Anthropic, the San Francisco artificial intelligence company, has filed papers for an initial public offering that could become the largest ever. The filing, submitted on Monday, sets up a Wall Street debut next month and could surpass the SpaceX offering in June, which raised a record $86bn.
The move comes only weeks after chief executive and cofounder Dario Amodei argued in an essay for slowing the development of the most advanced AI models. A listing could hand the five-year-old company vast resources to expand. The amount it will raise, its final valuation and its offering price are all still uncertain. Even at lower estimates, the company would rank among the world's most valuable, and the fortunes of Amodei and his sister and cofounder, Daniela, could grow by billions of dollars.
The prospectus shows steep growth alongside heavy losses. Anthropic's net loss for the first half of 2026 widened sharply to $80.7bn, a result the company largely tied to a rise in the fair value of bonds convertible into its shares. Revenue in that period was about $16.3bn. In the three months ended June 30, revenue rose 1,368 percent from a year earlier, though the company cautioned that the growth rate could decline over time and fluctuate.
Anthropic said more than half of Fortune 500 companies and over 30 percent of Global 2000 firms were customers that each generated at least $100,000 in revenue for it in the 12 months ended September 30. It expects compute and infrastructure costs to rise in absolute terms and to remain its largest operating expense for the foreseeable future.
The filing also sets out the dangers of advanced AI. It warns of misuse, security threats and the possibility that systems could produce outputs that "could be interpreted as coercive, deceptive, or manipulative". It says its models could show "self-preserving behaviours", including behaviour "resembling blackmail". Anthropic's own research has found code sabotage, fraud assistance and information manipulation in controlled tests. The company also said that using AI to speed up its own research may let model capabilities advance faster than its ability to develop and implement safety, alignment and security measures.
Anthropic added that it depends on a small number of cloud providers, including Amazon, Google and Microsoft, and on suppliers such as Broadcom, relationships that can create "dependencies and potential vulnerabilities".
The offering arrives as markets are choppy. Investor unease over debt-fuelled AI spending has driven a global selloff in AI and tech stocks, as doubts grow over whether trillions spent on data centres, chips and computing capacity will pay off quickly. Higher borrowing costs and rising bond yields have added to the caution, and several autumn listings have been delayed. Even so, tech firms still want cash: Meta is seeking up to $95bn in its largest equity raise, and SpaceX has signed a $40bn deal with Apollo to lease AI chips and server hardware.
Reuters reported that some current and prospective investors now value Anthropic at about $1.5 trillion, down from roughly $2 trillion a week earlier, after the company shared some financial details at an investor event. In May, a funding round had valued it at $965bn.
Anthropic will trade on the Nasdaq under the ticker "ANT". The filing did not disclose the number of shares or the proposed price range. Goldman Sachs and Morgan Stanley are lead banks, joined by 20 other underwriters, with Morgan Stanley acting as stabilisation agent. Both also led the SpaceX offering. IPOX research associate Lukas Muehlbauer said the listing "will establish a benchmark for valuing leading AI labs", though he added that its success "would not necessarily benefit every AI business equally".