Anthropic's $45 Billion AI Computing Deal with Nscale: What It Means for Aker and the AI Infrastructure Race
Anthropic PBC has agreed to pay $45 billion to rent AI computing power from Nscale's Monarch Compute Campus in Mason County, West Virginia, according to reporting published August 26, 2026 (Bloomberg). The following morning, Aker ASA confirmed that its portfolio company Nscale had secured the contract to supply AI compute capacity at the Monarch site, and Aker shares climbed on the news (MarketWatch).
The structure is a rental arrangement, not a capital purchase. Anthropic is paying for access to computing capacity at a facility Nscale is developing, rather than buying the physical servers and chips outright. Monarch is described by Nscale as a major AI infrastructure development in Mason County (Nscale). The specific terms — duration, GPU count, power envelope, ramp schedule — have not been disclosed in the available reporting. What is confirmed is the aggregate commitment: $45 billion in compute rental, directed at a single West Virginia campus.
Aker ASA, the Norwegian industrial holding company, is Nscale's parent. The share price reaction on August 27 reflects the market's initial read on what a $45 billion contracted revenue stream means for Aker's portfolio valuation. The magnitude is large enough to move a parent company's stock on announcement, which tells you the market had not fully priced this into Aker prior to disclosure.
Nscale's broader infrastructure footprint provides context for how Monarch fits into the company's strategy. In July 2025, Nscale announced Stargate Norway alongside Aker ASA and OpenAI — a planned 100,000 NVIDIA GPU AI gigafactory in Northern Norway, powered by renewable energy (Nscale). In September 2025, Nscale and Aker ASA signed an agreement with Microsoft to deliver scalable AI compute capacity powered entirely by renewable energy, with a start date in 2026 (LinkedIn). These are separate arrangements from the Anthropic contract but signal a pattern: Nscale is building relationships with multiple frontier AI labs and hyperscalers (the massive cloud providers like Amazon, Google, and Microsoft), financing gigawatt-scale infrastructure, and siting it where renewable power is available.
The Anthropic deal follows that template. West Virginia offers land, grid access, and power costs that coastal data-center markets increasingly cannot accommodate at scale. The Monarch campus, if delivered on the timeline the contract implies, becomes one of the larger dedicated AI compute sites in the United States.
The broader context here is that the $45 billion figure warrants real scrutiny. For comparison, that sum exceeds the market capitalization (the total value of all a company's publicly traded shares) of many S&P 500 companies. It is a multi-year commitment, and the compute capacity does not exist yet at the scale implied. The risk cuts both ways. If Anthropic's computing demand materializes as projected, Nscale and Aker capture a revenue stream that fundamentally reshapes the latter's portfolio economics. If Anthropic's demand softens — whether because model training becomes more efficient, because workloads shift to other providers, or because Anthropic's competitive position in the AI model market changes — the contract's enforceability and Nscale's financing assumptions become the binding constraint.
The deal also signals where the real bottleneck in AI infrastructure sits. Frontier AI labs are no longer competing primarily on algorithms or talent. They are competing on access to power, land, and the chips that consume both. A $45 billion rental commitment is effectively Anthropic paying a premium to lock in capacity years before it comes online, accepting the risk that Nscale can actually build and operate the Monarch campus at the scale promised.
For Aker shareholders, the central question is how much of the $45 billion flows through to Nscale's enterprise value versus being absorbed by construction costs, power purchase agreements, GPU procurement, and operating overhead. The gross contract number is not gross margin — the revenue a company keeps after the direct costs of delivering the service. The share price reaction on August 27 reflects optimism; the durability of that move will depend on disclosure around margins, phasing, and construction milestones that has not yet been provided.
For the broader AI infrastructure market, the deal reinforces a trend that has been building since 2024: compute supply contracts at the tens-of-billions level are becoming the norm for frontier labs seeking to secure multi-year training and inference capacity. Inference refers to running trained AI models to produce outputs, as opposed to training them in the first place. The Microsoft-Nscale and OpenAI-Nscale arrangements preceded this announcement. Anthropic's entry into that set of counterparties suggests the major labs view Nscale's portfolio of sites — Monarch in West Virginia, Stargate in Norway, the Microsoft-linked capacity — as a credible addition to the supply stack currently dominated by the hyperscalers themselves.
What remains unknown is the delivery timeline. Nscale's Monarch campus is a development project, not an operating facility at the contracted scale. Renewable power commitments and site readiness in rural West Virginia carry execution risk that no press release can eliminate. The $45 billion is an agreement to pay for compute that still needs to be built, energized, and operated at a performance level sufficient for Anthropic's workloads. The market's initial enthusiasm is rational. Whether it persists will turn on Nscale's ability to deliver.
