OpenAI’s ‘Stargate Norway’ aims to build a sovereign AI compute hub in Europe

Narvik, Norway, August 18, 2025 — OpenAI, Nscale, and Aker plan to develop a large renewable-powered AI data center in Northern Norway to expand sovereign compute capacity for Europe. The project, called Stargate Norway, is part of OpenAI’s “OpenAI for Countries” program and is designed to serve researchers, startups, and enterprises with high-performance infrastructure.

Located near Kvandal outside Narvik, the site targets 230 megawatts in its first phase with a roadmap to add 290 megawatts. The partners say the facility will host about 100,000 Nvidia GPUs by the end of 2026. Design choices include closed-loop direct-to-chip liquid cooling and a fully renewable energy model supplied by local hydropower. Plans also include waste-heat reuse to support nearby low-carbon businesses.

Nscale will lead development. Operations will run through a joint venture with Aker following its merger activity involving Aker Horizons. Each company expects to contribute at least 250 million dollars to co-fund a first-phase budget of about 1 billion dollars. Nvidia is a strategic backer on the hardware side. OpenAI indicates it may increase use of the facility as new phases come online.

Backers frame the project as both an industrial investment and a sovereignty play. Northern Norway offers cool ambient temperatures, grid access, and comparatively low energy costs due to hydropower. Aker’s leadership links the initiative to Norway’s history of converting clean energy into export-grade industry, while Nscale calls the deployment a regional milestone that could strengthen Europe’s position in advanced computing.

OpenAI positions Stargate Norway as a response to rising demand for compute across the continent. The company says additional capacity can support work in multimodal AI, agentic workflows, and scientific research while aligning with European regulatory preferences for expandable and locally governed infrastructure. Priority access for local researchers and startups is part of the stated plan, alongside surplus capacity available to clients across the Nordics, the United Kingdom, and Northern Europe.

Key execution variables include the pace of grid approvals, delivery timelines for Nvidia hardware, and the ability to maintain low latency and stable uptime at scale. Governance will also be in focus. Buyers are likely to scrutinize data handling, access controls, and the practicalities of allocating scarce GPU capacity among public and private users. The partners say sustainability features and energy efficiency will be central to operations as build-out proceeds.

What to watch

• Final timeline for the first 230 megawatts and the follow-on 290 megawatts.

• Shipment cadence and configurations for the planned 100,000 GPUs.

• Details on access policies, pricing, and governance for European public and private users.