Anthropic has struck a roughly $45 billion deal with British cloud computing startup Nscale, securing six years of dedicated artificial intelligence compute capacity at a data center campus in West Virginia, according to a report by Bloomberg. The agreement is among the largest single compute commitments disclosed by any AI laboratory to date and underlines how fiercely frontier developers are competing to lock up scarce data center capacity.

According to two people familiar with the matter cited by Bloomberg, Anthropic will begin using 460 megawatts of capacity equipped with Nvidia's upcoming Vera Rubin chip generation at the West Virginia site late next year. The scale of the reservation places the Claude maker in the same league as the world's largest technology companies when it comes to guaranteed access to cutting-edge accelerators. For more context on this story, see our ongoing AI industry coverage.

Inside the Nscale Agreement

The deal was first reported by Bloomberg, which said the contract runs for approximately six years. Nscale, a London-headquartered neocloud operator that grew out of the cryptocurrency mining sector, has positioned itself as one of the fastest-growing suppliers of purpose-built AI infrastructure outside the traditional hyperscalers.

Three factors make the arrangement notable:

  • Long duration: Committing to six years gives Nscale bankable revenue to fund construction, while giving Anthropic predictable pricing in a volatile market.
  • Next-generation hardware: Reserving capacity for Nvidia's Vera Rubin platform means Anthropic is betting on silicon that is only just arriving in volume deployments.
  • Dedicated megawatts: Rather than renting virtual machines, Anthropic effectively controls physical capacity — 460 MW is enough power to serve hundreds of thousands of accelerator units at modern density.

The Mason County Campus

The capacity sits within Nscale's planned 1.35-gigawatt data center campus in Mason County, West Virginia, which the company says will cost approximately $69 billion to build out at full scale. The project has drawn pushback from local residents concerned about electricity demand, water usage, and industrial development in a rural region — tension that has become a recurring theme of the American data center boom.

For Nscale, the Anthropic contract provides exactly the kind of anchor tenant that large infrastructure projects require. The company raised $2 billion in March at a $14.6 billion valuation, led by Norwegian energy group Aker, and is reportedly preparing an initial public offering of its own.

Part of a Larger Compute Buying Spree

The Nscale pact does not stand alone. Bloomberg reporting assembled by The Decoder describes Anthropic rapidly locking in compute across multiple providers:

| Partner | Arrangement |
|---|---|
| Nscale | ~$45 billion, six years, 460 MW of Vera Rubin capacity in West Virginia |
| Google | Reported commitment exceeding $150 billion to lease Google's AI chips |
| Amazon | Existing large-scale compute arrangement |
| SpaceX | Reported satellite and orbitally assisted compute cooperation |

The same reporting notes that Anthropic has held early discussions with Meta about a possible multibillion-dollar arrangement as well — remarkable given that Meta competes with Anthropic for both talent and users.

What It Means for the Industry

The economics of frontier AI have forced laboratories to behave like utilities planners. Training the next generation of models requires sustained clusters measured in gigawatts, and inference demand from hundreds of millions of users compounds the pressure. Guaranteed multi-year contracts are becoming the standard instrument for hedging that risk.

Three consequences follow:

1. Revenue-backed infrastructure finance: Startups such as Nscale can raise enormous sums against contracts with investment-grade counterparties, accelerating construction timelines across the sector.
2. Concentration risk: Critics note that circular arrangements — AI labs committing billions while their partners invest back into them — spread losses widely if AI revenue growth disappoints.
3. Community resistance: Megacampus projects in places like Mason County face growing scrutiny from residents and regulators, adding permitting risk to already massive capital plans.

Looking Ahead

Neither Anthropic nor Nscale has publicly detailed the financial terms beyond the figures reported by Bloomberg, and questions remain about timelines for powering successive phases of the West Virginia campus. What is clear is that the scramble for kilowatts — not just chips — now defines competitive advantage in frontier AI.

Anthropic's approach contrasts with that of some rivals building their own data centers outright. By contracting specialist operators, it converts fixed capital expenditure into long-term operating commitments, preserving flexibility while still guaranteeing supply. Whether that balance holds as model demands escalate will shape how the next wave of foundation models gets built.

Why Vera Rubin Capacity Is the Prize

Each successive Nvidia architecture resets the economics of both training and inference, and reservations follow silicon generations rather than physical buildings alone. By tying its reservation to the Vera Rubin platform specifically, Anthropic ensures that capacity comes online with per-watt performance substantially ahead of today's installed fleets — an advantage that compounds across six years of exponential model scaling.

For Nscale, meanwhile, delivering next-generation hardware at scale would validate its pitch that specialist AI cloud operators can build faster and cheaper than incumbents. The company's rapid rise from niche operator to anchor tenant for a top-tier laboratory shows how quickly the infrastructure market has bifurcated into hyperscale generalists and AI-focused specialists competing for the same megawatts.

The Due Diligence Questions

Analysts watching this sector will press on several open points. The reported terms have not been confirmed directly by either company. Construction pace at Mason County depends on grid interconnection approvals and supply chains that stretch years into the future. And the arrangement adds to the growing stack of interlinked commitments — compute purchases here, equity stakes there — that regulators including competition authorities have begun examining for systemic risk if AI demand disappoints.

None of those caveats diminishes the headline significance: in frontier AI, the decisive currency of 2026 is not algorithms but guaranteed access to electricity-backed accelerators, and Anthropic just spent $45 billion making sure it has some.

---

Stay Ahead of AI

Get the latest AI news, analysis, and breakthroughs — all in one place.

Read more AI news →