AI lab Anthropic has signed some $517 billion in compute capacity leases in the past 11 months.
As reported by The Information, which has compiled public announcements and its own reporting, since October last year, the company has signed on for 14.8GW of compute capacity that it will access over the next few years.
According to an unnamed source, the company had secured between 1GW and 2GW of capacity prior to that as well.
That Anthropic has been rapidly scaling up its compute capacity is not news. Towards the end of August, the company reportedly signed a $45bn compute capacity agreement with Nscale, and prior to that, it said it had signed more than a dozen letters of intent for data center leases with multiple US developers, signed a 191MW agreement with Riot Platforms for its data center in Texas, and has a $200bn agreement signed with Google for TPUs that it could deploy itself.
Beyond data center leases, the company has also turned to cloud providers for its compute capacity. This includes large cloud capacity deals with Akamai, Amazon Web Services – including a massive cluster of the cloud firm’s custom Trainium hardware, CoreWeave, and a $50bn partnership with AI cloud firm Fluidstack. Other companies it has signed agreements with include SpaceX/xAI, Google Cloud, Lambda, Volta Infra, AMD, and Microsoft.
