Can China beat the United States in the new-age arms race for AI dominance? Some commentators think so, saying they came back from recent trips “stunned” at how far China has come.
The issue these tech journalists uncovered wasn’t lack of innovation – after all, the U.S. boasts 40 leading AI models to China’s 15. It wasn’t tardy investment, either. To the contrary, U.S. industry poured nearly 12 times as much money into AI as China did last year alone.
The real problem facing AI in America is electricity.
China is investing heavily in a next-level power grid and currently enjoys an oversupply of electricity at least twice its capacity needs. The U.S., on the other hand, scrapes by with electricity reserves of 15% on a good day. We face severe bottlenecks that, according to Goldman Sachs, threaten an AI energy crisis. And with AI data centers representing more GDP growth than consumer spending, it’s clear that AI will be a critical component of economic success going forward.
Producing energy, however, is only a small part of the problem. Distribution is a huge headache. String together the power lines the U.S. has installed since the advent of the lightbulb, and you’d reach the moon. To power AI from the grid, we would need to hang new lines going two-thirds of that distance within five years – a financial and physical impossibility at the pace of current infrastructure build outs.
So, is China’s AI ascendance a foregone conclusion? From my vantage point in the energy industry, I can say with confidence that it’s not. The U.S. is confronting serious challenges but in them lie incredible opportunities – if only we are willing to reimagine the future of energy the American way.
We can do this by generating the power right where it’s required. Bypass long transmission lines and the complications of heavy grid dependence. Slim projects down from huge undertakings to modular add-ons for rapid deployments that take months, not years. And do it all with clean energy producing significantly less CO2 emissions than traditional methods and integrating technologies that accept emerging fuel sources like highly sustainable biogas produced from waste.
Sound like a pipe dream? It’s not. Power conversion technologies available today, called fuel cells, are already notching these achievements and more at thousands of sites nationwide. These fuel cell systems efficiently, quietly, and affordably turn natural gas and/or biogas into electricity. They do so away from the grid – on data center campuses and at other facilities, from factories to hospitals – 24/7, onsite and without backup. And fuel cell blocks can be installed in as little as 90 days and added to as demand increases for carefully calibrated expansion and investment.
Fuel cell systems are ideally suited to power AI data centers, but the benefits extend beyond the tech sector. Onsite generation imposes no cost or energy-availability impact on residential or business ratepayers anywhere.
Fuel cells also take advantage of domestic fuel sources, preserving and enhancing America’s energy independence. And decentralized power production is more resilient and safer from a national security standpoint, offering no large, interdependent target for terrorists or malevolent state actors to take down.
Leaders across the political spectrum are beginning to recognize the potential in onsite power to secure America’s global AI leadership – not by mimicking China’s centralized, technocratic infrastructure planning but rather by expanding on the U.S. model of widely distributed, private sector innovation.
Nestled within recent government initiatives, for example, were Clean Electricity Investment tax credits worth up to 70% of eligible project costs, which should spark more onsite power builds. This measure follows on the heels of a July 2025 Executive Order designed to expedite permitting and roll back red tape hampering critical AI energy infrastructure.
These efforts are a nice start, but the U.S. must do more.
Elected officials at all levels can help by requiring better collaboration across jurisdictional lines separating federal and state regulators. Even as longtime chip manufacturing competitors NVIDIA and Intel join forces in a $5 billion investment deal aimed at taking on China-based Huawei’s AI chip development, U.S. leaders of all political persuasions can come together to avert the AI electricity crisis.
With common-sense, forward-thinking policies, we can advance the U.S. model of private sector innovation and investment that has made our tech sector second to none – and will keep us on the forefront as the AI era unfolds.
James Roth is an award winning author and head of government affairs and policy at Bloom Energy, a leading provider of onsite energy solutions in the U.S. and globally.