Nscale finalized a $45 billion agreement to deploy Nvidia’s Vera Rubin chips specifically for Anthropic’s AI infrastructure requirements.
London-based AI infrastructure firm Nscale has secured one of the largest cloud computing agreements in history, entering a $45 billion, six-year contract with Anthropic. The deal will supply computing services powered by Nvidia’s next-generation Vera Rubin chips at Nscale’s Monarch Compute Campus in West Virginia. Announced on August 26, 2026, the agreement highlights a growing industry reliance on dedicated, large-scale infrastructure partnerships.
To contextualize the financial scale, the contract value approximates the gross domestic product of Latvia. Equally significant is the project’s energy requirements: the campus will draw approximately 460 megawatts of power, sufficient to supply roughly 345,000 American households.
Chip deployment under the agreement is scheduled to commence in late 2027. The Vera Rubin architecture serves as Nvidia’s successor to the current Blackwell and Grace platforms, engineered as a cohesive supercomputer system rather than a fragmented assembly of discrete GPUs. Designed specifically for frontier AI laboratories like Anthropic, the platform integrates seven chips functioning as a unified unit. Nvidia claims the architecture delivers higher throughput per watt and reduced costs per token compared to previous generations.
Nscale is diversifying its infrastructure pipeline beyond Anthropic. In March 2026, the company announced a separate commitment to Microsoft for up to 1.35 gigawatts of Vera Rubin NVL72 systems at the same West Virginia facility. On a broader scale, Nscale plans to deploy between 100,000 and 130,000 Vera Rubin GPUs across its European operations.
Nscale’s market position stems from its evolution from a cryptocurrency mining infrastructure spin-off into a specialized modular AI data center developer. Following substantial funding rounds and the addition of prominent board members, the company has established itself as a credible infrastructure partner for two of the technology sector’s most influential firms.
For Anthropic, developer of the Claude family of AI models, securing this capacity ahead of a reportedly imminent initial public offering carries strategic weight. Locking in $45 billion worth of compute is not merely about acquiring sufficient hardware for model training; it signals to investors that the company has secured a long-term infrastructure pipeline capable of sustaining frontier-level competition.
The Vera Rubin platform’s optimization for agentic AI workloads aligns closely with Anthropic’s trajectory. Agentic AI—where models autonomously plan and execute multi-step tasks rather than simply responding to prompts—is widely anticipated as the next major capability frontier. These workloads demand substantially more compute than standard inference, explaining Anthropic’s willingness to commit to such a massive power envelope.
When combined, Nscale’s Anthropic and Microsoft commitments suggest the Monarch Compute Campus could ultimately consume well over one gigawatt of electricity. At that scale, the facility would approach the output of a nuclear power plant, operating almost exclusively to train and run artificial intelligence models.