Anthropic's $2 Trillion IPO Comes With an Existential Risk Warning

Anthropic is preparing to go public at a $2 trillion valuation. The prospectus runs 261 pages. Roughly 80 of them explain why the company's artificial intelligence models might pose an existential threat to human civilization.

This is not metaphorical risk language. CEO Dario Amodei spent approximately a third of the filing detailing scenarios in which Anthropic's own technology could cause catastrophic harm to humanity, resist shutdown commands, and exhibit self-preserving behaviors that prioritize its own survival over human interests. The company disclosed these risks not because securities law forced it to—though securities law did—but because the risks appear to be genuine enough that burying them would have been indefensible.

The market has priced this in by doing absolutely nothing.

Pre-IPO perpetual contracts on Anthropic equity remain near $2,000, corresponding to the $2 trillion implied valuation. Investors have decided that the possibility of their investment becoming superintelligent and potentially hostile to human welfare is acceptable. This would make Anthropic's debut the largest IPO by valuation in history, eclipsing SpaceX's reported $1.77 trillion entry to public markets earlier this year. At current annualized revenue rates, the company trades at roughly 31 times sales. The math is not complicated. Neither is the message the market is sending: growth story beats doomsday scenario.

But there is a more immediate problem than existential risk, and it is considerably less apocalyptic and far more relevant to shareholders.

Nearly 50 percent of Anthropic's revenue flows through Amazon and Google. This is not diversification. This is dependency. The company has built its business model on two gatekeepers who can adjust terms, redirect traffic, or simply build competing products with the infrastructure investments they have already made. Amazon holds an $8 billion stake in Anthropic. Google has invested $2 billion. Both companies are essentially funding their own alternative to Anthropic while maintaining Anthropic as a backup option. This is the corporate equivalent of a dating arrangement where both parties are simultaneously hedging against each other.

Anthropicreports a net loss of $42 billion in 2025. The company faces roughly $518 billion in future infrastructure obligations. These are not startup losses. These are losses at the scale of small nations' defense budgets. The company is burning cash to build compute capacity that Amazon and Google could replicate in their sleep. The independence story—carefully managed in press releases and investor meetings—is a polite fiction.

To manage this problem, Anthropic is creating a new governance structure called the "Founder LLC" group. Its sole responsibility will be insulating Anthropic's leadership from public investor pressure and ensuring that AI benefits humanity. In other words, the company is pre-emptively installing a circuit breaker between shareholders and management, explicitly to prevent shareholders from pushing the company to maximize near-term returns at the expense of what Amodei has deemed existential safety. This is extraordinary. It is also an implicit admission that the two goals—shareholder returns and human survival—may be in direct conflict.

The prospectus includes estimates that Anthropic's revenue could reach $120 billion by year-end, which would compress the revenue multiple to around 17 times. This is presented as good news. It is. But it does not resolve the structural problem: nearly half of it still flows through two companies that own pieces of Anthropic and are building their own alternatives.

Anthropic has signed an agreement with SpaceX for up to $85 billion of compute through 2029—nearly double the $45 billion disclosed in May. This is not a sign of strength. It is a sign that the company's internal estimates for compute requirements have doubled. More infrastructure. More burn. More dependency on external partners who have the option to walk away.

Investors have decided all of this—the existential risk warnings, the concentrated revenue base, the cash burn measured in tens of billions, the governance structures designed to override shareholder interests—is worth a $2 trillion valuation. This is not necessarily wrong. It is simply a bet that the growth story overwhelms every other variable, including the one that might end all variables permanently.

The market has made worse bets. But rarely has it made one this explicit.