Key Points:
- Artificial intelligence lab Anthropic signed a $45 billion six-year agreement to lease computing power from cloud provider Nscale.
- The contract secures approximately 460 megawatts of dedicated capacity at Nscale’s flagship Monarch data center campus in West Virginia.
- The facility will deploy Nvidia’s next-generation Vera Rubin chips, with the first hardware phase coming online late next year.
- The full Monarch computing campus will scale to 1.35 gigawatts, with total development costs estimated at $71 billion.
Artificial intelligence developer Anthropic is executing one of the largest infrastructure commitments in technology history. The creator of the Claude foundation model family reached a definitive agreement to spend $45 billion over six years to lease artificial intelligence cloud computing power from specialized infrastructure provider Nscale. The multi-billion-dollar deal secures dedicated high-performance computing capacity at Nscale’s flagship Monarch campus in West Virginia as the enterprise prepares for a high-profile initial public offering.
The six-year commercial contract secures approximately 460 megawatts of dedicated electrical power capacity across the first of three planned buildings at the West Virginia site. To put that scale into perspective, 460 megawatts provides enough electrical energy to supply roughly 345,000 American homes simultaneously. Securing large blocks of energized capacity allows the artificial intelligence laboratory to guarantee server availability for training next-generation frontier models and scaling commercial enterprise workloads.
Under the agreement, Nscale will construct and operate specialized data center infrastructure powered by Nvidia’s cutting-edge Vera Rubin processor architecture. The next-generation silicon chips are scheduled to begin coming online late next year, with additional server clusters across subsequent campus buildings expected to enter service throughout 2028. Deploying advanced liquid-cooled racks ensures maximum computing throughput for complex reasoning and agentic software tasks.
The West Virginia deployment represents the cornerstone of a massive computing development. Once fully completed, the Monarch campus is projected to reach approximately 1.35 gigawatts of total capacity, supported by on-site power generation and dedicated high-voltage substation connections. Constructing the entire multi-building complex and its dedicated power infrastructure is estimated to cost roughly $71 billion, with approximately $47 billion dedicated directly to purchasing specialized artificial intelligence processors and server hardware.
The transaction highlights the rapid reshuffling of computing capacity across the technology sector. Anthropic secured the West Virginia capacity after taking over project allocations previously adjusted by other technology conglomerates. In an environment where obtaining multi-megawatt utility connections faces four-year waiting lists in traditional hubs like Northern Virginia, acquiring shovel-ready, energized campuses provides an immediate strategic advantage.
The $45 billion lease illustrates how dramatically the economics of artificial intelligence have evolved. Building leading machine learning models is no longer just about hiring elite software researchers; it requires managing massive physical utilities, securing long-term power purchase agreements, and leasing specialized real estate. Artificial intelligence developers operate more like industrial utility operators, requiring tens of billions of dollars in capital to keep models running 24 hours a day.
The Nscale partnership joins an aggressive series of infrastructure deals orchestrated by Anthropic over recent quarters. The company previously arranged a $50 billion computing program with Fluidstack, partnered on a $60 billion to $100 billion custom chip financing facility with Broadcom, and maintains multi-billion-dollar cloud commitments across major hyperscalers. Diversifying computing partners ensures that the company avoids reliance on a single cloud platform.
Securing hundreds of megawatts of guaranteed compute capacity arrives at a critical juncture as corporate leadership conducts preliminary investor meetings ahead of a public stock market debut. With annualized revenue run-rates surging past $65 billion, institutional backers are modeling a potential public valuation approaching $2 trillion. Demonstrating locked-in computing infrastructure through 2030 reassures prospective shareholders that compute bottlenecks will not constrain future revenue expansion.
As the race to build advanced artificial intelligence models accelerates, access to raw electrical power and silicon hardware has become the ultimate currency of the digital economy. By committing $45 billion to the Monarch campus in West Virginia, Anthropic is securing the physical foundation required to train its most powerful reasoning models. The landmark agreement marks a new chapter in technology infrastructure, proving that the future of software intelligence is built on gigawatt-scale physical engineering.
