Epstein文件余波、Nvidia风险、Burry的糟糕赌注、Google的突破、Tether的繁荣

All right, everybody. Welcome back to the number one podcast in the world. We are together in person. Yes, the besties are together in Vegas. It's going to be a great time. We're here for F1. Our friends at the Venetian have been amazing, gracious hosts. They gave us their beautiful studio here. We're going to play some cards. We're going to have Phil Hellmuth, Jason Koon, and all of our besties are coming. I've never stayed at the Venetian before. It's amazing. They gave us wonderful suites. It's beautiful. They VIP'd us out, and this is the place you want to play cards. They've got a beautiful, brand-new poker room. They've got a high-stakes room. We'll be playing here later. We'll be playing the secret game. And yes, trackside for Formula 1. We're here for F1. You brought a car dealer with you? You know Matthew? Your car dealer from your home game is here. Yeah, Matthew, car dealer. Is he dead money? What's the story here? Sadly? No? Okay. I haven't been in the home game for a little bit, and it looks like people got out of line. But anyway, thank you so much to our friends at the Venetian. They're doing a ton of poker content here, so you can look at that on their YouTube. All right, everybody. You've wanted us to talk about the Epstein files, and we're going to talk about it today. In a stunning turn of events, the House and Senate voted nearly unanimously to release the Epstein files. The vote was 427 to 1. Chamath, who abstained from the act? No, the person who abstained—well played—was Republican Clay Higgins from Louisiana.

He said it reveals and injures thousands of innocent people: witnesses, people who provided alibis, and family members. He makes a great point, but Attorney General Pam Bondi has addressed that already. They're not going to release any open investigations, and they're going to remove names if releasing them would harm anybody.

The Senate passed it by unanimous consent, which requires a sign-off from every senator, and Trump, in a reversal, signed the bill last night saying, "Give them everything." We did see some emails come out from the Epstein files last week. Friend of the pod Larry Summers was a main character in them, and he was communicating with Epstein up until 2019, asking him for advice on dating.

He's since stepped down from OpenAI and several other public-facing roles, and was just, I think, put on leave from Harvard. What do you guys think is going to be the fallout from this? I guess the question is, what's going to be the fallout from the release of the files?

I mean, I guess we'll put on our tin foil hats.

Let's break this down. I think the first question is, what is the relation between the Epstein files and Donald Trump? I think the answer is that it's flimsy. The reason is because this is the most investigated, most litigated human being on Earth.

If you had something that was incredibly salacious and accusatory of Trump, it would have been released during the Biden administration because it would have made a lot of sense politically to try to damage his candidacy. So the fact that we haven't seen much of anything other than some photos means that there's nothing there related to Trump.

So then, why didn't they release more of the files when they had them for 4 years? It's probably because there are a nontrivial number of Democratic operatives who are touched by these things.

Well, also, I think the reason they didn't release them was because there was an open Ghislaine Maxwell case, and she was appealing it, so they couldn't release them.

There were probably 2 reasons, but you know how this works. There are innumerable ways to leak stuff.

My point is, now what you're starting to see in these documents is that it seems to be tainting the Democratic establishment elite more than the Republicans. It explains why there were so few leaks in the last 4 years.

Jeffrey Epstein was a total creep. That island should be covered in cement and drowned. The house should be burned to the ground and replaced with something nice.

You're the one who said this—it's 1,000 women? That's what I saw in a report.

Saying 1,000 women.

Okay, someone said 1,000 women. Apparently, in there, there are claims that it's—

It was one of the victims who said, "There's 1,000 of us."

You have to be incredibly careful and thoughtful to protect their rights and respect what they've gone through. But I think now we need to release these files in an orderly manner, put this episode behind us, learn what we need to learn from it, get better, be better, treat these people with respect, and move on.

Do you think the release of the files is meant to help the victims, or do you think it's meant to identify fodder to go after political enemies?

Neither of those two things. The releasing of the files at this point is one of these things that's about a compact between those who have power and those who ask for something.

This is an issue that has animated millions of Americans. When they constantly keep asking for these things to be put out there, I think it's a good signal for the government to listen to folks and release them, again, in a respectful way.

Similarly, there are other things that I think fall into this. We've heard about the JFK files, right? The killing of Martin Luther King, the Amelia Earhart files, and all the UFO files.

I think what it does is show a pattern of being responsive to the voting public, and I think that's a good thing.

So Higgins, who, by the way, was a sheriff, was an Army staff sergeant, and has been in Congress for, I think, 9 years, said, in addition to talking about the victims, "This abandons 250 years of criminal justice precedent and procedure in America. A broad reveal of criminal investigative files released to a rabid media will absolutely result in innocent people being hurt."

Do you think this is a singular situation with Epstein because it's so extraordinary and there are so many people tied up in it? Or does this set a precedent where, anytime people and the media start to say, "Hey, we want to know what's going on in the middle of an active investigation or former investigation," these files get released?

Does this become a new standard where we're just going to start opening up investigative files like this? Do you think it's a singular thing? Because that's his whole point.

A lot of these issues have to bake for a decade or 2 before people want them to come out. This has been going on for now—how many decades?

Well, it's like a 20-year story. I feel like the investigative piece that's missing is how he got away scot-free in Florida.

When he was criminally charged, he was convicted and pleaded out, and he was basically let go. I can actually say I know a little bit about it because I met Epstein a half-dozen times at the TED conference.

Oh, you did?

Yes. I've talked about it on my Twitter incessantly because I'm in—

You are in his book.

I am among the thousands of people in his black book. I met him at the TED conference.

You know who else? Jeffrey Epstein, David, and I.

Okay, congratulations. You guys weren't in New York.

I went to the TED conference. I just avoided that room.

There was a billionaire's dinner at the TED conference. I didn't actually go to the TED conference. My book agent would host the billionaire's dinner. If you type in "billionaire's—"

You used to go to the TED conference. Of course.

You used to go to the TED conference. It was like I was never invited to the billionaire's dinner.

This is in the 1990s. That's how old this is. He was there giving donations to scientists—Marvin Minsky, MIT, all that stuff.

When he went away and got busted in Miami, the way it was framed in the TED community was that he had been set up, that this was just an underage girl whose ID had been checked, and that it was some sort of setup. They said he had been given a work-from-home sentence: he could go to work every day and then report to jail, and that it was all just a misunderstanding.

Looking back on it, I think he's a spy. I am not the conspiracy theorist of this podcast, but—

Shared intel with?

It could be anyone in that spectrum. We all know people in the intelligence community, and there are easy ways to pass intel without it being obvious, like working for an agency.

Then I think the question becomes, was he an asset? Was he sharing information? To what extent? The reason I think this is because why would he have an interest in the top scientists at the top universities, getting close to them, and then deciding who he wanted to pass that information on to?

Who would want top intelligence from scientists? Russia, Israel, the CIA. The compromat thing also seems likely because he had cameras everywhere, and they've talked about this.

So he had cameras, and you're saying he recorded famous people and then used that to get things?

I think there is a nonzero chance that part of the story is true. But I don't know that it's a 90% chance that happened. I think when we go back and look at this, there's a bunch of people who had embarrassing interactions with this person, who spread money everywhere.

Everybody wanted his money. That's why they were lining up. All of the scientists—Joi Ito, Reid Hoffman, all of these folks—were trying to get his money, which is peculiar.

He was also giving tax advice to the Microsoft people, to Peter Thiel. He was always trying to integrate himself into powerful people with money and scientists. Why? Was it to make money, or was it for some other purpose? I think it was not to make money. I think there was some other purpose here.

Now, I am not a conspiracy theorist, but because this thing has gone on so long and it has not been released, I think there are mechanisms that are keeping it at bay.

That's to me like the Occam's razor version of this. I think there are going to be very embarrassing or compromising things for intelligence agencies, which I think is the same thing behind—if I could have one question about JFK's assassination answered for me.

If I could have 1 question answered for me out of the Epstein files, if I could just have the question answered, I'd want to know: Where did all his money come from?

Because it is not very clear how a guy who was managing money for a billionaire, Les Wexner—we all know, if you're a money manager, maybe you're making 0.5% a year. No, that was documented. How did he get all the money?

Leon Black, who's the founder of Apollo, in 1 year paid Epstein $168 million for tax advice. That came out in the lawsuit that ultimately led to Leon Black resigning from Apollo. Right. But what did the tax advice look like? That's what he said it was. Now, I'm not trying to high-roll anybody, but I've had all the tax advisers come and give me their advice.

It cost you $168 million? It costs you $1,200 an hour. No, no, no, it's cost me millions. But I'm hard-pressed to understand what advice could have been given to me where I would have paid $168 million.

Here's the news: Your accountant just sent the bill. It's $172 million. No, no, no, it's crazy. Right, you're right. If you go to the best estate lawyers in the United States, it will cost you $5 to $10 million.

What do you think he was getting paid for? Was he getting a portion of the savings for using tax loopholes? Maybe he was charging on some sort of commission, but he was a money manager for many of these Microsoft executives, et cetera.

And when Peter Thiel said, “Why did you meet with him?” he said, “Tax advice,” which makes total sense. That's a total Peter Thiel—legitimate. That totally tracks. Peter Thiel was known for his Roth, right? He's known for studying these kinds of things. It makes total sense to me.

I think we're going to have a bunch of Larry Summers-like embarrassing things. There could be embarrassing things there for Democrats, Republicans, everybody in between—the scientists, obviously, who went to the island and all that stuff. It's all going to be embarrassing.

And I think at the end of the day, we're going to find out that some intelligence agency was somehow involved in this, and that's why it's being covered up and that's why it's so toxic.

That's your prediction. That's my prediction. Yes. Nostra-Chamath has spoken. Which intelligence agency will you pick? Oh, good sir. Look into your crystal ball.

I mean, CIA. It would be one of the big 3. The CIA, which he was talking to—I think in this latest volume of emails, he was talking to people from Israel, from the CIA, and he was talking to Russians. He was talking to all 3 in the emails that have been leaked.

Are you talking to Russian intelligence? He was talking to Russians. Oh my God. Yes, Russians, I think, generally. He was very involved with Russians.

Could you imagine how much anxiety we'd have if that was our job? I can't even imagine. I just want to go to my office, build some stuff, make a few investments, and play with my kids.

People, you know, mess around. It just seems like—what? Oh my God.

When they make the movie, are you going to play yourself? I mean, if you look online, if you go to the Edge.org site and look at those billionaire dinners, you'll see me in a couple of pictures with Larry, Sergey, Zuck, and Ev Williams, when we were all 29 or 30 years old.

Jesse Eisenberg to play you? I know his hair is curly and yours is straight.

More like Leo, probably. People might say Ethan Hawke. I get a lot of those, but I digress.

Who plays Ghislaine? Quick—I mean, who plays young Ghislaine?

Reid Hoffman? Reid Hoffman—who would play Reid Hoffman, huh? I just don't see a world in which Reid was involved in shenanigans, I'll be totally honest. I think he was just trying to raise money.

I think it's unfair that everybody who met him is being dragged into, “Oh, they were somehow a pedophile.” That's just crazy.

The guy was a consummate networker, obviously doing a bunch of stuff. He funded all of those dinners. He was funding all of these dinners. He was hosting dinners in New York. In New York, he was known for having these dinner parties with all kinds of famous people. You can look online.

His business was to meet with people and throw dinners. To throw dinners, Chamath?

That seems calorically taxing. You know what I mean? Dinner parties—you overeat at a dinner party. Imagine having 3 of those a week. I don't know.

Jason, do you know anybody who's having 3 dinner parties a week? It's a lot.

Jason and I went to Carbone last night. We'll talk about it when Keating gets here. That got a little heated. Speaking of—I’ll talk about it when he gets here. Somebody got out of line.

I had dinner last night with Paolo Ardoino, CEO and founder of Tether. Can't wait to meet him. Amazing, amazing guy. That is an incredible business. But don't they have $150 billion in treasuries now? It is incredible. Here's why that business is incredible. This is what I learned.

About Tether, the stablecoin.

Tether, the stablecoin. Yes. There are millions of people using U.S.-dollar-backed stablecoins from Tether all around the world—all over Africa, all over Central America, all over Asia.

Number 1, his user base is growing by 30 million users a quarter. The financial inclusion that then ties back to U.S. dollar hegemony is unbelievable.

Explain it to me like I'm an idiot who's never bought a stablecoin.

Yeah. So it's like—here's USDC, which is the one I have. Here's USDT. Let's look at these businesses as roughly the same. There's Circle, there's Tether, there's World Liberty Financial. They all have a stablecoin. What is it?

Okay, let's say that Jason is a cash worker in India. Let's use that as an example.

Sure. He gets paid 100 rupees, and he's like, “The rupee is constantly getting devalued. I'm constantly losing purchasing power. I want to swap that into a U.S. dollar.”

So he would create a crypto wallet. What Tether will say is, “Great. Give me your INR, your 100 rupees.” They immediately swap it to a U.S. dollar. So now there's a U.S. dollar, and there's a token for that dollar.

Right. I give Jason the token for that dollar. Now I have this dollar. What do I do with it? When I accumulate enough of these dollars—$50 billion, $100 billion—I can take that and invest it in treasuries so that it's completely safe.

U.S. treasuries.

U.S. treasuries. Now, if Jason decides to send it to you and then you redeem it, I can sell $1 of those treasuries that I own and undo the chain.

And does Tether earn all the interest on the treasuries? So I'm getting to this. Now Tether, Circle, and World Liberty earn interest on that. When the number gets big enough—when this number gets ginormous—what they do, in Tether's case, is reinvest this capital into all kinds of diversified assets: Bitcoin, gold, real estate.

But what they also do is invest in things like financial inclusion in Africa. He walked me through a bunch of things that he's doing yesterday. It is an incredible business.

And so, as a holder of the stablecoin in my wallet, I'm not earning any of that treasury yield. I just have a flat, dollar-denominated, or dollar-protected stablecoin.

A dollar-pegged stablecoin. You have a dollar-pegged stablecoin. Right. And that is sufficient risk management and risk mitigation for half a billion people.

Right. They're not trying to get a 4% or 3% interest yield.

No. In fact, you're bringing up the big point, which is—in the United States, what is the big fight now?

The big fight in the United States, in this thing called the CLARITY Act, which is meandering through the House and the Senate, is what should happen in the market structure. Meaning, if you, David Friedberg, are the one that gave me the dollar and I am, let's say, for example, Coinbase, and I issue you a stablecoin, do I share that revenue with you? Do you earn all of it?

Obviously, the banks, like the J.P. Morgans of the world, the Citibanks, they don't want that, right? Because that's their net interest margin. That's what happens today. You deposit money in the bank.

I said this on the program. The bank goes off to invest it. In the stablecoin legislation, they weren't able to give the stablecoin providers the ability to pay interest to consumers. They made that concession, but that will change over time.

But the banks were able to fight for that concession.

They were able to fight for it. But then you have the emergent crypto companies who say, “Hey, let’s find a way where we can do a sharing mechanism.”

How they hack around it is they do kind of a sharing mechanism, but via this kludgy way called rewards. So you earn rewards and reward points, but it’s not really what it should be. It should be that if you earn that interest margin, you should be able to share that.

And, by the way, you should be able to have different rules in different markets because, again, if you’re in Kenya, the last thing you’re probably thinking is, “Do I get the 4%?” What you’re more worried about is that the Kenyan currency, whatever it’s called, is about to depreciate another 60% this year.

Right, right. Let me just hedge that. That’s more than enough value. Anyway, I thought Paolo was incredibly impressive.

Well, I will say this: Tether’s business is really impressive. I have been super critical of Tether publicly, and they had a lot of challenges as a business. They were banned in many markets. They didn’t do any audits, and people didn’t know what was in there.

They’ve done an incredible job cleaning all that up. Now they’re starting to go from attestations to audits, and they desperately want to be legal in America. Under that legislation, they have 3 years to do it, and they then have to unwind being banned in New York, banned in Canada, and all these places where they got banned.

I want to give credit to David Sacks. What we saw under Biden, and what we saw with the anti-crypto approach that they took—and Trump in his first presidency was anti-crypto as well—that decade of anti-crypto led to a lot of people doing offshore stuff like Tether and a lot of shenanigans.

Actually, Sacks, who can’t make it this weekend, has now created a framework that is helping people do it the right way and taking out all of these questions.

Yeah, and Tether is example one. Tethers have been the default for all kinds of—

Do you know that for sure? This has been in our congressional hearings. They have documented it very clearly. Let’s not make the accusation if we don’t know.

What I saw yesterday was a very, very, very credible and thoughtful entrepreneur and a great business. The other thing I’ll say is that I would like to invite you to come with me to the Tether conference at the end of January.

Okay, we are going to go. He’s never turned down an invitation.

Here’s what we’re going to do. We’re going to fly together to El Salvador.

We’re going to do a tour of the prisons?

No, no, we’re going to do an interview with Bukele. And then we’re going to do an interview with Paolo, and then we’ll fly home. Will you come with me?

If I can ask him any question I want.

You have to go check out the prison.

No, I’ve been told you cannot go to CECOT on the first trip. I don’t want to go anywhere near that prison, but if I can ask him any question and he’ll be fine with it—

He’s great, dude.

Yeah, I’m happy to go. Yeah, of course.

In any other world, he would have been in Silicon Valley doing the same thing, building a trillion-dollar company.

The other challenge they’re going to have is when interest rates go down, these businesses are going to have to figure that out as well. But $183 billion in circulating USDT—that’s the ticker symbol right now—$135 billion of that is in Treasuries, and then another roughly $10 billion is in Bitcoin and gold.

And land.

That means they’re throwing off roughly 5%. They were making $7–8 billion a year just on the holdings. I can’t tell you all the details, but I’ve never seen a business—

No, they said it’s a $500 billion market cap. The word on the street is a $500 billion market cap, which would be roughly a 50-times price-to-sales ratio. They’re making $10 billion.

What do you think their profit margins are? Forget the growth quantum.

You only need 100 people to run the business.

Yeah, if the interest is the revenue, it’s probably a 60–70% margin business.

More than 95%. It makes total sense because, I mean, how many people do you need?

I thought last night at dinner, here we are grinding to get to 30%, 40%, 50% margins in so many of our businesses. And he’s like, “Yeah, yeah.” It’s incredible.

Good for him. Congratulations.

Congratulations. The good thing about that is there’s a financial theory, though, that high-margin businesses like that invite more competition.

Well, that’s literally where I was going. This is where competition gets ground down.

Stripe bought a stablecoin provider. It’s pretty well known. Stripe, Visa—everybody’s going to have their own stablecoin. So Tether will not have the market all to themselves, and obviously Jeremy Allaire at Circle is a very viable competitor.

Unfortunately, margins like that only have one direction to go.

Correct. And if interest rates go down, which we’ll talk about as well, that’s going to be a headwind for that whole space.

All right, let’s talk about Nvidia.

We’ve been talking a bit about Brad Gerstner personally deciding to blow up the AI bubble and then destroying the stock market. I’m joking. Shout-out to Brad Gerstner. My gosh, the short Bitcoin thing has been a bonanza. Crazy.

And then, is it below $90,000? It is, right? It’s like $87,000. I know it hit $88,000 or something, but watch out below. Let’s see the price. Watch out below.

Okay, let’s talk about Nvidia. Nvidia had a blowout quarter: revenue up 62% year over year, 22% quarter over quarter, and net income of $31.9 billion, up 65% year over year. They expect $65 billion this quarter.

Jensen, friend of the pod, has said that they can’t keep their product on the shelves. It’s sold out everywhere. At the same time, Michael Burry, who has the short on it, has been mixing it up. He is posting in response, I think, to you, Friedberg, who last week made a defense of the reasonable life of an H100, of these new chipsets that Nvidia sells.

Is it 4 years, 5 years, 6 years, 7 years? When do they get replaced? When do they have a useful life under GAAP accounting?

He believes, just to make it easy for the audience to understand, that major tech companies—big tech—are cooking the books in order to spike their earnings; that this is a house of cards; and that he’s going to short Palantir because it’s a 100-to-1 price-to-sales ratio. He’s also going to short Nvidia because of the depreciation.

What are your thoughts? I know you’ve seen his comments, Friedberg. I downloaded the GAAP depreciation rules. I was going to play the Accounting Corner jingle, which a fan sent me over the week.

Oh, great. We’ll put it in post. I want to hear it then.

Very nicely done, by the way. Thank you. Thank you for joining me here at Accounting Corner, and thank you to Roxana Martinez for that incredible jingle. I think we should adopt it.

Love it. Love it.

Send in your jingles, folks: Jason@allin.com.

Here we go. Accounting standards under ASC 360: “Depreciation must reflect the asset’s estimated useful life, not market innovation.”

Can you call us at 11:30 tonight and put us to bed? Okay, this is even—you found a corner even more boring than Science Corner.

People love Accounting Corner.

No, no, explain it, because—

Okay, yes, it is actually important. Under the GAAP standards—the generally accepted accounting principles—you set a useful life, and you reset that useful life as you reassess when you’re actually using that asset, not necessarily when there’s a better asset that creates more value.

Let’s explain this again. You make a big investment in property, plant, and equipment—PP&E. You write down that investment over a period of time that you, as an accountant, estimate to be the useful life of that asset.

If you’re going to use a building for 20 years, every year you write down the cost of that building by 1/20. You don’t get to write it all down in the first year. In fact, what Burry is arguing is that if you wrote it all down in the first year, your profit would go down and your business would look worse.

When you make an investment that you can use over a period of time, unlike salary, the accounting is different. When you pay someone a salary, you’re paying them for the hours they’re working that quarter or that year, and so that money is an expense. It gets recognized as paid out in that period.

But when you make an investment in a building or a piece of equipment that you’re going to use over time, you depreciate it. Just to go through that principle again, there are standards in GAAP for how you recognize the depreciation schedule. What’s the useful life? The useful life is based on when you’re actually realizing return value from that asset.

Burry’s point is incorrect. On Twitter, he said, “The idea of a useful life for depreciation being longer because chips from more than 3 to 4 years ago are fully depreciated confuses physical utilization with value creation.”

That is incorrect. There is value creation because they are generating revenue from those chips this year, 6 years later. So there is, in fact, a useful life for that chip that has extended into year 6.

And so it doesn’t matter—and this is part of the GAAP point that I wanted to bring up. What he’s arguing is that you should depreciate it over, say, 3 years, which means you’re doubling the cost every year, and then it’s all written off in 3 years.

But if you did that, to give you a point of example, in Google’s case, their total net profit would come down by roughly 10% to 12%.

So, it's not like they're cooking the books and recognizing some massive delta in their profit by doing this. The difference between 3 and 6 years is roughly 12% of their profit. And they're still using these chips.

What GAAP says is that only if the new asset—meaning the new chips—replaces the old one does the old asset's remaining useful life have to be marked down, with accelerated depreciation taken that year. Or if the maintenance costs spike, which means you have to spend money to fix the asset, which is not the case with chips.

The third is if the throughput requirements exceed the old equipment's capabilities, forcing early retirement. They're not retiring them. They're still making revenue off the old chips. Or if technological obsolescence means that you're putting the asset up for sale, then you stop using it after a period of time.

Yeah, and if you put it up for sale, you would actually know its market value. You could take that from the depreciation.

That's textbook GAAP: if you're still using the asset after 6 years, you can depreciate it over 6 years or whatever. This conversation lacks technical literacy.

So, let's assume you're Google, and let's say that the equivalent of an output token is the equivalent of a link. The first thing you would tell me is, “Not all links are made equal.” For example, if you generate a link for a pharmaceutical drug, Google charges a price per click that's way different from the link they generate that goes to Amazon to buy toothpaste.

For Google, it actually costs the same amount of money to generate that link. I think the thing that he needs to understand is that he's equating this to energy, but the reality is that in AI models, the thing we care about is: What is that output token? What is the revenue that's being generated?

That's right. What is the revenue that's being generated by the output token? Ultimately, what he doesn't appreciate is that Google, Facebook, Microsoft, OpenAI, and X are obviously not going to be in the business of generating negative-revenue output tokens just for the sake of it.

How do you know that? My wife got to the end of the internet this week. She launched X, put it on voice mode, and was stuck in traffic going from our house all the way to San Mateo and back. That's 25 minutes up, 25 minutes back. She said, “Hey, you know what? I ran out of tokens.” It said, “You can't use it anymore.”

You mean on Grok? Yeah, yeah. Why do they do that? It's because they're very conscious that there's a certain energy output and a certain revenue potential, and beyond that, they start to gate it.

Yeah, yeah. You do it on OpenAI; you do it on all these things. So, they're already keenly aware of the value of these output tokens. They know the revenue it's generating. Sorry, just one thing.

Yeah. And then the second thing is, in the bowels of these organizations, everybody has completely rebuilt all of the decoder infrastructure.

Before something gets to you, the user, there are all kinds of different manipulations that people are doing in the models, after the models, and before the models. All of that stuff has been rebuilt.

So, I think what he needs to understand—and, in fairness to him, what I would say Burry needs to appreciate—is that when those laws were written, they were written for a factory. They're for a turbine. It's a static thing. It probably doesn't do a very good job of understanding the world of chips.

But could he take a little bit of effort to call somebody and actually learn how this works? Yes. Is he doing it? No.

So, we don't care about it. Those machines are going to have 90% of their utilization in the first 3 years. For the next 7 years, they'll have 10% of their value. They'll be doing some small jobs in the background that won't be as important. Accounting isn't built to do this kind of refined depreciation schedule.

What do you mean by 90% of their utilization? Because if they're still making revenue—

Value to the consumer. So, let's theoretically think about the value of that H100. What value did the users get out of it?

The value today is that I'm making goofy Sora videos that generate no revenue. It's all money-losing. But down the road, that might actually be advertising, and it might create some number of clicks or some number of subscriptions, so we'll actually be able to attribute revenue to it.

There's no way to look at these devices right now and know how many of them are actually generating revenue in the first 2 years versus the next 2 years.

I think we know much more now than we did even 6 months ago about how to value an output token. What are the instructions we give to the accounting community on how to deal with that?

Dude, this is not that complicated. In the past, there have been efforts to try to change straight-line depreciation, but your point about utilization isn't necessarily the correct one. If they're still making revenue on that chip every year—years 4, 5, and 6—

Yeah, so even if it just makes a dollar a chip, it's still generating revenue for them. Remember, the cost of electricity and the cost of running the data center is still an expense in that period. All of that shows up as an operating expense.

If it's generating negative profit, negative gross profit, the market sees that. And I will say one more thing that I think is really important: they would turn it off. They'll tolerate it to a point, and then they'll stop.

Look, there's no hidden information here. Burry's implication that they are cooking the books or hiding accounting is completely false because all of the accounting is apparent in the cash flow statement and in the balance sheet.

Remember, companies have 3 financial statements: an income statement, a balance sheet, and a cash flow statement. The cash flow statement reconciles the income statement and the balance sheet, makes the linkage, and shows you all the cash that's going in and out of the company.

Many analysts and investors who are intelligent and do their homework will look at the cash flow statement. They will see the CapEx and all the investments going out, and they will calculate a number typically called free cash flow. That allows them to estimate the true cash generation of the business in a particular period and assess whether the company should be valued on free cash flow or on the GAAP standard of EBITDA.

The investor has the choice of how they want to value the company. Burry is incorrect in thinking that they're hiding anything because it's all there. They're following GAAP standards, and then investors make a market. They all decide, “What do I want to value this company on? Cash flow? EBITDA?” Let them choose, and then the market sets the price.

I think we've given this guy way too much airtime. He's not very good at what he does.

Come on the program, Burry. We'd love to have you. Why would we click on it?

Is there any other random person out there on the internet you want to take on? Let's just use a Magic 8 Ball to generate numbers and names.

I would say that there are a lot of people who think highly of his analysis.

They're right, and that doesn't mean he's good.

I think it could be a good conversation. You know who we've never had on the pod? Stan Druckenmiller. Let's get Druck before we get Michael Burry.

Have both of them. I mean, why not?

Google released Gemini 3, and it's pretty great. They regained the lead on most of the benchmarks. Polymarket now has Google at 89% to finish the year as the top LLM.

All the speculation that Google was going to have its search franchise absolutely slaughtered by ChatGPT has turned out not to be true, at least not this year, with searches and revenue both going up. But the big story is the speculation around Gemini 3 being trained only on Google's TPUs, not Nvidia's GPUs. Your thoughts, Chamath?

I think TPU is an incredible product. Unbiased, but I think it's an incredible architecture. This latest spin is very profound.

But I also think that what we're quickly seeing is that there's going to be a highly fragmented layer of decoding chips in the marketplace. Groq is one, TPU is one, Microsoft has a spin, Amazon has Inferentia, and Facebook, I think, is spinning up its own silicon. So, we're going to get to disaggregated decode pretty quickly.

The question is, who will win? There'll be a bunch of different solutions. What's incredible about Google is—I don't know if you saw the stats—but they went from roughly 8% share to roughly 16% share of the entire chat market as of this last month. That's an incredible stat.

On the enterprise side, Anthropic is absolutely crushing it. So, what are we seeing? We're seeing a nascent market get created. We saw an allocation of traffic that basically favored one company over everyone, and now we're starting to see a sorting function and a classifier in all of these different markets.

It's breaking apart. There'll be winners in science, winners in enterprise-level coding, and winners on the chat side. Where are the advantages going to be? On the enterprise side, it's going to be model quality. Anthropic's is excellent. On the chat side, it's probably going to pivot around your existing inherent distribution.

That means your operating system, your browser, and your phone—which means Apple, Google, and Microsoft.

I agree that Google has done an absolutely incredible job defending search. But I think what this creates is the setup where now they can cannibalize themselves versus having their market cannibalized for them.

I'm going to take the other side of it. I think what's going to happen is the AI gains in advertising targeting and the number of searches is going to go up. So, while the revenue per search might go down, the number of searches goes up, and then the targeting goes up.

I'm going to take the other side of it. I think their search franchise is going to grow and that Google is not going to lose to ChatGPT. I think the big loser in all this is going to be OpenAI because they started with 100% of the market, and they're only going down. They're facing Google firing on all cylinders, Anthropic, and Grok beating them in the leaderboards pretty consistently.

I think the short in all of this, if you were going to put on the pair trade, is short OpenAI, which I think is overvalued and is going to go down. I think I would be long Google, Grok, and Anthropic.

I think they’re going to have many challenges. I’ll just add this as my final thought on it: the startup community is not trusting OpenAI with their data. If you use OpenAI and you see them releasing products like Sora, if you were in the space of doing image generation or social networking, why would you trust OpenAI with your data?

If you're doing a Cursor and OpenAI has that product, they're not going to trust them. They're going to go with a model like Anthropic, which is taking a more neutral approach: we're not going to go to the application level. They're also using DeepSeek and open-source models because they don't want to give their data and their advantage over to a person, enabling somebody who might compete with them.

Yeah, go ahead. Make two comments.

Sure. Number 1 is, I think what we will see over time is probably a differentiation from general-purpose workhorses in chip architecture to more of these special-purpose chips that work well with certain models and certain applications.

You can think about inference in machine vision and robotics. You don't necessarily need an H100 to do that. You can use a purpose-built chip to do that in a way that reduces power costs and ultimately reduces the capital cost to deploy that in an edge environment.

In the core data center environment, you may end up having models that are different for graph neural nets versus LLMs. There are going to be different chips that'll likely fit very differently with different architectures. So, I would say that the general workhorse is what we had, but now that everyone's making these investments, you should expect that the investment dollars in chip design are only going to ramp up, not down.

Massively.

There's going to be differentiated chips for different markets and different applications.

100%. And that's where there's a risk to Nvidia. Who do you think has the best chance of challenging Nvidia?

So, this is where I was going to go. The other black swan that I think is missing in the equation today—and my early prediction for 2026—is Huawei.

I think there is lithography technology that exists in China that is not publicly discussed, and that is going to be deployed in Huawei and all these fabs that they're building in mainland China. Huawei can create, at a very low cost, probably very high volume, and probably in reasonably short order, chips that can start to rival, for certain market applications, chips that might be expensive and have long lead times.

Give me a timeline for that. 2 years, 3 years out?

They're going to start to have an impact on Nvidia. I think they're going to start to make announcements. And, by the way, remember, chip architecture—and even Jensen's talked about this—is being redesigned with AI. So, AI can design better chips.

Okay. So, announcements in 2026, impact in 2027?

Probably fair.

Sure. Yeah, love it. Is this what it'll be like when we have to be in a studio, when we get to this level of scale where our show actually matters and we need to be in a studio?

We could be in a studio together.

Yeah, I mean, we'd have to. The show would have never happened or worked.

Yeah, I mean, you have 4 people with actual schedules and jobs. It's not like we do this for a living. Oh, you do.

No, I mean, I invest in 100 companies a year.

Not well, but I'm saying you do it.

No, actually, I literally just got back from Founder University in Tokyo.

That's what I'm saying: watching. That's the keyword, watching. You didn't say doing.

You said Founder University in Tokyo, thanks to my partners there. Did you watch that, too? I'm investing in 100–150 companies per year and launching the 5th Launch Fund next year.

I got a check from you, by the way, recently.

Yes, good. I appreciate that. That first fund is like 5–6X now.

Yeah, but it was on $8 million. Have you hit a 5X fund yet?

Yes.

Okay, great. Welcome to the club. Welcome to the club.

On $500 million, so it's—

Okay, great. Awesome. Well, let me tell people it's actually more than a 5X.

Okay, great. I'm happy for you. I'm happy for you.

I should have just done it with all my own money.

That is actually the question that a lot of people have. Do you feel you're a better investor when you're investing your own money, or do you think you're better when you have the discipline of having to report to LPs?

It's actually a good question. My returns have been better when I've been by myself, but I think that there is something really valuable about working for other people, which does keep you accountable.

What's happened is my dispersion has increased massively investing on my own, which means I cut the losers off far later than I would have if I was running a fund. I think what I signed up for when I was running a fund was: never lose money, ever; return the money as quickly as possible; and then run the upside.

I would have traded a 7X with high volatility for a guaranteed 3–3.5X because I think that was my responsibility as the GP. My LPs were Memorial Sloan Kettering and the Mayo Clinic. I wanted to give them the money back because they have programs.

Right.

And it's not my job to hold the money back. With myself, I can keep it out, so then the ups are higher, but then the lows are also lower because some of these things just get annihilated. Look at Relativity Space. I took a $400 million goose egg.

And this is the challenge for Friedberg. Eric Schmidt shows up, and he's like, “Here, it's a billion dollars, pay to play.” And I'm like, “Okay, I'm not going to do it.”

Friedberg, you had a venture studio for a little while. You had to deal with outside investors. Now you're obviously in the driver's seat, CEO of Ohalo. You also had that pressure. You have to answer to LPs. Did it make you better at the job, or did it make you—

Same investor. My venture studio owns the majority of Ohalo. It's our biggest driver of value, so I'm spending all my time on Ohalo. That's kind of my gig.

Do you run The Production Board still? Are there investments that—

I'm on a few other boards, but no active investments.

Are you doing any new investing?

Actually, a lot of folks moved into Ohalo or moved out to stop doing new investing. Slowly, as we have a liquidity event, we'll do a distribution, but the goal is for TPB to end up being a holding company with just Ohalo in it. That's where all the value is going to come from.

We actually just did a distribution, and then we're going to do distributions as we have other events for the other things that are in our portfolio. Then we'll just focus on Ohalo.

When you came into this venture studio model, did you anticipate—which is what most people do anticipate with venture studios—that you'd have 1 breakout and you'd go all in on that?

No, I was delusional. In 2011, 2 years before I sold the Climate Corporation, I started a company called Metromile. I was the chairman of the board of the company. I hired an outside CEO, fired him in a year, and promoted the CTO to be CEO. For years, I worked with him as the chairman of the board.

I invested close to $10 million of my own money in this company and spent years on it. It had raised a Series B, Series C, and Series D and was doing great. I thought, “Man, this is awesome. I can be a chairman, not a CEO, and run these companies. This goes great.” It scaled to whatever it was, $100 million of revenue.

I started Eatsa, which, as you guys recall, was this quinoa fast-food restaurant.

Robotic as well.

Robotic, 20 years ahead of your time. I put $3 million of my own capital in the business, and then I had a CEO run it. We raised outside money, which I was not planning to do in that business.

I thought, “Man, I am so good at starting companies and being a chairman. This is what I should do.” That's what led me to start the venture studio.

Both those companies ended up being net-negative returners for me. Over time, many of the other projects that I was a founder of but chairman of didn't succeed financially.

And over my years being on boards, I realized how frustrating it was to be on a board where you would tell a CEO a bunch of stuff. They wouldn't listen. They would do whatever they wanted to do. I was frustrated, pulling my hair out, watching them do things I wouldn't do and not doing the things I would do.

So, after many years of business failure after failure, I realized this was the moment when we had this amazing outcome—a series of outcomes at Ohalo. It had been a research project for several years. We put close to $40 million into this project before these results started to come in. I'm like, “Holy—this is the game-changing business of my career. This is the power law.” That's when I made the decision: I'm going to go all in on this, and I'm going to run it.

How did the LPs take that? How did you communicate it to them?

Everyone was very supportive and very active. They were like, “This is exactly what we always hoped you would do with the fund: find a winner.” I never thought that was what I would do, because I swore after I sold The Climate Corporation that I would never be a CEO again. It's too stressful. It was damaging to my health. It's overwhelming. I'm so into it; I cannot stop building the business. It consumes me. Everything about it—I have to win. I have to make the business an enormous success. It consumes me.

I knew that it would happen to me again, and I've got kids and all this stuff. So I really had to dig deep to make the decision to do it. Actually, you know what changed my mind about this? I saw the movie Oppenheimer in IMAX, and I left that movie and cried. I realized I wasn't doing what I should be doing with my life by being a board member who was useless. I'm like, “What am I doing here?” I said, “I'm going to make this time.”

I'd been thinking about this, and I'm like, “That's it. It tipped me over.” I made the decision to step in as CEO. So, yeah, the LPs and the investors were all thrilled, because they had all said, “We hope you would run something one day.” Then more capital came in, and we've been running Ohalo for 2 years now. This month marks 2 years that I've been running Ohalo as CEO. I'm really happy I did it.

That's incredibly inspiring. Chamath, would you like to mock Friedberg for crying at Oppenheimer in any way? I saw you doing it. He's doing it. He's like, “Yeah. Great movie, by the way.”

I cried when I got married. I cried when my kids were born. He cried at Oppenheimer. He's like, “Oh my God, he split the atom.”

Do you ever ask yourself, “What do I do with my life?” Do you ever think, “The impact I thought I would have in my life has a missing piece to it. There's something that I didn't accomplish that I always expected I would. There's something I didn't achieve as a person”? Then you see these extraordinary outcomes that others achieve, and you're like, “Man, what was I doing? What am I doing with my life? What am I doing?”

I think that's a pretty profound insight you had, and it sounds like you made the right decision.

All right, next up on the program, one of our favorite human beings. You know him from high-stakes poker: the one, the only, the madman of the poker tables and the mensch in our poker group, Alan Keating. Welcome.

What's up, Chamath?

Thank you for the kind introduction. Have a seat. Have a seat. What's up, brother? How are you?

You're sitting next to me. You're sitting next to me.

Oh, good. Good. Good. Now, Alan, just like at dinner last night—yeah, you were on my right.

Let me introduce you. Chamath, you were not at dinner last night.

Hold on. Let me just do a proper introduction so the audience understands who Alan Keating is. We'll get to it. Alan Keating—just type that into YouTube and watch a bunch of Alan Keating clips. Alan, very famous player. By the way, Alan Keating ran the high-stakes game here, the elite big game, for many years. Then he started investing in companies. He was the seed investor in a little company known as Polymarket, and he's gotten into our friend group. I don't know how many years ago Chamath brought him in, but he fit right in, and we started hanging out 10 years ago and got to know him.

In that time, he also started to play high stakes on TV and stopped running the big game here in Vegas. On TV, you're known for playing way above the rim in a way that, to call it nontraditional, would be an understatement. Take us through it.

Yeah, look. Listen up. Keating, at his core, is an exceptional player. Not a traditional player like some of the other guys who are more solver-oriented, but Keating has incredible live reads. He's kind of like a modern, younger generation of Phil Hellmuth.

I mean, Hellmuth has good live reads, but now he's like an aging horse.

Yeah, he's older. We're going to send him to the glue factory soon.

But Keating is in his prime, and what he can do is soul-read people, which, when you're playing at the high stakes, honestly, that's all that matters, because you can't play solver-based. At the stakes at which we play, you cannot. You're just going to get run over, and that's why you see him being able to do these things, because everybody else steps into the game and they're like a deer in the headlights. They're so afraid, and he is very comfortable. Like when he picked off Doug Polk with a four. How does that happen? It's because he can soul-read people, and he's attuned to play this game. And it was great. It was incredible.

Keating, true or false: Can you soul-read people?

Yeah, I think I've been navigating fear at the poker table better than most. When people are afraid, they tend to give things away, get scared, or act differently.

The Doug Polk hand was that he messed up the bet sizing on the turn. He ships the turn, you fold. He puts himself in a horrible situation where he's folding the river. I mean, you just soul-read him.

There were a few different tells that I wondered if I should delve into, but by the way, I think this is important: There was the same tonality and the same cadence about an hour and a half before then. He was just stealing and air-balling the situation. There was a myriad of things where it was like, “That might be something. I'm not sure. That might be something. I'm not sure.” Well, here are a lot of things that might be something, and I'm pretty sure the combination of them leads to this.

By the way, I think this is important, because a lot of people think poker became solved because of computers and AI, and everyone uses trainers now. But to your point, and to Chamath's point, at the core of the elite level of the game, it's still very much tells and psychology and reading.

We'll get Nick to play this hand picture-in-picture. How big was the pot in the end? A million-something?

No, probably $600,000 or $700,000.

$600,000. Doug Polk, who's a phenomenal heads-up player, had ace-king—like, the best in the world.

You, like a dingdong, had 4-2. 4-2. Playable.

$75,000. Raised flop? It was like what—$150,000 on the flop? $75,000 pre, $35,000, $75,000 all-in.

Okay, so take us behind the hand. What's the read? And why are you playing 2-4 to begin with? Why don't you explain that to the people watching?

Well, it was a big bluff.

Yeah, but explain the thinking there, because a lot of people want to, like you said, put everything into a solver and reduce something into a vacuum and navigate that situation. I don't really have a passion for that. I have a passion for what's happening in this moment, what's happening with this person, what's happening with me, and what they're perceiving me as.

In that moment, Doug had gotten some confidence around a couple of hands, and there was a player in between that I knew he didn't think much of his hand, and he didn't think much of my hand. It seemed like an obvious situation for him to pull it away from me.

I thought about reraising all-in preflop just to simplify it, and I think that was probably a better way to do it. But at the same time, I do a lot of things for the fun of it. I thought it would be a little bit more fun to get him in a spot later on, a couple of streets down, where I could bluff him out or call him down.

Did he have a tell on the turn?

Yeah. He goes, “$75,000,” kind of directly. About an hour and a half before then, he had the same tonality and the same cadence, and he was just stealing and air-balling the situation. There were a few different tells that I wondered if I should delve into, but there was a myriad of things where it was like, “That might be something. I'm not sure. That might be something. I'm not sure.” Well, here are a lot of things that might be something, and I'm pretty sure the combination of them leads to this.

What happens if you call and he shows pocket jacks or something? When you call off and lose, how do you process losing $700,000 that way? Do you think, “This is so stupid. Why did I do that?” What's the self-talk? What's the internal monologue?

I don't know. I guess I don't know where it came from, but I've always had a sense of humor around whatever happens to me—the things I can't control and the things I can control. If I put myself in a spot, I've gotten to a point where I can immediately recognize how ridiculous what I just did was and kind of laugh about it.

You're forgiving yourself during the hand?

Yeah, sure. There'll be an internal part that's just like, “Yeah, of course that was stupid. Why am I doing this? So dumb to do that.”

To be able to pull off the hands you pull off, you have to have an “I don't give a—” kind of attitude about it. This is all a game. I don't take it too seriously, because that's where fear comes from.

The fear, yeah. I think I've just recognized that people make bad decisions when they're scared.

But sorry, say that again: mastering fear. You put in reps? How do you put in reps to master fear? Because that extends to many other things in life.

Right. Absolutely. We talked about this in investing—different strategies about how big of a bet you want to make relative to your bankroll, and in venture and stuff like that. I like making the bet where, if it doesn't work out, I'm in a little bit of trouble. You like to feel the pain. You like that it feels real.

It's a motivator. It's something that drives me, and I like poker. I'll put myself in the same type of situation.

Just like 2 years ago, Keating calls me. He's like, “Hey, this is about portfolio construction and a specific company.” We won't say the company. We talked for an hour, and I'm trying to give him my best advice: “Look, here's how you structure it to minimize volatility. Take some of these chips off the table. Do this. Do that.” He goes, “I really appreciate this.” He calls me 2 days later: “Yeah, so I doubled down on this thing—and, okay, you don't know this, but I've tripled down.”

Here's the interesting thing: If you felt that was where you were going to go, why are you checking it? Were you trying to check your sanity, or what were you doing there? Why do you check?

I'm inviting him to the deep end. I'm saying, “You want to come with me?”

No, no, you're talking about me? No, no, I'm asking why he called you. No, no, I'm asking why he called Chamath in that situation where you're going to double down, triple down.

I have access to someone who's infinitely smarter than the thing that I'm trying to understand, right?

You're jumping in the deep end.

Yes, yes.

Why are you asking him, like, “Should I jump in the deep end first?”

Because I want to earmark all the reasoning. I'm trying to understand everything about this decision because I'm going to live with the outcome of that decision no matter what. I want to remember his take, my take, my feelings, and other people's thoughts, and put that into a little folder that I can come back to.

You've unconsciously—or you've discovered—something that's referred to as superforecasting in behavioral sciences. If you write down and understand all the permutations of your decision-making, and then reflect on it years from now, you'll just be better at decision-making. That's actually what you're doing, and I think it just comes naturally to you. You were going to say, when you thought the question was about the poker hand, something about inviting him to the deep end.

Yeah. Unpack that concept of saying, “Hey, we both know that this hand's out of control. We're in the deep, dark waters. There could be sharks in there.” Explain what you're doing, because I've been in hands with you where I feel like you just dragged me out to the deep water, where I've got jacks or queens, and then all of a sudden I'm going to be playing for my entire stack, and they don't feel good anymore, even though I have an overpair to the board or whatever it is.

Yeah, that's a great point. I just think there's some purity or beauty in the chaos after everyone's—after you get past where everyone's prepared. I'm interested in that space, and I have no interest in the space that everyone's prepared for.

Everybody's got a plan until they get punched in the face. You remind me of Alex Honnold.

And you see—oh my God, his bestie. Look, guys, Phil Hellmuth decided, for this special occasion with us here taping All-In for the first time in Vegas, to wear a tracksuit that's only 12 years old. It was one of his newer tracksuits.

Phil Hellmuth

No, no, this shirt you guys bought for me.

At Hellmuth's birthday, we each chipped in, I think, $3,000 to buy him a new wardrobe. It was like $70,000 in total. There were 20 of us. We put in $3,000.

Phil Hellmuth

No, no, hold on. Between that wall and that wall, you guys gave me a whole ton of new clothes. No, no, I gave them to my sons.

Oh, okay. Sit down. PHIL HELMUTH, PLEASE SIT DOWN. I GOT TO close the show. Hold on. Supplemental. Yeah, three, two. All right, Alan Keating, you're a mensch. It's a pleasure to know you. Great to play with you. And we're going to have some exclusive content on our YouTube channel of the besties playing poker with incredible professional poker players like Jason Koon, Alan Keating, and Phil Hellmuth. [Applause] You let your winners ride. Rain Man David Sax. And as I said, we open sources to the fans and they've just gone crazy with it. Queen of quinoa. [Music] Let your winners ride. Besties are back. And it's my dog taking a nice scenic drive, David Sax. [Music] We should all just get a room and just have one big huge orgy because they're all just useless. It's like this sexual tension that they just need to release. What? You're beat beat. What? Bare your feet. Beat. What? We need to get merch. These are back. I'm going all in. [Music] I'm going all in.