📢 Flash
Anthropic Is Also Frantically Snatching Hashrate: $89.54 Billion Committed, Even Old Coal-Fired Power Plants Are Not Spared
By Su Yang | Edited by Xu Qingyang
Image generated by AI; person depicted is Dario Amodei, CEO of Anthropic
AI companies are rapidly transforming into large-scale infrastructure buyers. Anthropic’s business scope has expanded beyond model training and enterprise customer acquisition to encompass capital-intensive domains such as data center construction, procurement of high-end chips, and coordination of power resources. To ensure continuous and sufficient compute supply, the company is now directly assuming core functions previously handled by cloud service providers and specialized infrastructure operators.
On November 2, local time, Anthropic filed an updated prospectus with the U.S. Securities and Exchange Commission (SEC), adding financial forecasts through Q3 2026 and detailing several long-term compute procurement and facility leasing agreements. The filing explicitly states that compute expenditure has become Anthropic’s largest and most consequential cost driver in its expansion—meaning the company no longer relies solely on cloud platforms like Amazon and Google to support the training iterations of its Claude models and meet enterprise customers’ high-frequency API usage demands.
Image generated by AI; person depicted is Dario Amodei, CEO of Anthropic
AI companies are rapidly transforming into large-scale infrastructure buyers. Anthropic’s business scope has expanded beyond model training and enterprise customer acquisition to encompass capital-intensive domains such as data center construction, procurement of high-end chips, and coordination of power resources. To ensure continuous and sufficient compute supply, the company is now directly assuming core functions previously handled by cloud service providers and specialized infrastructure operators.
On November 2, local time, Anthropic filed an updated prospectus with the U.S. Securities and Exchange Commission (SEC), adding financial forecasts through Q3 2026 and detailing several long-term compute procurement and facility leasing agreements. The filing explicitly states that compute expenditure has become Anthropic’s largest and most consequential cost driver in its expansion—meaning the company no longer relies solely on cloud platforms like Amazon and Google to support the training iterations of its Claude models and meet enterprise customers’ high-frequency API usage demands.
