Brad Gerstner: More Trillion Dollar IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts

Episode 280 opens as a scheduling joke: Friedberg is on vacation somewhere unspecified and Brad Gerstner takes the fourth chair. Chamath calls in from Paris after roughly eight interviews at the RAISE conference; Sacks is in his flag room on the East Coast after a week in Washington for America's 250th anniversary. The banter looks like filler, but it establishes three distinct vantage points that never fully reconcile later in the hour. Gerstner is the public-market operator who prices risk. Chamath is the enterprise-software operator who is currently living through the cost side of the AI trade. Sacks is the policy operator who has spent the week in rooms with sovereign governments and the White House. When they disagree about token economics, they are disagreeing from three different data sets, and that is the most useful thing about the episode.

The first substantive thread is Sacks describing a Geneva dinner with Marc Benioff and a UN Commission on AI that seated him alongside Jensen Huang, Brad Smith and Grab's Anthony Tan. His read on Benioff is admiring and slightly anthropological — "the impresario of impresarios" — but the payload arrives when Jason asks what the commission is actually for. Chamath's one-word answer is "Open source," and Sacks then offers the thesis that organizes the back half of the show: no country wants to subordinate itself to a closed American model, so "the future is open source for all these countries."

Jason parks the thread immediately — "we're gonna get into that. That's on the docket for sure" — and pivots to IPOs. The sequencing is deliberate. Open source as national industrial policy is the frame; the IPO segment is the evidence about who is capturing economic value right now. Everything after the break is an attempt to answer whether those two stories are consistent.

The scoreboard is unambiguous. SpaceX trades near its IPO price — roughly $150 a share against a $200 peak — for a market capitalization of about $2 trillion, making it the seventh-largest company in the world. Anthropic filed confidentially on June 1. Polymarket puts a 65% probability on an Anthropic listing this year on thin volume. Gavin Baker's estimate, relayed by Jason, is that Anthropic exits 2026 above $100 billion in revenue, "very profitable," and would trade at $3 trillion if it listed today. Those three claims are doing a lot of work in the conversation and none of them is audited: the revenue figure is a rumor, the $3 trillion is a guess made on air, and the probability market is thin enough that it should be read as sentiment rather than odds.

Gerstner supplies the deal mechanics that the next two issuers are reportedly copying. SpaceX "raised $75,000,000,000 at 1,750,000,000,000," is up about 25% from issue, and trades near $2 trillion against roughly $35 billion of forward revenue. His emphasis is on process rather than multiple: total raise size, pricing, float, a lockup structured as staged release tied to milestones, and early index inclusion. On index inclusion he does something unusual for a bull — he concedes the objection was legitimate. A newly public company has no post-IPO track record, and "the peak to trough drawdown in the six months post IPO is 50%." The resolution was to modify index rules rather than exclude a company whose size makes exclusion untenable, which is a defensible answer to an genuinely hard problem rather than a favor to Elon Musk.

His justification for why the old rules existed is the most transferable idea in the segment: the rules were written for companies that were younger, earlier, less tested, with less revenue and less profit, where inclusion risk was asymmetric. A $2 trillion issuer inverts that asymmetry, because excluding it distorts every index that tracks the market. That is the template he expects Anthropic to inherit — not the valuation, but the plumbing.

On sequencing, Gerstner is careful in a way that reads as genuine rather than coy. He expects both to list, doubts there is a real race, and thinks OpenAI's corporate restructuring adds complexity that makes an earlier Anthropic listing more likely — while saying plainly "I don't know." His stated position is that Altimeter would be "a buyer at scale and at size in both of those IPOs." Chamath's counter is a market-clearing argument rather than a valuation argument: the binding constraint is how much new paper the market can absorb and at what price, and the two companies sit in different cash-burn positions because OpenAI is more consumer-weighted and more diffuse.

The most concrete number in the episode comes from Chamath's own company. His CTO reports that token costs are "doubling every forty five days." When Chamath asks what the downstream productivity gain is, the answer is "maybe 5% max." The explanation is the part worth keeping: reaching the next iteration of improvement now requires far more tokens because the easy gains have already been captured — "we've effectively already asymptoted." This is a claim about marginal returns at the current capability frontier, and it is testable in a way that most AI commentary is not. It also cuts against the optimism elsewhere in the hour without contradicting it, because both can be true: aggregate demand can compound while the marginal return on the next dollar of inference collapses.

His conclusion is a timing argument, not a technology argument. If costs compound while the productivity upside stays flat, the rational corporate move is to raise capital and exit before that arithmetic becomes visible in public filings — "if you can get out now, you should get out now before all of that starts to seep into the water table." He expects every company to run this reckoning within three or four years; the only variable is who gets there while the market is still paying for growth. Note what this implies about the IPO window he discusses earlier: the argument for listing in 2026 is not that the businesses are mature, but that the disclosure regime has not yet caught up with the cost curve.