On August 26, 2026, Anthropic signed a roughly $45 billion deal with UK infrastructure firm Nscale to lease 460 megawatts of compute using Nvidia's Vera Rubin chips, in a six-year contract that anchors Nscale's own IPO and lands weeks before Anthropic pursues its own public offering.
According to Bloomberg and CNBC, Anthropic will lease compute capacity from a data center Nscale is building in West Virginia, with 460 megawatts of power dedicated almost entirely to Anthropic's workloads. The contract runs six years and is built on Vera Rubin chips, Nvidia's next generation of silicon for large-scale training and inference. The facility is expected to come online by late 2027.
Nscale is one of the so-called "neoclouds": infrastructure companies specialized almost exclusively in building and operating data centers for AI workloads, unlike traditional hyperscalers (AWS, Azure, Google Cloud) that split capacity across multiple business lines. According to TechTimes, this deal comes after Microsoft pulled out of a similar negotiation with Nscale, and it serves as a commercial anchor for the company's own IPO. For Anthropic, locking in six years of compute is part of a broader race to secure energy and silicon capacity before frontier AI demand keeps outpacing supply.
What's notable about this contract isn't just its size, but what it reveals about where the real limit on scaling frontier models actually is: energy. A deal structured in megawatts, not just GPU units, confirms that securing dedicated power supply has become just as critical as securing the silicon itself, and that AI labs are willing to commit to years of capacity before the infrastructure physically exists.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel