If there’s a single bottleneck preventing the U.S. from reaching its full potential with Artificial Intelligence (AI), it is energy. A recent report from the U.S. Department of Energy (DOE) projected that the energy burden of data centers will triple by 2028. In a November 2025 meeting, Google’s head of AI infrastructure, Ahmin Vahdat, said that the company will have to double its compute every six months to account for the demands of AI.
These demands for new power mean not only new generation, but immense mileage of new transmission and distribution infrastructure as well. Such a large-scale demand and compressed timeline are unprecedented in world history, and present questions about tradeoffs and competing priorities. Assuming we can secure the power needed for homes and businesses first, the new demand for data centers alone threatens the very potential for meaningfully robust AI.
Having AI in the national toolbox is not a matter of access to a chatbot or a virtual assistant to simplify and synthesize information. At scale and in government, industrial, and military contexts, it is a matter of fundamental resilience, security, and efficiency. Having it or not is the difference between taking the grid to the next level and obtaining a permanent military advantage or remaining in a hybrid analog-digital world where every action and victory, be it agricultural or atomic, is hard won and demanding.
As the national security implications of AI become more pressing, it calls for an examination of the U.S.’s ability to account for such an exponential increase in energy demand. This is especially the case given that our main competitor, China, may have a superior energy grid already. Since 2004, the communist country has been engaging in a massive build-out of ultra-high voltage (UHV) transmission lines and juicing its energy sector with coal and cheaply sourced renewable technology. Long-distance UHV lines enable the movement of large quantities of power at a scale that the United States can’t yet match. As of November 2025, the U.S. has built zero miles of operational UHV lines, while some numbers estimate that China has reached around 30,000.
Regardless of whether the particulars are true – the CCP has a tendency to inflate data on this exact subject – the country has undeniable advantages over the United States. Namely, it has a stronger industrial base and a centralized economic structure that allows it to wield that base from the top down. Outside of a market economy, robust environmental and worker safety regulations in the U.S. limit largescale and high-speed infrastructure projects today. The exact kind called for to meet the challenge of the day.
While the centralized model hurts them in the long run for a host of reasons, particularly regarding overall economic efficiency, it is advantageous for energy infrastructure development. When China wants to build a transmission line, it certainly isn’t bogged down in regulation and lawsuits.
Cheap electricity is often touted as one of the reasons that the country has been able to participate in the AI race at all, despite its second-mover disadvantage and the U.S.’s stringent export controls. Intellectual property theft and other factors also play in. To make up for the computational gap, Chinese data centers are using larger quantities of lower-quality chips and cheap energy. This presents a challenge—but also a much greater opportunity: with stronger energy infrastructure, the U.S. could widen its lead in AI, where we already hold the advantage in core technologies and intellectual property.
So, how do we get the power?
The Trump administration has recently initiated several changes in the right direction. For example, the president has continually been reforming the National Environmental Policy Act (NEPA) throughout his presidency, a bloated piece of legislation that has been a thorn in the side of energy projects for decades. Trump has implemented broad executive-level NEPA reforms and improved the fast-tracking process in the One Big Beautiful Bill Act. Moreover, the Standardizing Permitting and Expediting Economic Development Act (SPEED) act, which is designed to make it far easier for tech companies to obtain federal permits for AI infrastructure, passed through the House in December 2025.
The president has also led an admirable charge towards Nuclear and Natural Gas (LNG). 2025 even saw the restarting of nuclear reactors like Three Mile Island, and the DOE is fast-tracking eleven high-tech nuclear projects. On the LNG front, the EIA reported record highs last October, and the administration ended the LNG export pause.
With all that being said, it still isn’t nearly enough. Since 2025, the U.S. has built roughly 423 miles of transmission and 53 gigawatts (GW) of additional total generating capacity. That’s a solid amount, but it’s a far cry from the 5,000 miles and 80 GW projected annually to account for the predicted growth in demand. To further complicate the energy capacity picture, the administration has handicapped wind and solar, cancelling $7.5 billion in funding, freezing permitting on federal land, and cancelling multiple key projects this year, such as Empire Wind, which would have generated 810 megawatts of electricity.
Rejecting available source of energy, especially given that renewables can often be built faster than thermal or nuclear projects, may only slow innovation. With China close on our heels and willing to go to any end to beat us, it may be time to stop letting the perfect be the enemy of the good, however well-reasoned, and accept more energy sources wherever they are.
The private sector has already embraced this heuristic. Data-center operators and AI firms are increasingly pursuing dedicated supply – either by contracting directly for large blocks of power or by turning to on-site generation – because they understand that time is of the essence, and if the federal government is unwilling or unable to step up to the plate, they may have to take matters into their own hands.
But a patchwork of private energy generation does not a power grid make. Cooperation between federal and state governments is necessary to streamline the process and drive down prices in a dramatic fashion. Siting, permitting, interconnection, and cost-allocation all cross state lines, and the cheapest, most reliable power is often far from the load. A federal framework can set predictable timelines, standardize reviews, and clarify who pays for interregional lines, which is the only way to build transmission at the scale needed without every project becoming a political battle. Without that backbone, the U.S. will keep adding pockets of generation while failing to deliver power to where it is actually needed.
If the United States wants to win the AI race, it has to treat electricity as strategic infrastructure rather than a partisan talking point or a local zoning problem. That means committing to a buildout on a national scale and adding generation fast enough to keep data-center load from becoming a drag on the broader economy. The relevant question is not whether an energy source is perfect, but whether it can be deployed quickly, integrated reliably, and expanded at scale. America doesn’t need to copy China’s politics to copy its urgency; all it needs is the political will to build more power plants, more lines, more substations, and to do so much faster.
Guest post by Atticus Vernacchio
The Alliance for Innovation and Infrastructure (Aii) is an independent, national research and educational organization working to advance innovation across industry and public policy. The only nationwide public policy think tank dedicated to infrastructure, Aii explores the intersection of economics, law, and public policy in the areas of climate, damage prevention, eminent domain, energy, infrastructure, innovation, technology, and transportation.