Anthropic in Talks With Nscale for Second Multibillion-Dollar AI Deal

Anthropic is negotiating a second multibillion-dollar infrastructure deal with the neocloud provider Nscale, a source told CNBC. If it closes, the agreement would follow an earlier deal between the two companies for a data center on the same U.S. site.

That's the whole of what the sourcing supports at this stage — a reported negotiation, not a signed contract. But the shape of the story tells you a good deal about where the AI compute market is heading.

Who are the two companies here?

Anthropic builds the Claude family of AI models and has positioned itself as one of the main U.S. frontier labs. Training and running large models requires enormous amounts of computing power, and that compute lives in data centers packed with high-end chips.

Nscale belongs to a category the industry has come to call "neoclouds" — smaller cloud providers that rent out GPU capacity, often to AI companies, rather than offering the full general-purpose cloud platforms of Amazon, Microsoft or Google. For labs like Anthropic, neoclouds are one of several ways to lock in compute outside the big-three hyperscalers.

The takeaway: this is a frontier lab working with a specialist compute provider, not a household-name cloud giant.

What do we actually know about the deal?

Two facts, per the CNBC report:

• Anthropic and Nscale are in talks over a deal worth multiple billions of dollars.
• The two companies have already signed a similar arrangement for a data center on the same site in the United States.

The reported negotiation concerns a second deal, which suggests Anthropic wants more capacity in a location where it has already committed. No dollar figure, timeline or site name has been confirmed in the reporting so far, and neither company has publicly announced the agreement.

The takeaway: this is a follow-on commitment at an existing U.S. location, still at the negotiation stage.

Why does a second deal at the same site matter?

Companies rarely double down on a location by accident. When a lab signs one multibillion-dollar deal and then returns to the same site for more, it usually signals that the first tranche of capacity worked — and that demand for compute keeps outrunning supply.

That dynamic is the engine of the current AI infrastructure buildout. Frontier labs are competing for GPU capacity, power and land, and they're signing long-term contracts to secure years of compute in advance. Neoclouds like Nscale have grown into that gap, offering dedicated AI capacity that the hyperscalers can't always guarantee.

For Nscale, a second multibillion-dollar contract with Anthropic would be a significant validation. For Anthropic, it would mean more locked-in compute for training and serving Claude models.

The takeaway: repeat deals at the same site point to compute demand that hasn't come close to being met.

What happens next?

The talks could still fall apart or change in scope — that's the nature of unannounced negotiations sourced to a single person familiar with the matter. The signal to watch is whether Anthropic and Nscale confirm a deal publicly, and whether the dollar figure matches the multibillion scale reported by CNBC.

If it closes, it would extend a pattern that now defines the AI industry's physical layer: labs signing ever-larger, multiyear commitments to secure the compute their models depend on.