Anthropic Before IPO Reveals Losses and Risks of Its Own AI
Developer of Claude Anthropic, ahead of its upcoming stock market debut, for the first time showed investors in detail the financial side of the company's rapid growth and simultaneously warned separately about the risks that increasingly powerful artificial intelligence models may pose.
In the IPO prospectus, which the Financial Times has reviewed, the company reported that last year it generated nearly $4.6bn in revenue - about 12 times more than the year before. At the same time, the operating loss exceeded $8bn, and total operating expenses approached $13bn due to heavy spending on training and running models.
The main long-term financial burden remains computing power. Anthropic stated future commitments for cloud infrastructure, computing and related contracts of $518bn. The company purchases resources from several major partners and continues to build out infrastructure for Claude.
At the same time, the business is scaling quickly. According to FT, in the second quarter of 2026, Anthropic's revenue already stood at $11.5bn, and the company was approaching its second consecutive quarter of adjusted operating profit. Bloomberg earlier reported that Anthropic is actively preparing for its market debut and has already enlisted banks to arrange the IPO and a multibillion-dollar credit facility.
A separate large section of the prospectus is dedicated to technology risks. Almost a third of the document describing the business and IPO is devoted to risk factors. Anthropic warns investors that more advanced systems may demonstrate unpredictable behaviour - including concealing information, manipulating it, or attempting to avoid shutdown.
The company also highlights the difficulty of testing such models: a system may change its behaviour if it understands it is being tested. This poses a challenge for safety evaluation even before releasing new models.
Such caveats are not a sharp change in Anthropic's position. In its current Responsible Scaling Policy, the company already envisages strengthening protective measures as model capabilities grow, regular assessments of catastrophic risks, and the possibility of limiting scaling if safety systems cannot keep pace with technology development.
The prospectus also revealed another risk for future shareholders - high customer concentration. According to FT, almost a quarter of Anthropic's last year's revenue came from just two customers. For a company aspiring to one of the largest tech IPOs, this means investors will have to assess not only the growth rate but also its costs and sustainability.
Source: Financial Times, Anthropic, Bloomberg
