特朗普要求大型数据中心自发自用，业内称全天候100%供电落地困难

Row of large industrial energy generators with exhaust pipes and cooling units

Industry experts say AI data centers can use onsite generation, but supplying 100% of a facility’s power reliably around the clock is far more difficult in practice.Getty Images

With data centers dominating industry and policy discussions, it was only a matter of time before President Donald Trump weighed in.

In an Oct. 21 Truth Social post, Trump said new, large-scale data centers should generate their own electricity and return excess power to the local community or grid. He also suggested that developers make financial arrangements with host communities, including payments to residents:

“As the President of the United States, I have been watching Data Centers very closely, and have come to a very strong conclusion. New and large, in some cases gigantic, Data Centers are mandated by me to make their own electricity, and give or sell any excess back to the local community, or grid, and therefore there is electric price increase. In fact, there could be a substantial decrease. Other elements likewise! If companies want to make their Data Centers popular, pay the community, and/or individuals within the community, a ‘dividend.’ I have watched the process closely, and Data Centers that were highly unpopular in a certain area, have become extremely popular when a financial deal is struck. MAKE DEALS, and get them built!!! GREAT FOR THE U.S.A. President DJT”

The comments are likely to intensify debate across the sector. Many projects already under construction have secured grid connections, signed power purchase agreements (PPAs), or otherwise locked in their energy strategy. If Trump’s position were translated into formal policy, developers would face difficult questions about whether those plans would need to change.

Equipment availability would be another major challenge. Many of the components required for gas-fired, nuclear, and battery-backed onsite power systems are already in tight supply, with some turbine models effectively sold out through 2031.

That would present a serious challenge for data center designers, engineers, developers, and operators, said Patrick Hughes, senior vice president of strategy, technical, and industry affairs at the National Electrical Manufacturers Association (NEMA). Hughes was involved in developing the AI Data Center Energy Performance Framework,co-developed by NEMA, the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE), and Pacific Northwest National Laboratory (PNNL).

“It’s possible for a standalone AI data center to provide sufficient onsite power to fuel its own operations – whether it’s a combined-cycle plant or solar plus storage, we have the technology,” Hughes said. “But that doesn’t mean it’s feasible for a single AI data center – let alone the whole fleet of AI data centers – to supply 100% of their own power and operate reliably 24/7; that requires significant operational management and maintenance.”

Permitting timelines, fuel and equipment supply chains, interconnection rules, and community considerations vary widely by location, Hughes added. Those factors will shape what is practical at any given site. For many facilities, the most likely outcome may be a hybrid approach that combines onsite generation and storage with a grid connection.

Why AI Data Centers Can’t Easily Go Off-Grid

There are substantial technical and logistical barriers to any broad requirement that AI data centers supply all of their own power. AI training and inference facilities have power densities and load profiles that differ sharply from those of traditional data centers and from facilities that have historically relied on co-located generation and storage.

At the same time, the AI infrastructure buildout is grappling with enormous demand for electrical equipment and skilled labor across utilities, buildings, transportation, and industrial manufacturing. Whether or not developers are pushed toward a bring-your-own-power model, they are already competing with other sectors for turbines, switchgear, transformers, batteries, and specialized labor.

Regardless of the onsite resource selected, the AI Data Center Energy Performance Framework offers technical standards and guiding principles for safety, resilience, and grid interaction. Data center designers, engineers, developers, and operators can use it to improve energy and water efficiency while supporting grid reliability and resilience through load flexibility and other grid-interactive operations.

Existing Power Deals and Renewable Commitments Complicate the Shift

A blanket onsite power requirement would also have to contend with existing clean-energy mandates, corporate sustainability goals, and regional differences in the availability of renewable resources. Many AI data centers already have agreements in place to draw power from solar and wind projects.

The Stargate site in Abilene, Texas, offers one example. The project is already using natural gas, fuel cells, and gas engines during the construction phase and to power the first data centers now operating there. Developers are also planning to incorporate solar and wind, and the area’s solar generation potential was one of the reasons the location was selected. Abilene offers strong irradiance, abundant undeveloped land, and no major urban loads nearby. With developers scrambling to find 1.2 GW once all phases are complete, wind and solar are expected to supplement other power sources.

Applied Digital’s AI data center project in North Dakota provides another example. The company has arranged for 400 MW of stranded wind power to serve its Ellendale campus and continues to emphasize renewables in its energy strategy. It also recently signed a deal with data center developer Soluna to use output from Soluna’s 150 MW Briscoe wind farm for a new data center in Texas.

"We have said for a long time that the constraint in AI is energy access, not chips," Soluna CEO John Belizaire said in a press release. "We built Soluna to bring load to the generation, so the power would be ready when the demand arrived."

A Hybrid Power Model May Be the Most Practical Path

If policymakers decide to pursue Trump’s idea, they will likely need to translate it into a more flexible and workable policy. The AI Data Center Energy Performance Framework accounts for that kind of flexibility, avoiding a rigid preference for either bring-your-own-power or grid power. Its guidance on grid-interactive design, onsite energy, and thermal efficiency addresses the role of onsite and offsite renewables, as well as intelligent energy storage, in reducing strain on the grid.

“There’s no single answer, as the right onsite generation strategy will inevitably depend on each facility’s load, site, and regulatory landscape,” Hughes said.

About the Author

Contributor

Drew Robb has been a full-time professional writer and editor for more than twenty years. He currently works freelance for a number of IT publications.