[Editor’s note: This post was written before news began to emerge of China’s new DeepSeek AI system. Watch The Energy Mix for updates as all aspects of the story develop.]
The US$500-billion plan to make the United States an Artificial Intelligence powerhouse faces a glaring flaw: not enough funding, power, nor grid infrastructure to back it, onlookers warn.
Stargate, a $100-billion proposal by OpenAI, Oracle, SoftBank, and MGX to build AI data centres in Texas, was one of Donald Trump’s first big announcements as U.S. president last week, with $400 billion more pledged by the tech giants over four years. Given its high-profile backers—and one highly visible skeptic—Stargate is making waves in the technology sector.
But some of its future users, including officials in the U.S. Department of Defense (DOD), are questioning how the rest of the power demand on the U.S. grid will be met while avoiding potential blackouts, reports Defense News
“This is not a problem that industry or the DOD can figure out by itself,” Jeff Waksman, a program manager leading an effort in the Pentagon’s Strategic Capabilities Office to develop a mobile nuclear reactor, said at a recent panel discussion in Virginia. “It’s about the nation’s grid as a whole.”
And “it’s probably the most underrated challenge of this huge $500 billion announcement.”
Data centres have massive energy demand that can put a strain on grid capacity. And that demand is rising: U.S. data centres will use 606 terawatt-hours (TWh) by 2030, up from 147 TWh in 2023, projects consulting firm McKinsey & Company. That would amount to 11.7% of total U.S. power demand.
Tesla founder Elon Musk, a prominent member of the Trump administration who reportedly has a rivalry with OpenAI CEO Sam Altman, openly suggested that access to financing could be another problem for Stargate. Trump was reportedly furious with Musk over his criticism, but he shrugged off the accusation, saying Musk “doesn’t like one of the people” involved in the project.
The government would not provide any money for Stargate, but would help its development “a lot” through emergency declarations, Trump said. Those declarations could address electricity access, which has so far been a bottleneck for AI development.
“We have to get this stuff built,” Trump said. “They have to produce a lot of electricity, and we’ll make it possible for them to get that production done very easily—at their own plants if they want.”
Business Insider reported on Jan. 27 that the first Stargate data centre, to be built in Abilene, Texas, will use an onsite gas plant. Renewables like solar are another option, but experts question how that would align with Trump’s energy policies, like his pause on wind energy development and Biden-era funding for clean energy projects and his move to literally define “energy” as excluding solar and wind.
Those policies are “a huge contradiction,” and also raise concerns that tech companies would get preferential treatment when connecting data centres to the grid, Line Roald, a power systems expert at the University of Wisconsin, Madison, told IEEE Spectrum.
According to TechCrunch, at least some of the power for Stargate will come from large-scale solar plus storage projects. The nuclear sector is also anticipating it will have a part to play, reports OilPrice.
Stargate could also spur development of more oil and gas, contributing to Trump’s promises to “Drill, Baby, Drill.” Doug Burgum, who awaits U.S. Senate confirmation to lead the Interior Department, has also suggested using “clean” coal as an option, says the Sierra Club.
As an important determinant of geopolitical power, AI development is being framed as a race with countries like China and Russia. But it is also a growing risk for its inherent dangers if it becomes uncontrollable by humanity. Former president Joe Biden had issued executive orders to control AI’s development, requiring companies to stop development if they couldn’t prove their programs were not harmful to the public. Days after the Stargate announcement, Trump signed an order revoking those guardrails that he called a “barrier to innovation,” and instead ordered an AI action plan to be developed within 180 days.
