Anthropic has finalized a contract to lease approximately $45 billion worth of artificial intelligence computing resources from Nscale, a UK-based AI infrastructure firm, according to an insider.
Founded in 2024, Nscale has already established partnerships with major players, including Microsoft, and will provide Anthropic with computing power through Nvidia’s advanced Vera Rubin chip system.
The Vera Rubin system utilizes a combination of six different chips operating in unison, representing the forefront of chip technology. The computing power is anticipated to start fueling the AI lab’s services by late 2027, as mentioned by the source.
This agreement, first highlighted by another news outlet, marks the latest in a series of computing partnerships for Anthropic. This six-year deal involves computing resources from Nscale’s main data center located in West Virginia.
In recent months, Anthropic has significantly bolstered its computing capacity to enhance its competitive edge against rivals, particularly OpenAI.
Earlier this month, Anthropic secured a $10 billion agreement with the AI cloud startup Volta, which was launched in January, for a six-year supply of cloud computing services from a data center in Norway. Additionally, in July, Anthropic finalized a $5 billion computing agreement with AMD.
Prior to that, in May, Anthropic disclosed a major computing contract with SpaceX, run by Elon Musk, positioning him as an unexpected ally against OpenAI CEO Sam Altman. This agreement allows Anthropic to draw computing capacity from two SpaceX data centers and provides an estimated $1.25 billion worth of capacity monthly.
In April, Anthropic also expanded its collaboration with Amazon, gaining access to an extra 5 gigawatts of computing power. That same month, the firm enhanced its partnerships with both Google and Broadcom, further increasing its computing capabilities.
Anthropic is not alone in chasing advanced computing power; many other industry giants, including Google, OpenAI, and Meta, are also striving to secure increased compute resources.
