The scramble to find more electricity to power AI data centers could well become the dominant concern of Trump’s second term.
Just look at that photo of the billionaires seated near President Trump as he took the oath of office on Monday in the Capitol Rotunda and later declared an “energy emergency.” The billionaires got better seats than any of Trump’s cabinet nominees. In that frame, Mark Zuckerberg is flanked by Jeff Bezos,(who is engaged to Lauren Wendy Sanchez), flanked by Sundar Pichai and flanked by Elon Musk. Tim Cook was seated nearby as well, often looking distressed throughout Trump’s speech.
It matters that all the men have given large sums to Trump’s campaign (former President Biden warned of an oligopoly taking shape) , but what matters more is that all have huge stakes in how fast AI can flourish in the US. All their companies operate huge data centers, or want to, and contribute to various cloud backbones used by every major corporation. What’s driving the huge demand for data center compute? AI, of course.
Republican Virginia Gov. Glenn Youngkin delivered a State of the Commonwealth address on Monday calling for Virginia to continue to be the “data center capital of the world.” So far, the growth has been nothing short of explosive, with Northern Virginia counties of Prince County and Loudon at the epicenter of the trend. Meanwhile, the Virginia General Assembly is facing a slew of bipartisan bills to provide greater state oversight of large data centers, even as a centerpiece of the legislative package got tabled by a House panel on Monday.
Where is the power for data centers coming from? Nearly all of it comes from regulated electric utilities, like Dominion Energy in Virginia, which rely on traditional fuels like oil and gas, wind and solar to meet demand. Trump’s condemnation of wind energy in remarks on Monday were almost as prescient as his “Drill, baby drill!” refrain. Never has the need for oil and gas seemed more essential as the persistent demand for data center growth. Even nuclear power generation, locally, is back on the table for the modern AI era.
Trump’s pick to lead the Interior Department, former N. Dakota governor Doug Burgum, last week made a plea for “baseload electricity” generated by 24x7 burning of coal and natural gas. “Without baseload, we’re going to lose the AI arms race to China,” Burgum told the Senate Energy and Natural Resources Committee. “AI is manufacturing intelligence, and if we don’t manufacture more intelligence than our adversaries, that affects every job, every company and every industry.”
Fierce Electronics and Fierce Network have written plenty on ways data centers are working to be more efficient in how they cool their hot servers and how energy-sucking semiconductors can be made more efficient, even with future materials. On that latter point, GPU maker Nvidia and CEO Jensen Huang have consistently argued that while each of its latest GPUs uses more watts of power, the overall efficiency and capability of compute has increased, shortening the time needed to process any given algorithm.
Yet, even as algorithms can be processed more efficiently with the latest GPUs, the overall demand for compute continues to rise.
What can be done? The answer, apparently, is generating more electricity, but how the nation gets there and stays ahead of China is yet to be revealed…Trump is offering up a major infrastructure announcement on Tuesday and the billionaires may already be on board. (The White House posted his plans for an "energy emergency" that includes his plan to "use all necessary resources to build critical infrastructure." ) On Tuesday, Trump joined leaders of Oracle, OpenAI and SoftBank to announce a five-year, $500 billion joint venture to boost AI infrastructure.
RELATED: Stargate joint venture to pump $500 billion into AI infrastructure
Will Trump's chant become: “Generate baby, generate”? Best to pay attention to see how energy policy is going to be at the center of everything under 47.