US energy developers seek alternative power sources for AI data centers

As the artificial intelligence boom accelerates, US data center operators are seeking alternative power solutions to bypass significant electricity grid constraints. Linking large-scale data centers to the existing US power grid can take between three and six years, a timeline that conflicts with the rapid deployment needs of AI companies.

One proposed solution involves the use of fuel cells. Rystad Energy projects that investments in fuel cells for data centers could rise from $2.8 billion in 2025 to approximately $30 billion by 2030. Companies like Bloom Energy and FuelCell Energy are positioned to benefit from this shift toward on-site generation. Rystad estimates that roughly 40% of projected US data center capacity in 2030 could utilize dedicated on-site power.

Additionally, Texas-based developer HGP Intelligent Energy has submitted a proposal to the U.S. Department of Energy to repurpose two retired Navy nuclear reactors. Known as the CoreHeld Project, the initiative aims to supply 450 to 520 megawatts of continuous electricity to a planned AI data center near Oak Ridge, Tennessee. The project would utilize proven naval pressurized-water reactor technology, potentially providing power comparable to the consumption of 360,000 homes.