Plox
Anthropic3 mins read

Anthropic’s $45B Nscale Compute Deal Signals a Bigger Race for AI Infrastructure

Anthropic has reportedly struck a roughly $45 billion compute deal with British cloud startup Nscale as it locks in capacity ahead of a planned IPO.

The Decoder

The Decoder

Anthropic locks in 45-billion-dollar compute deal with Nscale ahead of IPO

Anthropic has struck a deal worth around 45 billion dollars with British cloud startup Nscale, Bloomberg reports.
Image credits:The Decoder

The Deal: Six Years of AI Compute Capacity

Anthropic has struck a deal worth around $45 billion with British cloud startup Nscale, Bloomberg reports. Under the six-year arrangement, the Claude maker will rent compute capacity from a data center in West Virginia. Starting late next year, Anthropic will use 460 megawatts of Nvidia’s coming Vera Rubin chip generation there, according to two people familiar with the matter.

Why It Matters Before Anthropic’s IPO

The reported agreement comes as Anthropic moves quickly to secure compute ahead of an IPO planned for this fall. The company is also locking in capacity through deals with Amazon and SpaceX, plus a commitment to lease Google’s AI chips for more than $150 billion. It has also held early talks with Meta about a possible multibillion-dollar deal.

Nscale’s Own Expansion Push

Nscale is preparing an IPO of its own and raised $2 billion in March at a valuation of $14.6 billion, led by Norwegian energy company Aker. The company says its planned 1.35-gigawatt data center in Mason County will cost $69 billion. The project is facing pushback from local residents.

Key Takeaway for AI Watchers

The report highlights how access to power, chips, and data center capacity has become central to frontier AI competition. For Anthropic, the Nscale agreement would add another major infrastructure commitment around Claude. For Nscale, the deal would tie its growth plans to one of the most visible AI companies in the market.

Discover More

    Anthropic Model Hardware Standard interface connecting AI agents with lab and factory hardware
    Anthropic MHS for AI hardware

    MHS gives AI agents a shared way to read from and control physical devices, but physical reasoning remains a key limitation.

    AnthropicAI agents
    U.S. court rules Pentagon's blacklisting of Anthropic was unlawful
    Court Faults Anthropic Blacklist

    A San Francisco federal court ruled the Pentagon’s blacklist designation of Anthropic was unlawful, though the listing remains in place pending a Washington case.

    AnthropicPentagon