Jason: All right, everybody. Welcome back to the number one podcast in the world. It's the All-In podcast, episode 274. Sachs is out today, but we're very lucky to have Gavin Baker from Atreides Management joining us. It's been a huge week in tech. We can start with the SpaceX and OpenAI IPOs. We've got Andre Karpathy joining Anthropic. Nvidia crushing it. Karpathy is only 39 years old. He's already a legend in the tech industry. He recently built auto-research, an open-source training tool that helps AI models improve themselves by running five-minute experiments. That got over 82,000 stars on GitHub. He's going to be in charge of a new pre-training team at Anthropic. The focus obviously being recursive self-improvement. In other words, they're going to have Claude improve itself. Is this super important in 2026?

Chamath: You know what's interesting if you go back to like Google, the culture of Google which they got right was the singular technical talents there, they were called Google Fellows. What's interesting about Andre is he's been at the wave upon wave of AI. He was probably the first person that really commercialized the Richard Sutton bitter lesson essay when he was leading FSD at Tesla, which was really about the brute force computation. I remember him telling me he spent a quarter of his time labeling data—could you imagine like 2016-17 like hand-labeling video data from Teslas? This idea of recursive self-learning puts these models on a combination of overdrive and autopilot. The model quality just goes absolutely parabolically just like this straight up.

Gavin: The success is extraordinary. It's undeniable. I think the fact that they were EBIT positive per the Wall Street Journal in the most recent quarter is a really important fact for kind of the whole AI narrative because now there's ROI. If OpenAI and Anthropic are at call it a hundred billion dollars of ARR now with 80%ish gross margins on inference, the returns are there. It's not hard to see 200, 300, 400 billion of ARR at the end of this year across all language models. I do think what Karpathy is working on recursive self-improvement is really important. Maybe the two final frontiers for AI are that and continual learning, where the model learns from experiences the way humans do.

Jason: Anthropic has a decent lead on everybody else. But at what point do we think this super recursiveness occurs? Friedberg, what's your take here on the AI PR crisis? We had three different commencement speeches that were booed. Eric Schmidt being one of them. Why are young people booing AI vociferously?

Friedberg: I think that there's an underlying view that technology creates leverage for a small group of people which creates power imbalances, and nothing represents that more than AI. The economic benefit that's accruing to the few today becomes the narrative. Secondly, I think that there's a deep amount of external energy that's fueling this anti-technology sentiment in the United States. I think this goes all the way back to KGB design during the Cold War. Thirdly, AI shifts and messes with the ego of the human. It's almost anti-humanist, and I think that's a deep psychological current in a lot of people's disdain for this technology.

Jason: I think we have to recognize that the layoffs are not just the bloating issue anymore. Matthew Prince, CEO of Cloudflare, laid off more than 20% of his workforce. He says he's getting rid of "measurers"—people who manage people and measure data—because they are unnecessary because of AI. At the same time, Zuckerberg did another round of layoffs and he told everybody we're putting recording software on every single person in the company's computers to study and train our model. This is scaring the be-Jesus out of people.

Chamath: I thought the Matthew Prince note was horrible. This was like from the PR school of retards. You reduce humans to a label called the "measurer" and then you're like, I'm going to lay off all the measurers. It puts a scarlet letter on their back. I would say shut the f**k up, get behind the keyboard, just do your job. Don't write these missives. You're all terrible at it.

Jason: Topic two, SpaceX just filed their S1 on Wednesday. They are aiming to raise 75 billion at a 1.75 trillion valuation. This will be the largest IPO ever. Starlink is the money printer right now, 11.4 billion in revenue. AI did 3.2 billion in revenue but had 6.4 billion in operating losses. But here's the big one: EWS, Elon Web Services. Anthropic is paying SpaceX 1.25 billion a month to rent out Colossus 1 and parts of Colossus 2. It's a $45 billion deal over three years.

Gavin: Elon Web Services makes me laugh, but it means the AI business is going to quadruple. They build data centers dramatically faster than anyone else—the first one was 122 days, the second 91 days, the third 66 days. Cursor's Composer 2.5 model came out this week and it is Pareto dominant after doing reinforcement learning on Colossus 2. As far as when will orbital compute be a reality, I would say second half of '28 to first half of 2030.

Chamath: How do I underwrite SpaceX at 2 trillion? I'm buying the most important internet infrastructure project (Starlink). I'm buying a delivery infrastructure. And I'm buying an AI business. They are the closest to delivering a gigawatt data center. What he creates is a capital moat that then accelerates a technology moat. That thing will look very cheap, I think, in a few years.

Jason: Nvidia blew out its earnings again. 81.6 billion in revenue, up 85% year-over-year. 58 billion of net income with 75% gross margins. They've announced another 80 billion in additional buybacks and raised the quarterly dividend 25x.

Gavin: The AI market is cross-sectionally inefficient right now. I think there was one other really important thing in the Nvidia quarter: they said their CPU business was going to be $20 billion this year. It means overnight they're one of the world's largest CPU manufacturers. The DSA (Domain Specific Architecture) market evolution is actually happening inside of Nvidia.

Friedberg: Global debt to GDP is 310%. The spending problem ultimately creates a cascading effect. As it starts to break, you have massive inflation. 30-year Treasury is at 5.2%. This Japanese yield might be a catalyst for a credit crisis. This is water leaking out of the bucket.

Chamath: There are signals that are flashing. Generally everything else, I think you should not speculate. I have a few companies that I really believe in. Five or less.

Jason: We had a 48 hour jaunt by a bunch of tech CEOs and the president to hang out with Xi. Other than a little business development, was it just performative?

Friedberg: The grand deal that de-escalates tension probably didn't manifest as some had hoped it would. And it's no surprise that Putin is with Xi today. The story continues.

Gavin: I think selling deprecated Nvidia GPUs to China lowers the odds of them developing their own alternative ecosystem. I think there's sound arguments that this is stabilizing for the world.

Jason: All right. Let your winners ride. See you next time.