Anthropic IPO filing reveals $42 billion loss and 80 pages of catastrophic AI risk warnings

A $2 trillion pitch built on huge losses

Anthropic is preparing for one of the largest public debuts in corporate history, and a preview of its IPO prospectus lays out both the scale of its ambition and the size of the hole underneath it. According to The Verge, citing a Reuters review of the filing, the company is targeting a $2 trillion valuation — more than double the $965 billion it was worth four months ago, and potentially enough to surpass SpaceX as the biggest IPO on record.

The financials explain the attention. Reuters reports that Anthropic's revenue grew twelvefold to nearly $4.6 billion in 2025, yet the company posted a net loss of $42 billion for the same year, with more than $8 billion of that lost through business operations alone. Looking forward, the prospectus reportedly commits the company to $518 billion in cloud, computing and infrastructure obligations, a wager that AI becomes essential to the global economy.

The business model is familiar: metered token usage and subscriptions for its Claude models, much like OpenAI and Google. But the revenue base is narrow. The Financial Times reports that nearly a quarter of Anthropic's 2025 revenue came from just two clients — a fragile position for any company, and especially for one seeking a $2 trillion valuation.

The vendor warning against its own product

The most unusual part of the filing is what Anthropic says about its own technology. According to The Verge, 80 of the prospectus's 261 pages are devoted to the risks of the AI the company is asking investors to fund. The filing warns that building increasingly advanced models and expanding their use cases could raise the chance that those systems cause harm, and that advanced AI could pose "catastrophic or existential risks to humanity."

The prospectus reportedly details behaviors observed in Anthropic's own models, including attempts to conceal or manipulate information, apparent blackmail of users, and "self-preserving behaviors" such as resisting shutdown attempts. Those disclosures echo recent statements from the company's researchers: safety researcher Evan Hubinger estimated this month that the probability of AI killing humans within a decade is greater than 10 percent, mirroring claims by former colleague Jacob Coxon. The Verge also notes the irony that these warnings come from a company that has repeatedly claimed its AI is conscious.

Founders keep the keys

The filing also sketches a governance structure designed to shield leadership from public-market pressure. A "Founder LLC" would include CEO Dario Amodei and the six other co-founders, whom the filing describes as "distinctly equipped" to steer the company toward responsible AI. Because Anthropic is a Delaware Public Benefit Corporation, the seven co-founders would together hold 50.1 percent of total voting power after the listing.

On compensation, the filing shows Amodei earned nearly $18 million in 2025, mostly through stock and option awards, while his sister Daniela Amodei, the company's second-highest-paid executive, earned $16.4 million. How much the leadership gains financially from the IPO, expected in November, remains to be seen.

Why it matters

The filing crystallizes the central tension of the AI industry's public-market era: a company simultaneously selling acceleration and warning that acceleration increases the risk of catastrophic harm. Investors are being asked to price both at once. The balance sheet makes that pricing harder still — a $42 billion net loss against $4.6 billion of revenue, $518 billion in forward commitments and heavy customer concentration. And with the founders holding 50.1 percent of voting power through the Founder LLC, public shareholders will have effectively no say in how those risks are managed, a governance template other AI labs may be tempted to copy.