August 27, 2026, (Inside AI) — Anthropic has agreed to spend $45 billion over six years to rent AI computing power from Nscale, a cloud infrastructure provider, according to a person familiar with the matter. The deal covers roughly 460 megawatts of power capacity at Nscale's data center campus in West Virginia.
The arrangement will deploy Nvidia's upcoming Vera Rubin chips to support Anthropic's computing needs. The AI startup is securing capacity ahead of expected demand for products such as its coding tool, Claude Code.
The agreement signals how quickly frontier AI labs are locking in compute. Anthropic has moved aggressively in recent months to overcome capacity constraints. In May, it agreed to rent the full computing power of SpaceX's Colossus 1 facility in Memphis, which houses more than 220,000 Nvidia processors and 300 megawatts of new capacity.
Anthropic declined to comment. The development was first reported earlier on Wednesday.
Compute Deals Reshape AI's Power Map
The $45 billion commitment dwarfs many previous AI infrastructure deals. It highlights a broader industry shift: model builders are no longer just renting cloud capacity. They are signing long-term, multi-gigawatt agreements that resemble utility contracts more than traditional cloud service agreements.
Nscale's West Virginia campus is part of a wave of data center construction in regions with abundant energy and land. The company has positioned itself as a specialist in large-scale AI clusters. Deploying Nvidia's Vera Rubin chips suggests the facility will be optimized for next-generation training and inference workloads.
For Anthropic, the deal reduces reliance on major hyperscalers. The company already has deep ties to Amazon and Google, but direct agreements with infrastructure providers give it more control over supply. That matters as demand for AI coding tools and agentic systems accelerates.
Capacity Fears Drive Long-Term Bets
Industry analysts have warned that AI compute demand could outstrip supply through the end of the decade. Chip production constraints, power availability, and cooling requirements are all bottlenecks. Anthropic's deal with Nscale is a hedge against those risks.
The six-year term is notable. Most cloud contracts run one to three years. A six-year commitment suggests Anthropic expects sustained demand for its models and tools. It also gives Nscale the revenue certainty needed to finance expensive data center buildouts.
The West Virginia campus will likely support both training and inference. Claude Code, Anthropic's AI coding assistant, has driven significant usage growth. Coding workloads are compute-intensive and require low latency, making dedicated capacity valuable.
Anthropic's earlier deal with SpaceX's Colossus 1 facility in Memphis followed a similar pattern. That facility houses more than 220,000 Nvidia processors. The combined capacity from both deals positions Anthropic among the best-resourced AI labs in the world.
The financial scale is also striking. $45 billion over six years averages $7.5 billion annually. That is a substantial portion of Anthropic's projected revenue. It reflects confidence that AI coding tools and enterprise AI services will generate enough demand to justify the expense.
The deal does not include equity or ownership of the data center. It is a pure compute rental agreement. That structure lets Anthropic scale without the capital burden of building its own facilities. Nscale bears the construction and operational risk.
The broader AI infrastructure market is booming. Competitors such as CoreWeave, Crusoe, and Lambda have raised billions to build AI-specific data centers. Nscale's deal with Anthropic is one of the largest single customer agreements in that emerging sector.
Power availability remains a key constraint. West Virginia has attracted data center investment due to its energy infrastructure and relatively low costs. The 460 megawatts allocated to Anthropic is enough to power hundreds of thousands of homes.
Anthropic's aggressive capacity building also reflects competitive pressure. Rivals such as OpenAI and Google DeepMind are expanding their own compute footprints. The race to secure Nvidia's latest chips is intense, and long-term contracts are becoming a strategic weapon.
The Vera Rubin chips are expected to deliver significant performance gains over current-generation hardware. Deploying them at scale will require advanced cooling and power systems. Nscale's West Virginia campus is designed for exactly that kind of high-density AI workload.
As AI models grow larger and more capable, the cost of training and serving them rises. Anthropic's deal with Nscale is a clear signal that the company intends to compete at the frontier for years to come.
