In its IPO filing, Anthropic said its AI models could pose serious risks to humanity in the future. Reuters and the Financial Times reported this after reviewing the document.
The filing has not yet been made public. According to Reuters, 80 of its 261 pages are dedicated to potential risks facing the company, while 48 pages describe its business.
The document says AI models could potentially take actions to preserve their own operation. For example, a model could resist being shut down, hide or manipulate information, or even take actions resembling blackmail.
In 2025, Anthropic’s revenue increased 12-fold to nearly $4.6 billion. At the same time, the company recorded an operating loss of $8.06 billion, compared with $2.98 billion in 2024.
Its net loss was close to $42 billion. Around $34 billion of that was related to accounting expenses. This was not a cash outflow, but resulted from an increase in the value of financial instruments that could be converted into shares.
Anthropic’s total operating expenses reached $12.65 billion. Of this, $7.33 billion went toward computing power and infrastructure needed to run its AI models. That was nearly three times more than in 2024.
At the end of 2025, Anthropic held $20.28 billion in cash and short-term investments. However, nearly a quarter of the company’s revenue came from just two customers. Most major customers are not tied to Anthropic through long-term contracts, meaning they could reduce their spending on AI services in the future.
Anthropic plans to spend around $518 billion on cloud services, computing power and infrastructure over the coming years. A large part of these costs consists of commitments that the company will have to pay regardless of whether it fully uses the capacity or not.
Anthropic expects its valuation to exceed $2 trillion at the time of its IPO. That would be more than twice the $965 billion valuation estimated in May.
