US President Trump Proposal Would Require AI Data Centers to Self-Generate Power

Former President Donald Trump has put forward a proposal under which large-scale data centers built for artificial intelligence workloads would be required to produce their own electricity, returning any surplus power to the grid or to the surrounding community.

As part of the same proposal, he suggested that developers reach financial arrangements with the communities hosting their facilities, including direct payments to residents, It indicates that, should the proposal be converted into formal policy, it would raise substantial questions about the energy strategies and equipment supply chains currently used by data center developers.

Supply Chain Constraints Could Undermine the Policy

Any requirement to build on-site power systems would run up against existing shortages in equipment supply. According to Patrick Hughes, Senior Vice President of Strategy, Technical, and Industry Affairs at the National Electrical Manufacturers Association (NEMA), some turbine models are effectively sold out through 2031.

Gas-fired, nuclear, and battery-backed components for onsite power systems are already constrained, meaning developers seeking to comply with a self-generation mandate could struggle to secure the necessary hardware.

A self-generation requirement would also directly challenge the grid connections and power purchase agreements (PPAs) that data center developers currently depend on for their electricity.

Technical Reality Check on Full Self-Sufficiency

Hughes acknowledged that a standalone AI data center can technically supply its own power, whether through a combined-cycle plant or through solar paired with storage. He cautioned, however, that it is not feasible for an entire fleet of AI data centers to meet 100 percent of their own power needs while operating reliably around the clock, 24/7. Achieving that level of independence would demand significant operational management and maintenance.

The facilities in question differ from traditional data centers as well. AI training and inference facilities have distinct power densities and load profiles, and the broader AI infrastructure buildout faces immense demand for electrical equipment and skilled labour across multiple sectors. Permitting timelines, fuel and equipment supply chains, interconnection rules, and community considerations all vary by location.

Friction with Clean-Energy Commitments

A blanket onsite power requirement would also collide with clean-energy mandates and corporate sustainability goals. Many AI data centers already hold agreements to draw power from solar and wind projects, and a strict self-generation mandate could complicate those existing commitments.

Hybrid Approach Framed as the Workable Path

Under the current landscape, developers may need to adopt a hybrid model that combines on-site generation and storage with a grid connection. This model is accounted for in the AI Data Center Energy Performance Framework, which provides technical standards and guiding principles covering safety, resilience, and grid interaction. The framework supports flexible energy strategies both on-site and off-site, including renewables and intelligent energy storage.

Hughes noted that the optimal onsite generation strategy depends on each facility's load, site, and regulatory landscape, a point that underscores why a one-size-fits-all mandate may not suit the diversity of projects now underway.