The United States needs more electricity now, not tomorrow. The country is locked in an existential competition with China over supremacy in artificial intelligence and the technologies that will define economic and military power in the 21st century. AI is extraordinarily electricity-intensive. Yet, America’s power grid is fragile, aging, and constrained by a political and regulatory system that makes it painfully slow to build new generation.
The data centers that power the AI revolution are coming online faster than regulators, utilities, and lawmakers can adapt. Instead of confronting the hard question of who should pay for the electricity required to win the AI race, politicians are pandering to the populist wings of both parties. The result is a growing threat to grid reliability, AI leadership, and long-term economic growth.
AI promises to benefit every American. The data centers that make it possible bring jobs, tax revenue, and global prestige to the communities where they are located. It is plainly in the national interest to accelerate, not slow, the deployment of AI. That requires cutting through regulatory barriers that delay new power generation by years. It also requires political leaders to stop misleading the public about rising electricity bills. Yes, AI data centers increase demand for power. But they also drive productivity, economic growth, and higher wages. Their spillover effects are overwhelmingly positive. Data centers should be celebrated, not demonized.
America’s largest technology companies are building these facilities. Hyperscalers such as Alphabet, Amazon, and Microsoft have the financial strength to pay for the power infrastructure AI requires. And they are willing to do so. Good public policy should facilitate that investment, not obstruct it.
AI data centers are fundamentally different from traditional industrial loads. A single hyperscale facility can consume as much electricity as a midsized city, with demand that is constant, concentrated, and nonnegotiable. Training large AI models and running inference at scale requires uninterrupted, 24/7 power. Reliability is not optional.
Last week, President Donald Trump urged governors in 13 Northeastern states, where the nation’s largest power grid is located, to pursue an emergency power auction in which technology companies would sign 15-year electricity contracts. Under this approach, data center operators would pay for the power they consume. Long-term auctions are transparent, efficient, and fair. They send clear price signals, encourage new generation, and benefit all stakeholders. This is sound policy.
But if populists and far-left politicians continue to demonize AI by blaming it for higher electricity bills while ignoring its productivity gains and growth effects, then hyperscalers will respond rationally. They will go off-grid. That process has already started. AI companies are increasingly building self-contained power systems, typically fueled by natural gas turbines connected directly to pipelines. These off-grid data centers may be efficient for individual firms, but if the trend accelerates, the consequences for regional electricity grids will be severe.
Large data centers are ideal customers for utilities. Their demand is predictable, continuous, and creditworthy. Companies such as Microsoft and Amazon reliably pay their bills and can help finance new generation, grid upgrades, and transmission expansion. Their presence spreads fixed costs, including substations, transmission lines, and maintenance, across a broader base.
TRUMP’S UNNECESSARY GREENLAND THREATS UNDERMINE US INTERESTS
If AI data centers leave the grid, those fixed costs do not disappear. They are simply shifted onto households and small businesses. Local tax revenues decline. Investment in new generation becomes harder to finance. Political pressure mounts to cap rates. Utilities struggle to maintain infrastructure. A classic utility death spiral emerges.
The choice is clear. The U.S. can embrace AI data centers and build the electricity system they require. Alternatively, it can drive them off-grid and undermine its own economic future. Winning the AI race requires more power, faster construction, and honest leadership. Anything less harms the nation.
James Rogan is a former U.S. foreign service officer who later worked in finance and law for 30 years. He writes a daily note on the markets, politics, and society.
