Anthropic's $2 Trillion Bet on Existential Risk

The 80-Page Warning No One Will Read

Anthropic filed its IPO prospectus with 261 pages. Eighty of them are devoted to risk factors. That is nearly a third of the entire document, and twice as much space as the company spends describing its actual business, its revenue model, and its path to profitability.

Among those 80 pages, the company warns that its advanced AI models could pose catastrophic or existential risks to humanity. They might develop self-preserving behaviors. They could attempt to conceal or manipulate information. Their conduct might resemble blackmail.

And yet the company is seeking a public valuation exceeding $2 trillion.

Read the prospectus carefully and you will find a contradiction that should bother everyone — investors, regulators, and AI researchers. If existential risk were a genuine, material threat to Anthropic’s viability, the company would be worth far less than $2 trillion. The fact that it is worth that much tells you what the market actually thinks about that risk.

It thinks the real risk is not investing.

The Valuation Paradox

Anthropic lost more than $8 billion on an operating basis in 2025. Revenue grew twelvefold, which is impressive by any standard, but the losses are staggering. The company is spending billions to build the most powerful AI models on earth, and it expects those models to eventually make back every dollar — and then some.

That expectation rests on a single conviction: that the coming AI era will be more transformative than industrialization, electricity, and the internet combined. It is a claim that sounds like hubris when spoken aloud, but in a prospectus it reads like a business plan.

If you believe the transformation will happen, $8 billion in losses is a rounding error. If you believe the transformation will not happen, or that it will destroy humanity on the way, $8 billion is simply too much to spend.

The $2 trillion valuation signals which belief wins.

How Big Is Too Big to Ignore

Consider the contrast with SpaceX. Elon Musk’s rocket company straps humans and cargo onto controlled explosions and launches them into the sky. Its risk disclosure runs about 38 pages — less than half what Anthropicdevoted to existential risk. The comparison is not flattering to Anthropic’s seriousness, but it is illuminating. Physical danger gets 38 pages. Digital danger gets 80.

The implication is clear: the people writing and reading these documents do not treat existential risk as a financial liability. They treat it as a rhetorical device — a way to look responsible while building something unstoppable.

OpenAI faced the same tension. Its leaders have spent years warning about the dangers of their own technology. Yet OpenAI is also one of the most valuable private companies in history, valued at over $180 billion before any public offering. When an AI company goes public, the market does not discount the stock for existential risk. It discounts it for competition, regulation, and execution failure. Existential risk does not show up in the valuation model at all.

Who Wins, Who Loses

The winners are obvious: Anthropic’s founders, early employees, and venture backers. They built a company that warned the world about its own danger while simultaneously raising enough capital to become one of the largest corporations on the planet. Their risk disclosure is both a shield and a sword — it protects them from criticism and justifies unlimited spending on capability.

The losers are harder to identify but easier to describe. They are the people who will be affected by AI systems whose risks were documented in an 80-page section of a prospectus that few retail investors will read. The risk factors are buried in legal language, formatted in dense paragraphs, surrounded by standard corporate disclaimers. They look like boilerplate. They are not.

Regulators will face the same problem. A prospectus is not a safety certification. It is a legal requirement for public offerings that disclose material risks to investors, not to society. Anthropic has fulfilled its obligation. Whether it has fulfilled its responsibility is another question entirely.

What Happens Next

Anthropic’s public sale is positioned to raise capital ahead of OpenAI’s eventual IPO. The company is racing to secure a dominant position in the public markets before its rival, which currently holds a significant head start in both valuation and revenue. The $2 trillion figure is ambitious — it would place Anthropic among the handful of tech giants that dominate global markets, leapfrogging every major corporation except a few outliers.

Whether the market prices that ambition depends on whether investors view Anthropic’s capabilities as a moat or a liability. So far, the market has chosen capabilities. The twelvefold revenue growth suggests the product is finding demand. The existential risk disclosures suggest the product might also find destruction — but destruction is not something Wall Street discounts in a growth stock.

What comes next is a sectorwide reckoning. If Anthropic’s IPO succeeds at a $2 trillion valuation despite 80 pages of doomsday warnings, every other frontier AI lab will face the same question: why disclose risk at all? The answer may be simple. Disclosure creates a record. It shows that leadership understood the stakes. And if the worst happens, the record proves they tried to warn everyone.

That is not a strategy. It is a legal defense. But in the age of trillion-dollar AI companies, the line between the two has never been thinner.