AI giant Anthropic submitted an S-1 prospectus last week and is expected to complete a record IPO this month, stirring up Wall Street sentiment. The company updated its prospectus on Monday to reveal impressive growth data, but the risks on the expenditure side are also prominent. In the past year, Anthropic signed a large number of computing power procurement contracts to lock in computing power resource training and run large AI models, betting that surging demand can cover huge investments. According to the updated S-1, Anthropic's third-quarter revenue was 17 billion US dollars, 12 times the year-on-year increase, up 47% from the second quarter; however, the net loss was 23 billion US dollars, and the month-on-month growth rate declined from 144% in the previous quarter. As of 2045, Anthropic's total procurement commitments for data centers, chips, and computing power reached 895.4 billion US dollars; within the next 12 months, nearly 70 billion US dollars of contract payments will begin. Two people familiar with the matter revealed that by the end of this year, the company's usable computing power will reach 3.5 gigawatts, which is enough to supply all household electricity in Chicago; next year, the target computing power will double. Anthropic is even deeply involved in data center design, recruiting infrastructure and supply chain experts, and procuring hardware such as cooling systems from contractors. Its number one competitor, OpenAI, is also investing wildly in computing power. People familiar with the matter revealed that OpenAI expects to invest a total of about 750 billion US dollars in computing power by 2030.

AI giant Anthropic submitted an S-1 prospectus last week and is expected to complete a record IPO this month, stirring up Wall Street sentiment. The company updated its prospectus on Monday to reveal impressive growth data, but the risks on the expenditure side are also prominent. In the past year, Anthropic signed a large number of computing power procurement contracts to lock in computing power resource training and run large AI models, betting that surging demand can cover huge investments. According to the updated S-1, Anthropic's third-quarter revenue was 17 billion US dollars, 12 times the year-on-year increase, up 47% from the second quarter; however, the net loss was 23 billion US dollars, and the month-on-month growth rate declined from 144% in the previous quarter. As of 2045, Anthropic's total procurement commitments for data centers, chips, and computing power reached 895.4 billion US dollars; within the next 12 months, nearly 70 billion US dollars of contract payments will begin. Two people familiar with the matter revealed that by the end of this year, the company's usable computing power will reach 3.5 gigawatts, which is enough to supply all household electricity in Chicago; next year, the target computing power will double. Anthropic is even deeply involved in data center design, recruiting infrastructure and supply chain experts, and procuring hardware such as cooling systems from contractors. Its number one competitor, OpenAI, is also investing wildly in computing power. People familiar with the matter revealed that OpenAI expects to invest a total of about 750 billion US dollars in computing power by 2030.