It's rare in corporate IPO history for a document meant to convince investors to open their wallets to also state that "our product could destroy humanity." Anthropic's IPO draft recently leaked to the media—the document was originally submitted confidentially to the U.S. Securities and Exchange Commission back in June this year, and its contents have now come to light through Reuters reporting. Alongside the financial figures, the document contains a rare piece of self-disclosure: the company's own AI technology could pose an "existential risk" to humanity.
Let's start with the numbers. Anthropic posted an $8 billion operating loss last year, while its net loss came in far higher at $42 billion, even as revenue grew 12-fold year-over-year to $4.6 billion. One key driver of the massive losses is the company's plan to pour $518 billion into data center infrastructure over the coming years—the enormous financing costs are exactly why the net loss dwarfs the operating loss. Even so, Anthropic's pre-IPO valuation is still expected to reach as high as $2 trillion.
The picture isn't entirely grim, though. In Q2 2026, Anthropic turned an operating profit on $11.5 billion in revenue, and the company expects to stay profitable next quarter as well. But hidden beneath that profitability lies concentration risk: according to a Financial Times report, a quarter of this quarter's revenue came from just two customers. The company hasn't disclosed who they are, but a report back in August suggested Meta could be paying Anthropic as much as $10 billion a year. Making matters more precarious, many of Anthropic's big clients haven't signed long-term contracts, meaning they could walk away at any time.
