Anthropic의 IPO 공시가 첨단 AI 개발로부터의 실존적 위험에 대해 투자자들을 경고한다.

Anthropic's leaked IPO prospectus reveals ambitious spending plans and significant financial challenges as the company seeks to become a profitable business. The disclosure shows the AI model maker has included warnings about existential risks to humanity in materials aimed at investors—a messaging strategy that has drawn scrutiny over whether it serves primarily to highlight the claimed power of its technologies while encouraging sector regulation.

According to the leaked documents, approximately a third of Anthropic's prospectus addresses the risks AI models pose, including their potential to manipulate, blackmail, and exhibit unpredictable behaviours. The company plans to spend $518 billion on cloud computing and infrastructure obligations in coming years—more than a quarter of its stated target valuation of $2 trillion. This follows $7.33 billion in compute and infrastructure spending in the previous year.

The financial figures underscore the company's operating challenges. Anthropic reported an operating loss exceeding $8 billion last year, even as revenue increased 12-fold to approximately $4.6 billion. These figures suggest the pathway to profitability remains uncertain despite the company's market position as the maker of Claude, a model popular among AI-assisted developers.

The capital demands facing Anthropic extend across the broader AI sector. Oracle has committed to building $300 billion in datacenter capacity for OpenAI, Anthropic's primary competitor, though Oracle recently issued a force majeure notice to Blue Owl, an alternative asset management firm, citing circumstances beyond its control. The notice relates to concerns about power availability for a planned New Mexico datacenter, which Oracle maintains will proceed as scheduled.

The capital intensity of AI infrastructure raises broader economic concerns. Current estimates project $31.6 trillion in capital expenditure will be required before 2050 to meet projected AI service demands. Present investment levels now represent approximately one-third of all US economic growth, figures that have drawn comparisons to the dotcom boom. A comparable collapse in AI spending would have severe consequences for the global economy.