Why an AI Company Just Agreed to Spend $45 Billion on Computing Power
Anthropic, the company behind the Claude AI assistant, has signed a deal to rent roughly $45 billion worth of computing power from Nscale, a British company founded in 2024. The power is expected to start running Anthropic's services in late 2027. First reported by Bloomberg, the six-year agreement covers about 460 megawatts of capacity at Nscale's flagship data center in West Virginia. TechCrunch CNBC
A data center is a large building filled with powerful computers that run software. AI models like Claude need enormous amounts of computing power to generate responses, and that power comes from specialized chips inside these facilities. Nscale will provide Anthropic with computing power using Nvidia's Vera Rubin chips, a new generation of hardware that follows the current Blackwell line.
The West Virginia site is Nscale's flagship facility in the US. The company also operates in Europe, including a partnership with Nordkraft for data center operations in Narvik, Norway. Nscale has previously worked with companies including Microsoft and has been building out its AI cloud platform, partly through its acquisition of a company called Anyscale. Nscale also secured a $900 million credit line and gave $500,000 to support science and technology education in Mason County, where the West Virginia data center is located. TechCrunch Nscale
The Nscale deal is the largest in a series of agreements Anthropic has made over the past five months to lock up computing power. In April 2026, Anthropic expanded its partnership with Amazon to access an additional 5 gigawatts of computing capacity and separately broadened its relationship with Google and Broadcom. In May, the company struck a deal with SpaceX to use two of its data centers for $1.25 billion worth of capacity each month. In July, Anthropic signed a $5 billion deal with AMD. Earlier in August, the company signed a $10 billion, six-year agreement with an AI cloud startup called Volta, using a data center in Norway. TechCrunch
Together, these agreements total well over $60 billion in committed spending across at least six different partners. The combined power footprint reaches into the multiple-gigawatt range. For context, a single gigawatt can power roughly 750,000 homes.
The buying spree lines up with Anthropic's product launches. The company introduced Claude Sonnet 5 on June 30, 2026, followed by Claude Opus 5 on July 24, 2026. Both are designed for AI agents, programs that can carry out multi-step tasks with limited human supervision, with improvements in coding and professional work. Anthropic also published "The Making of Claude Code" on July 6, 2026, describing how its coding tool grew from an internal command-line utility into a full coding assistant. On August 14, the company shared details of its text watermarking method for Claude, and on August 7 it updated Claude Fable 5's biology safety filters to reduce false alarms. The computing power being secured now, much of it arriving in late 2027 and beyond, is the foundation for the next generation of these tools. Anthropic
Nscale's part in this deal is striking on its own. A company founded in 2024 has landed a $45 billion, six-year contract with one of the leading AI labs. That is possible because supply is tight: Nvidia's new Vera Rubin chips are not yet shipping in large numbers, and the big cloud providers like Amazon and Google have their capacity booked years ahead. Independent operators like Nscale that can secure land, power, and chip allocations are stepping into the gap. Nscale's acquisition of Anyscale and its $900 million credit line are the kinds of moves that make a deal of this size credible, even for a two-year-old company.
The geographic spread of Anthropic's computing power also matters. West Virginia, Norway, SpaceX's facilities, Amazon's data centers, and Google's infrastructure span different countries, power grids, and chip types. Anthropic published a report called "Build AI in America" in July 2025, arguing that the US must invest heavily in computing power and electricity to lead in AI globally. The West Virginia deal puts that argument into practice on American soil, while the Norway agreements spread risk across European power markets. Anthropic
There is a real gap, though, between signing a deal and actually getting the power. The Nscale capacity does not come online until late 2027, the Volta deal runs for six years, and much of the SpaceX and Amazon capacity is presumably ramping up on similar timelines. Anthropic is committing enormous sums of money based on a future demand it can estimate but cannot guarantee. If AI agents take off the way Anthropic's strategy suggests, the computing power secured in 2026 will be what limits or drives revenue in 2027 and 2028. If adoption falls short, the fixed costs of these long-term agreements become a heavy burden.
In my view, this pattern will be familiar to anyone who watched the cloud computing buildout of the 2010s, but the financial stakes are far higher. Amazon, Google, and Microsoft committed tens of billions to data center construction over a decade ago, betting that businesses would move their software to the cloud. They were right, and the infrastructure they built became a lasting advantage. Anthropic is making a similar bet that demand for AI will grow fast enough to fill all this capacity. The difference is that cloud infrastructure was general-purpose and could handle many kinds of software. AI computing power is tied to specific chip designs and model approaches that are still changing rapidly. A data center built for today's AI models might need significant rework for whatever comes next.
On the other side of the ledger, Anthropic has been expanding its policy and regulatory capacity alongside its computing power. The company announced that Mariano-Florentino (Tino) Cuéllar would join as Chief Global Affairs Officer on August 4, 2026, published its position on open-weights models (AI models whose parameters are publicly available) on July 27, expanded its partnership with Cognizant to bring Claude to enterprise clients on the same date, and announced funding for better evaluations of AI's impact on wellbeing on August 25. The computing deals and the policy moves are not unrelated. A company committing this much money and power needs the regulatory and institutional footing to match. Anthropic
What this makes possible, if the demand materializes, is straightforward: Anthropic positions itself to run AI models and agents at a scale that few competitors can match, with enough variety in its computing base to avoid depending on any single supplier. Whether the assumptions about demand for AI hold true is the question that will determine whether $60 billion in committed spending was far-sighted or premature.
