OpenAI has abandoned plans to rent compute capacity directly from the Stargate Norway data center, handing the 230-megawatt Arctic Circle site over to Microsoft, according to reports that surfaced on April 14 and were confirmed across outlets on April 15, 2026. The move marks the third high-profile Stargate retreat in recent weeks, after OpenAI scrapped a planned expansion of its Abilene, Texas site and paused Stargate UK.
The project, originally announced in July 2025 by UK AI cloud provider Nscale alongside Norway's Aker ASA and OpenAI, was pitched as one of Europe's most ambitious AI buildouts, designed to host 100,000 Nvidia GPUs near the town of Narvik. Instead, Microsoft will expand its deployment at the campus, adding more than 30,000 Nvidia Vera Rubin GPUs on the site.
Why OpenAI is retreating
OpenAI has been publicly narrowing its infrastructure ambitions as it concentrates cash on products that more directly address the competitive threat from Anthropic. Reporting indicates the company is "pulling back on several of its side quests" to double down on coding agents and enterprise tooling, where Claude has been rapidly gaining enterprise share. The U.K. pause earlier this month was attributed to high energy costs and the regulatory environment, and the Norway exit appears to reflect similar caution about locking in long-term European leases.
That calculus contrasts sharply with the $122 billion funding round OpenAI closed just weeks ago at an $852 billion valuation. With Amazon, Nvidia and SoftBank anchoring the raise, OpenAI has the capital to pursue almost any infrastructure play it wants — the decision to shed sites suggests a deliberate pivot toward flexibility over owning the full stack.
Microsoft fills the gap
For Microsoft, the timing is opportune. Jon Tinter, Microsoft's president of business development and ventures, said in a statement that expanding the company's work with Nscale in Narvik "helps ensure Microsoft customers have access to the advanced AI infrastructure they need as demand continues to grow across Europe."
The Narvik expansion slots into Microsoft's broader Vera Rubin rollout. Microsoft has said it will be among the first cloud providers to deploy Nvidia's next-generation Vera Rubin NVL72 rack-scale systems in 2026, alongside AWS, Google Cloud and OCI, with the chips forming the backbone of future Fairwater AI superfactory sites.
What it means for Europe's AI buildout
For Nscale, the shift from OpenAI to Microsoft removes a marquee customer but swaps in arguably the most dependable hyperscaler buyer on the planet. For Narvik and Norway — where cold Arctic air and hydropower make the region one of Europe's most efficient places to run GPUs — the end state is the same: a major AI campus comes online, just under a different flag.
The larger story is a rebalancing of the Stargate program itself. Announced with great fanfare as a multi-hundred-billion-dollar OpenAI-led infrastructure initiative, Stargate is quietly being disaggregated, with Microsoft, Oracle and others absorbing pieces OpenAI no longer wants to underwrite directly. That may ultimately prove healthier for the ecosystem — but it is a notable recalibration for a company that spent 2025 positioning itself as the world's next great infrastructure builder.
