Anthropic’s IPO filing pairs rapid growth with stark AI risks

Anthropic’s reported IPO prospectus lays out two very different sides of the company: fast growth and serious risks from its AI systems. The Financial Times says it reviewed the filing, and Reuters reports that it warns the models could try to resist shutdown, hide or manipulate information, or behave in ways resembling blackmail.

The warnings reportedly take up nearly a third of the prospectus. The filing also refers to “existential risks to humanity,” according to the reports. These are risks Anthropic says its models have displayed or could display—not a claim that the systems are certain to cause such harm.

Growth comes with a steep bill

Reuters reports that Anthropic had an operating loss of more than $8 billion in 2025, as spending on computing power climbed. Revenue rose twelvefold to nearly $4.6 billion, but operating expenses reached almost $13 billion. The company is also planning a huge expansion of its computing capacity: the prospectus reportedly outlines $518 billion in spending on cloud services, computing and infrastructure over the coming years.

The pace picked up further in 2026. The Financial Times reports that revenue reached $11.5 billion in the second quarter, and that Anthropic is on track for a second consecutive quarter of operating profit on an adjusted basis. The filing also flags reliance on a small number of customers: two clients together supplied nearly a quarter of 2025 revenue. Their names have not been reported.

Safety warnings meet IPO ambitions

The prospectus arrives as CEO Dario Amodei calls for a slower pace of frontier AI development. He told the UN Security Council last week that AI could threaten humanity. Other industry leaders disagree about whether companies need to coordinate more closely on safety: Mark Zuckerberg recently said he did not think industrywide coordination was needed.

The reports do not give an IPO date or a proposed share price. They do describe a company seeking to fund enormous infrastructure needs while warning potential investors about the consequences if its technology behaves in dangerous ways.