Nvidia Is Playing Matchmaker to Put Its GPUs Into Nordic Data Centers

Why Scandinavia and Iceland are attracting AI infrastructure

The Nordic region—including Norway, Sweden, Finland and Iceland—has become an attractive location for power-intensive data centers because of its combination of available land, access to electricity and cooler weather. Those factors can help operators manage the unusually high power density and heat generated by modern AI systems.

Renewable power is another part of the pitch. Stargate Norway, for example, was announced as a facility powered by renewable energy in Northern Norway. The region’s advantages do not eliminate the infrastructure challenge: grid access, construction schedules, financing and customer commitments still determine whether proposed capacity becomes operational compute.

Nvidia’s DGX-Ready program provides a formal channel

The matchmaking reports sit alongside an established Nvidia partner program. Nvidia’s DGX-Ready Colocation Data Center program certifies facilities worldwide for deploying DGX infrastructure in an environment designed for AI workloads. Nvidia describes the program as covering modern data centers, interconnectivity and advanced cooling, including liquid-cooling capabilities.

Reports about the Nordic initiative identify atNorth, Bulk Data Centers and Borealis Data Center among the operators that meet Nvidia’s regional standards. Certification is not the same as a confirmed commercial deal with every GPU owner, but it gives Nvidia a structured network of facilities that customers can consider when they need a place to deploy DGX systems.

Nvidia also has a more explicit capacity-matching marketplace. Its Compute MatchMake service allows teams to browse GPU capacity by platform, region and dates, submit workload requirements, connect with certified cloud partners and finalize a contract. That marketplace is distinct from the reported introductions between GPU holders and data-center operators, but both point toward a more organized role for Nvidia in arranging access to compute.

Projects showing how the market is taking shape

Several announced or reported projects illustrate the demand for Nordic AI capacity.

Stargate Norway targeted 100,000 Nvidia GPUs

In July 2025, Nscale, Aker and OpenAI announced Stargate Norway, an AI infrastructure project near Narvik in Northern Norway. The project was planned with an initial 230 megawatts of capacity, the possibility of adding another 290 megawatts, and a target of 100,000 Nvidia GPUs by the end of 2026. It was presented as a renewable-powered facility and OpenAI’s first European AI data-center initiative under its OpenAI for Countries program.

The announcement also highlights why compute offtakers matter. A facility can be designed and supplied with hardware, but it needs customers willing to purchase or lease the capacity for the investment to make commercial sense.

Microsoft later contracted additional capacity at the Norway site

In April 2026, Microsoft agreed to rent an additional 30,000 Nvidia Vera Rubin GPUs from Nscale at the campus in the Narvik area. The site had initially been associated with the OpenAI Stargate effort, according to reporting on the subsequent arrangement.

This change shows that the physical facility and its GPU capacity can remain commercially valuable even when the identity of the intended customer changes. It also illustrates the role of an offtaker: Microsoft’s commitment helps anchor demand for capacity being developed by Nscale and its partners.

Crusoe and atNorth expanded in Iceland

CNBC-related reporting also pointed to Crusoe’s expansion with atNorth in Iceland, using Nvidia DGX GB200 NVL72 systems, as another example of the Nordic buildout. The example connects a GPU infrastructure provider, a certified or qualified data-center environment and a location suited to large-scale compute deployment.

The available reporting does not establish that Nvidia directly financed or operated this Iceland project. It supports the narrower conclusion that Nvidia’s hardware and deployment ecosystem are being used in a region where operators are building capacity for AI workloads.

What this means for Nvidia’s business and influence

The immediate benefit of matchmaking is faster utilization. If customers can install GPUs sooner, Nvidia may be able to accelerate deployments and reduce the risk that purchased hardware remains idle while its owner waits for power or facilities. Faster deployment can also support additional demand for Nvidia’s systems, networking products and software.

The longer-term implication is greater influence over the deployment layer. Nvidia’s DGX systems, networking technologies, software environment and certified colocation partners can become part of a coordinated path from hardware allocation to functioning AI service. Nvidia may therefore influence not only which GPUs customers buy, but also where those GPUs are installed and which operators host them.

That influence is strategically useful in a market where the main constraint is increasingly the complete AI factory rather than an individual component. Power availability, cooling, buildings, networking, financing and customer demand all have to arrive together. Nvidia’s reported introductions address one part of that coordination problem, while its DGX and cloud-partner programs provide a more formal infrastructure network.

What remains unconfirmed

The evidence should be separated into two categories. Nvidia’s DGX-Ready program is documented directly by Nvidia, and the Stargate Norway project was announced by OpenAI and its partners. The details of Nvidia’s matchmaking activity, including the number of introductions, commercial terms and specific prospective offtakers, come primarily from reporting that cites unnamed sources.

The clearest conclusion is therefore not that Nvidia has become a data-center owner or operator. It is that the company is trying to coordinate more of the path between GPU supply and productive AI capacity—especially in Nordic markets where land, power and cooling have made large-scale projects more feasible. The scale of that effort, and how much revenue or control it ultimately creates for Nvidia, remains undisclosed.