Anthropic is preparing to land on Wall Street as early as November, after the mid-term elections in the United States, with a possible valuation of around or above 2,000 billion dollars and puts pen to paper on its accounts, governance and above all the risks that could affect the investment.
In an official document Anthropic formally warned investors that its technology could also pose “existential risks to humanity”. This is what emerges from the information prospectus prepared by the company in view of the expected landing on the stock exchange and viewed by Financial Times. In the document, the startup that develops “Claude” linguistic models puts down in black and white some of the most extreme fears linked to the development of advanced artificial intelligence.
Emerging hazards
Almost a third of the 261 pages of the prospectus are dedicated to “risk factors”. Among these Anthropic indicates the possibility that increasingly powerful models can demonstrate unpredictable behavior, manipulate information or people and, in controlled tests conducted by the company itself, lead to behaviors such as sabotage or blackmail. The company also warns that some dangerous capabilities may only emerge as systems become more autonomous and used on a large scale.
The warnings come as investors try to place a value on one of the companies that symbolize the artificial intelligence boom. In May, Anthropic had reached an estimated valuation of around $965 billion, while second Reuters the listing could take place at a valuation exceeding 2,000 billion.
Huge costs
The numbers contained in the prospectus show both vertiginous growth and enormous costs. In 2025, Anthropic generated nearly $4.6 billion in revenue, twelve times the previous year, but lost more than $8 billion operationally. Operating expenses reached $12.65 billion, of which $7.33 billion was absorbed by computing capacity and infrastructure.
The net loss was even higher, nearly $42 billion. However, approximately 34 billion derive from an accounting charge linked to the revaluation of convertible financial instruments and do not represent money actually spent to manage the company.
In 2026, growth accelerated further: in the second quarter Anthropic reported to investors preliminary revenues of more than $11.5 billion, compared to $787 million in the same period the previous year. However, the commitments necessary to fuel the models are also growing. The prospectus highlights future obligations of approximately $518 billion related to cloud, computing capacity and infrastructure. Anthropic has also entered into billion-dollar agreements with various suppliers: the agreement with SpaceX alone provides for payments of up to 1.25 billion dollars per month for the use of data centers intended for artificial intelligence.