Anthropic and Nscale Pursue Second $45 Billion Compute Deal

Anthropic and Nscale are eyeing a second compute deal worth $45 billion, according to a report from The Tech Buzz. The figure matches the scale of the AI industry's largest infrastructure commitments and signals that neither company considers its current capacity plans sufficient.

The report gives no timeline, no contract structure and no named suppliers for the potential agreement. What it does establish is intent: Anthropic, the maker of the Claude model family, and Nscale, the GPU cloud provider, are in discussions about a follow-on arrangement on top of an existing compute relationship.

What does a second $45B deal mean?

A deal of this size would rank among the biggest compute contracts in the AI sector. It would also deepen Anthropic's dependence on Nscale, or vice versa, depending on how the capacity is structured and paid for.

The framing matters here. The Tech Buzz describes the talks as eyeing a second deal, which implies a first arrangement already exists between the two companies. The outlet does not specify the value, duration or GPU volumes of that initial agreement.

Why is Anthropic shopping for more compute?

Training frontier models consumes capacity at a rate that outpaces most forecasts. Anthropic competes with OpenAI, Google DeepMind and Meta, all of which have locked in multi-year, multi-billion-dollar infrastructure commitments with cloud and chip partners.

A second $45 billion arrangement would give Anthropic a secured runway for training and inference. It would also give Nscale, a newer entrant in the GPU cloud market, a marquee anchor customer with contracted revenue stretching years ahead.

Neither company has confirmed the talks. The Tech Buzz cites no executives by name and includes no quotes from either side.

What happens next?

Expect confirmation or denial to surface through official announcements rather than further leaks, given the size of the figure involved. If the deal closes at or near $45 billion, it would reset expectations for what second-tier AI labs must spend to stay competitive in compute.

By Amara Osei, Business Bearings