Tenev, Joe Gebbia, the Treasury Secretary, Luke Pettit at the Treasury, talking basically every day for the last year. Our objective was not, "We want to build the best thing that the government's ever launched." We wanted to build one of the best consumer products, period, that's ever been launched. Mission accomplished.

Any Silicon Valley consumer company would be thrilled with the numbers that we're seeing, the ratings that we're seeing, the engagement that we're seeing. It's been fun doing it with that incredible team. Everybody's in this for the right mission as well. And so, I agree with you, Jason. It's rare that government recruits or embraces that mission. I think the tent's getting a heck of a lot bigger. It is bipartisan. It's bipartisan in support.

You know, this is important: what Governor Moore from Maryland said yesterday. He said, "Republicans and Democrats have been trying to do something that looks like this, feels like this, for 40 years. This guy got it done. Give him credit. This is great for America, great for our kids, and especially on the 250th anniversary."

You're standing in the Oval Office, looking at the original Declaration of Independence, and you realize what people laid down for this experiment. And then I read this chatter in my feeds about Mamdani and others literally wanting to set this great experiment on fire.

Yeah.

Right. Wanting to burn the place down because they think they have a better formula. No, the answer is evolving and doubling down on the formula that has worked for 250 years. And rather than making all these kids socialists, we get them all into the game of capitalism. They become owners—owners and shareholders in America. So that's what we got accomplished.

I appreciate the chance to talk about it.

Listen, incredible, Brad, and just amazing to watch you do it. And if you really think about it, the tech industry needs to win—and capitalists and creators and the makers, the people who build stuff. This is our chance to say, "Here's an example of something that helps the people at the bottom," right?

And I talked a bunch about the minimum wage here. We've got a $7 minimum wage. We need to address that as well. These are the things that, if we address what people are scared about—what people who don't have what we all have here—if we have empathy for those people and we actually care about them and we give them a path to believe in the American dream, they will take that path—

We have to give them a better path than these socialist lunatics. And this is so much of a better path.

Let's do it again. Let's do another one of these things. Let's keep growing this spirit of getting everybody in the country to have equity in these great companies. That's why the entire world is trying to replicate what we do here. Every single country I go to wants to recreate Silicon Valley.

We're not stopping here. And, David, to your point, whether it's the AI companies—or frankly whether it's the Intel shares, or whether it's the TikTok fee—I now have a place where all of those wins achieved by this administration can go. They ought to go directly to all the citizens, you know, the country, into their accounts, and compound for a lifetime.

As far as people over the age of 18, there's certainly a lot of talk about that as well. Again, Social Security is a sacred promise by both parties. Nobody's going to change that, but there's a huge opportunity to have a supplement here.

Why shouldn't people between 20 and 30, or 20 and 40, also have a Trump account that they can begin, on a supplemental basis, adding dollars that they own and control? Remember the big difference between this and Social Security: Social Security takes 12.4% of my W-2 income and puts it into something akin to the black hole of government. I don't own it. I don't control it. If I ask a room of 3,000 people how much they've contributed, they have no idea. If I die, I don't have title; it doesn't pass to my heirs, et cetera. So it's really not mine.

If we created a supplemental IRA—it doesn't even require new legislation, I don't think—it's just expanding the age. These are IRAs after the age of 18, right? Then people could start building supplemental wealth and participate in these gains. There's a lot of conversation going on about that as well. And so we're not done, but it was a hell of a milestone on the 250th birthday of America to ring the bell in the Oval Office and to watch all these kids get their accounts lit up. It was pretty special.

What I think is really cool about the Trump accounts—I think it's an amazing philanthropic platform, but in addition to that, it's an amazing platform for middle-class family planning. That's the point I'm trying to make. All the CPAs that I'm seeing talking about this are saying this is like one of the greatest things ever. If I could only tell my clients to do one thing, this would be the one thing.

Finally, something for the middle class, right? This is what people have been asking for: Hey, let's get something for the middle class.

Yeah. And I think the market gap, bro—correct me if I'm wrong—but basically, the market gap that was created here is that you can't get an IRA, which is basically a tax-advantaged savings account, until you have your first job and get earnings, right?

Which would be for most people at age 22-plus. So for that first 22 years—

Your kids can't have an IRA, right?

Correct. And you're effectively giving every child at birth a Roth IRA. You know, a Trump account has certain rules. It's actually better than an IRA.

Better than an IRA.

It's better than an IRA because with an IRA, your employer can't contribute $2,500 tax-free, can they? I mean—

I don't think so. Philanthropists can't contribute to it, and moms and dads can't. But the most important thing—you know, Buffett's like, the secret is to find a really small snowball on a really long hill. But the size of the hill in this country, we've cut off the first third of the hill forever. Nobody saves anything until they're 25.

Okay, the easiest compounding in the world—the easiest compounding in the world—is between 0 and 5. All right, listen, this has been amazing.

Yeah. Because they're dependent, frankly. They're still on Mom and Dad.

Exactly. So you pick up the first third of life.

Incredible, right? In compounding. And so, it's—

It's really remarkable. Let's go. I mean, amazing job. All right. We'll see you all next time. Bye-bye.