Nvidia to Transform Old Nuclear Power Plants Into AI Data Centres

New Life for Old Infrastructure

The rapid advancement of artificial intelligence has triggered a massive surge in global demand for computing power. As the market leader in this sector, Nvidia is seeking faster, more cost-effective ways to build the necessary infrastructure. The company has announced a partnership with German energy firm PreussenElektra GmbH (PEL) to study the feasibility of converting decommissioned nuclear power plant sites into data centres.

The project focuses on facilities where dismantling began in April 2023, following Germany’s official exit from nuclear energy. PreussenElektra is currently overseeing the decommissioning of eight such stations, which experts will now evaluate as potential locations for AI servers.

Why Nuclear Sites Are Ideal for AI

Building modern data centres is an exceptionally complex and time-consuming process. It requires not only significant land but also access to high-voltage power grids, water infrastructure for cooling systems, and specific industrial permits. Former nuclear power plants inherently meet these requirements.

"The sites of former power plants possess valuable industrial infrastructure, providing an ideal environment for new, future-oriented projects," stated Guido Knott, Chairman of the Board of Management at PreussenElektra. Markus Hecker, a representative for Nvidia, added that the collaboration would make a significant contribution to the development of high-performance AI infrastructure in Germany.

Experimental Phase and Future Prospects

It is important to note that the project is currently in the research phase. The parties have not yet planned specific investments, nor have construction timelines been established. Furthermore, the project does not involve restarting nuclear reactors or powering data centres directly with nuclear energy; the focus is solely on the rational use of existing land and electrical grid connections.

For Nvidia, this strategy could offer a way to overcome challenges posed by the global shortage of data centres. If the study yields positive results, similar practices could be adopted in other countries where closed industrial facilities remain idle. For users globally, this development is significant as it highlights how technology giants are increasingly prioritising resource optimisation, which will ultimately impact the accessibility of AI services.