All-In PodcastAnthropic S-1 Shocker, Greatest Business Ever? Google’s $500B Wipeout, Midterm Predictions
All-In Podcast
Podcast banter 0:00
The hosts trade jokes about being the world’s top podcast, fake news, and who said they had cooled on the word superintelligence. Sachs says the whole document is about superintelligence, and the others tease him about being photographed on an episode he was not on. They also greet Friedberg and Chamath, then drift into friendly talk about a long card game and how much one of them talked.
Anthropic’s S-1 3:32
The discussion turns to Anthropic’s S-1 and the claim that it may be the greatest business in capitalism. They describe a planned November 16 IPO, a target to raise about $100 billion, and revenue of $16.3 billion in the first half of 2026. Sachs says the old $80 billion loss story was not a real cash loss, and that the company looks capital efficient, with rising margins and strong demand.
Compute as the bottleneck 16:00
Claude is described as letting people do more work in less time, and the speaker thinks pricing could soon split into cheaper tokens for easy tasks and higher prices for harder ones. He then turns to Anthropic’s S-1 and says the biggest risk is compute. The word cloud of risk factors puts data centers, power, chips, the grid, electricity, and hardware at the center, with regulation, privacy, IP, AI safety, competition, and governance following behind.
Revenue, cash, and scale 26:31
The revenue side is used to show how broad the customer base already is. He says about 40 million individuals and small businesses spend under $100,000, about 10,000 spend between $100,000 and $1 million, about 1,650 spend between $1 million and $10 million, and only 250 spend more than $10 million. He argues that the business could reach hundreds of billions in a few years if a modest share of the market converts, and says Anthropic has over $125 billion in cash while spending very little to get there.
AI Capital Pressure 31:30
Anthropic is said to be raising 100 billion in its IPO, with another 80 billion in debt capacity, giving it more than 300 billion of dry powder. That, the speakers say, makes its future compute and CapEx commitments easier to back. The smaller capital raise also gives it more pricing power and lets it manage the float better, which they treat as a constructive setup for the offering.
Governments Versus Compute 33:00
The discussion then widens to the idea that the United States economy is tied to a small AI complex. The biggest companies are spending heavily, draining cash, and turning to debt, while Microsoft is singled out for cutting its debt. The group says the AI firms may need to raise amounts equal to about 30% of all Treasury issuance through 2027, which means they are competing with government borrowing for capital and may draw regulatory pushback as compute is assigned and controlled.
Market risk and rates 47:00
Two risks are laid out. One is simple to track: how fast Anthropic’s 10,000 customers move from small spend to multi-million-dollar deals as they prove return on investment. The other is harder to price. The AI complex has become large enough to compete with U.S. Treasuries for capital, so someone has to pay more. If Treasuries pay more, borrowing costs rise across the economy. If the AI firms pay more, defaults can come sooner inside the complex.
Scale, efficiency, and scrutiny 51:00
The business is described as huge enough to sit beside the best franchises in American business, with revenue targets imagined at 100, 300, and even 600 billion dollars. At the same time, the group returns to the risks: open models may commoditize parts of the market, especially coding, and Anthropic may have to lean harder on recursive self-improvement to keep ahead. Usage is still early, with many firms holding back because of cyber fears, but the demand could rise sharply if those limits ease.
Google's mixed picture 1:03:00
Google’s core search ads are under pressure, but the wider company still looks stronger than a simple decline story. The cloud business has moved fast, with operating income rising from 3.5 billion to more than 11 billion in a year. That makes the infrastructure more valuable, while also raising the question of whether Google can build enough consumer AI to replace any slowdown in search revenue.
Search and election noise 1:11:30
One speaker says search revenue is still rising, from 56 billion to 63 billion, even if it is a smaller share of total revenue now. Another says the market is overreacting and that the company’s other businesses are finally taking hold. The discussion then shifts to the midterms, with talk of a Republican Senate, a narrow Democratic House edge, and heavy distrust of polls, while Polymarket is treated as a better guide than surveys.
Midterm forecast 1:18:30
One speaker says the media is overstating the midterms and that a normal result is more likely. Another says Democrats will win the Senate and House, pointing to Trump’s weak approval, anti-Trump turnout, a split Republican Party, and the idea that moderates will stay home. He expects impeachment, subpoena fights, and a two-year mess.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
