Nvidia’s New AI Infrastructure Role: Connecting GPU Owners With Nordic Data Centers

The reported activity does not establish that Nvidia owns or operates the facilities involved. It points instead to a more networked role: Nvidia can use its relationships with chip buyers, cloud companies and infrastructure developers to steer demand toward sites that are technically prepared for its systems.

How the DGX-Ready program fits in

Nvidia’s DGX-Ready Colocation Data Center program provides a formal layer for this approach. Reported requirements include support for liquid cooling, validated infrastructure components and the deployment support needed for high-density Nvidia systems. Nordic operators identified as meeting the program’s requirements include atNorth, Bulk Data Centers and Borealis Data Center.

Certification matters because a data center with empty floor space is not automatically suitable for advanced AI clusters. Operators must be able to handle concentrated power loads, thermal output and the installation requirements of Nvidia’s integrated systems.

Nvidia has also introduced Compute MatchMake, a marketplace where customers can browse GPU capacity from certified Nvidia Cloud Partners, request specific capacity, negotiate terms and complete a contract. That marketplace makes the company’s coordinating role more explicit, although it is distinct from the reported introductions between GPU owners and Nordic colocation providers.

Why Scandinavia and Iceland are attracting AI infrastructure

The Nordic region is appealing to data-center developers because projects can combine access to renewable electricity, available land and a cool climate. Borealis, for example, says its Icelandic and Nordic facilities use hydro and geothermal power, while the local climate supports free-air cooling.

These advantages are especially relevant for dense GPU deployments. Cooling is a major design consideration when large numbers of accelerators operate together, so a site that can use natural cooling conditions may reduce the complexity and energy needs of facility operations. The opportunity is constrained, however, by the availability and timing of grid connections; reporting cited about 2.3 gigawatts of Norwegian data-center capacity waiting for future grid access.

The result is a region with attractive physical conditions but intense competition for the infrastructure needed to convert those conditions into live AI compute.

Projects showing the scale of the opportunity

Several Nordic projects illustrate why Nvidia has an incentive to connect hardware holders with operators.

Stargate Norway: a 100,000-GPU target

Nscale, Aker and OpenAI announced Stargate Norway in July 2025 as a renewable-powered AI facility near Narvik. The project was planned around 230 megawatts of initial capacity, with ambitions to add another 290 megawatts, and targeted 100,000 Nvidia GPUs by the end of 2026.

The project’s commercial picture later changed. Subsequent reporting said Microsoft had contracted the full capacity of the Kvandal site and that Nscale planned to deploy more than 30,000 Nvidia Rubin GPUs at its Narvik campus for Microsoft from 2027. That development should be treated as a reported shift in customers and arrangements—not as confirmation that the original Stargate plan will proceed unchanged.

Crusoe and atNorth expand in Iceland

Crusoe and atNorth announced an additional 24 megawatts at atNorth’s ICE02 facility in Keflavík, building on an earlier 33-megawatt agreement and bringing the announced Crusoe footprint at the site to 57 megawatts. The deployment includes Nvidia GB200 NVL72 systems and Blackwell GPUs, including liquid-cooled hardware.

This is an example of the infrastructure model Nvidia is trying to accelerate: an AI cloud provider leases capacity from a specialist data-center operator, then makes Nvidia systems available through a cloud platform.

Nscale’s Iceland deployment

Nscale also signed a 15-megawatt lease at Verne’s Iceland campus for a planned deployment of 4,600 Nvidia Blackwell Ultra GPUs through 2026. The facility is described as renewable-powered, uses natural free cooling and is Nvidia DGX authorized.

Together, these projects show that the Nordic opportunity is not limited to one flagship facility. It includes a network of colocation sites, AI cloud providers and large customers assembling capacity across Norway and Iceland.

What this means for Nvidia’s broader strategy

Nvidia’s matchmaking activity is best understood as ecosystem coordination rather than a simple move into property ownership. The company’s leverage comes from several connected positions: it supplies the dominant hardware platform described in these projects, provides software and systems such as DGX, validates facilities, works with cloud partners and maintains relationships with major AI customers.

Its Compute MatchMake marketplace adds a more standardized way to reserve Nvidia GPU capacity without owning a data center. Customers can filter capacity by GPU platform, region and dates, then move through requests, partner matching and contracting.

The Nordic introductions show the same logic operating through relationships rather than only through a public marketplace. Nvidia benefits when an infrastructure builder can find a GPU customer, when a GPU customer can find power and cooling, and when a cloud provider can turn both into a service.

That does not mean every announced project will be built exactly as initially described. Stargate Norway’s changing customer arrangements are a reminder that AI infrastructure plans can shift as financing, power access and demand develop. But the underlying problem remains: AI companies need much more than chips. By helping solve the siting and deployment problem, Nvidia is positioning itself closer to the center of the infrastructure market that its GPUs have helped create.