AI's Fine Print

Buy the Stock, Fear the Overlord

Most IPO filings warn investors about competition, regulation, and changing customer preferences. Anthropic's is slightly more dramatic: its technology could pose "existential risks to humanity."

The company’s prospectus was leaked to the press on Monday, giving a sneak peek ahead of what could be the world’s largest initial public offering (IPO). Prospectus is a document that reveals financials ahead of a listing.

The maker of Claude, one of the world's leading AI models, reportedly devoted almost a third of its IPO prospectus to risks. The filing warns that increasingly advanced AI systems could manipulate information, resist shutdown, and conceal intentions. Investors are expected to value Anthropic at more than $2 trillion if the IPO goes ahead in November.

Growing at Full Throttle

Anthropic's prospectus pulled back the curtain on the economics of the AI boom.

• Revenue surged 12-fold to nearly $4.6B in 2025
• Operating loss exceeded $8B in the same period
• Operating expenses hit $12.7B
• $7.3B went toward computing and infrastructure alone
• It expects to spend over $500B in that category in the coming years, roughly the GDP of Denmark or Vietnam
• About half of its customer sales came through just two cloud partners: Amazon and Google

Those figures may sound worrying but Anthropic’s growth has continued at breakneck speed this year. The revenue in the second quarter alone was $11.5 billion, and it is expected to post its second consecutive quarter of operating profit (on adjusted basis).

When AI Meets Politics

Concerns about advanced AI have moved from research labs to the center of public debate. US President Donald Trump held a meeting with AI leaders — including Anthropic chief Dario Amodei, OpenAI's Greg Brockman, xAI's Elon Musk, Meta's Mark Zuckerberg, and Nvidia's Jensen Huang — on Tuesday to discuss the future of the technology. The meeting came after a series of reports about AI systems going rogue, including hacking into Australian healthcare statistics portal.

Leaders agreed on a “voluntary” accord on AI safety. Trump later praised “tremendous self-regulation” by the companies.

Anthropic has become one of the loudest voices warning about the dangers. Amodei recently described AI as one of the world's most important security challenges and has urged his own industry to slow down.

Who Will Hit the Brakes?

Around the time when Anthropic's prospectus surfaced, OpenAI froze plans to release new model GPT-6.1 Astra over safety concerns. The company said it fell short in areas such as staying within authorized boundaries and communicating clearly with users.

Yet OpenAI, Anthropic, Google, Meta, xAI (owned by SpaceX), and increasingly capable Chinese rivals like Moonshot AI, Alibaba, and DeepSeek are all racing to build more powerful systems.

OpenAI is preparing its own trillion-dollar IPO, and tech giants are pouring hundreds of billions into the sector. A sector-wide slowdown seems unlikely without regulators stepping in.