Nvidia and PreussenElektra Scout Decommissioned German Nuclear Plants for AI Hubs

Nvidia and German energy firm PreussenElektra have launched a joint assessment to determine whether decommissioned nuclear power plants can be converted into high-performance artificial intelligence data centers. The partnership, announced on October 21, 2026, aims to bypass the lengthy permitting and construction timelines that typically delay new computing infrastructure.[3][4]

The initiative focuses strictly on repurposing the land, water access, and high-voltage grid connections left behind by the shuttered atomic stations. Unlike recent agreements in the United States, the German project explicitly rules out restarting any dormant reactors to generate electricity from fission.[3][4]

PreussenElektra, a subsidiary of the E.ON Group, is currently responsible for decommissioning and dismantling eight nuclear facilities across Germany. The country permanently closed its final three operational reactors in April 2023 following a long-planned phaseout accelerated by the 2011 Fukushima disaster.[2][4]

The utility is now managing the teardown of approximately 30 atomic stations nationwide. As the physical dismantling progresses, the company is seeking long-term commercial uses for the sprawling industrial zones that remain, which already possess the heavy infrastructure that modern data centers require.[2][4]

Bypassing the construction bottleneck

Securing suitable land and adequate power has become the primary constraint on the global expansion of artificial intelligence. Major data center operators report that mounting barriers to bringing new projects online—including zoning approvals and grid connection delays—are severely limiting the industry's ability to deploy new capacity.[5]

Former nuclear sites bypass many of these hurdles because the necessary infrastructure is already in place. The massive transmission lines that once exported gigawatts of nuclear electricity to the German grid can theoretically be reversed to import power for tens of thousands of specialized graphics processing units.[1][3]

"AI factories require access to power, land and high-capacity infrastructure, and former industrial sites could provide a strong foundation for this next generation of computing," said Markus Hacker, Nvidia's senior regional director. He noted the assessment will determine how these specific locations meet the operational demands of AI workloads.[1][2]

PreussenElektra management views the partnership as a way to replace the economic activity lost when the reactors went offline. "Our former power plant sites have valuable industrial infrastructure and offer opportunities for new, future-oriented uses," said Guido Knott, chairman of the utility's management board.[2][4]

Knott emphasized that the utility intends to conduct a rigorous evaluation of the technical and regulatory conditions. "In doing so, we are combining the responsible reuse of our sites with new economic opportunities for the regions," he stated during the initial project announcement.[1][2]

A different approach to nuclear power

The German strategy contrasts sharply with how American technology giants are approaching the intersection of nuclear energy and artificial intelligence. In September 2026, Microsoft signed a 20-year power purchase agreement to support the planned restart of Unit 1 at the Three Mile Island nuclear plant in Pennsylvania.[4]

Nvidia itself has recently backed initiatives aimed at deploying new nuclear generation. In July 2026, the chipmaker partnered with Amazon Web Services and several small modular reactor developers on the Prometheus project, which uses artificial intelligence to accelerate the deployment of advanced nuclear technologies.[2]

The PreussenElektra assessment, however, is purely a real estate and infrastructure play. Because Germany's legal framework prohibits the resumption of nuclear generation, any data center built on these sites will draw its electricity from the broader German grid, which relies increasingly on wind and solar power.[2][4]

This distinction means the project will not add new baseload generation to the European energy market. Instead, it will consume existing grid capacity, raising questions about whether the regional transmission networks can support the massive continuous draw of an AI factory without destabilizing local supply.[3][4]

Evaluating the remaining hurdles

The current phase of the partnership remains entirely exploratory, with no definitive construction agreements, selected locations, or committed investment amounts. The two companies are conducting an open-ended feasibility review to identify which of the eight available sites warrant more detailed engineering studies.[2][3]

Converting a site undergoing active nuclear decommissioning introduces unique logistical and regulatory complications. An operating data center and a dismantling operation require vastly different security perimeters, safety protocols, and environmental monitoring, meaning the two activities may struggle to coexist on the same footprint.[1][3]

The timing of the teardown process presents another significant constraint. While the grid connections exist, the physical space required for server halls and cooling towers may not become available until specific reactor buildings and containment structures are fully decontaminated and demolished.[2][3]

Nvidia and PreussenElektra must also determine whether the existing power infrastructure can deliver the specific reliability standards demanded by continuous AI training runs. A grid connection designed to export power does not automatically guarantee the uninterrupted, highly stable voltage required by sensitive silicon.[1][3]

The sheer scale of modern artificial intelligence clusters makes finding suitable real estate increasingly difficult. A single training cluster utilizing 100,000 next-generation GPUs can consume upwards of 150 megawatts of continuous power, equivalent to the draw of a small city, severely limiting where such facilities can be placed.[3][5]

By targeting locations that previously hosted gigawatt-scale generation, developers can bypass the multi-year queue for new high-voltage grid interconnects. The former Biblis nuclear power plant in the German state of Hesse, for example, has already been earmarked to host a laser-based nuclear fusion research facility by 2035.[1]

The next verifiable milestone for the Nvidia and PreussenElektra partnership will be the publication of their initial site assessments. Until the companies identify a specific location and secure the necessary grid import agreements, the concept remains a theoretical solution to the industry's most pressing physical bottleneck.[2][3]