The Numbers: How Much Power Is Actually at Stake

The damage is already measurable. Permitting changes and federal funding withdrawals have wiped out 7 gigawatts of generating capacity on federal land in 2025 alone, according to consulting firm Wood Mackenzie. That capacity is gone — cancelled projects that will not come back.

The exposure ahead is far larger. An additional 12 gigawatts on federal land sits under threat, along with 80 gigawatts on private property. Combined with what has already been cancelled, the total threatened electricity supply reaches 92 gigawatts. Wood Mackenzie’s analysis puts more than $121 billion in energy investment in the crosshairs of federal bureaucratic delays.

To understand the scale, consider this: 92 gigawatts roughly equals the entire power generation capacity of Texas — the largest electricity market in the United States. Texas does not run on marginal capacity. It runs an industrial economy, 30 million people, and a grid that operates independently from the rest of the country. Threatening an equivalent volume of new power supply is not a rounding error in the national energy mix. It reshapes what the grid can deliver over the next decade.

That matters specifically because electricity demand is surging again after two decades of flat growth. AI data centers are driving a wave of new load that utilities and grid operators have not seen in a generation. Training large language models, running inference at scale, and storing the data that feeds machine learning pipelines all require continuous, high-volume power draw. Data center operators are signing long-term power purchase agreements precisely because they know supply is tightening.

When 92 gigawatts of new generating capacity — solar, wind, storage, and transmission infrastructure — faces cancellation or indefinite delay, the pipeline of electricity that AI infrastructure depends on shrinks. Power prices rise. Interconnection queues stall. New data center construction slows or shifts to jurisdictions with more reliable grid capacity. The arithmetic is straightforward: less new supply meeting faster-growing demand produces an energy bottleneck, and an energy bottleneck is a ceiling on American AI development.

The Irony Problem: An Anti-Red-Tape Administration Tangled in Red Tape

The Trump administration built its brand on slashing red tape. Deregulation and permitting reform sit at the center of its economic agenda — the explicit justification for dismantling dozens of federal rules since January 2025. That makes what’s happening in the American energy sector a genuine contradiction, not a talking point.

New permitting scrutiny requirements and deliberate withdrawals of federal funding mechanisms have already killed 7 gigawatts of generating capacity on federal land this year, according to consulting firm Wood Mackenzie. The cascading effect threatens another 12 gigawatts on federal land and 80 gigawatts on private property — a combined 92 gigawatts of electricity supply that the U.S. grid desperately needs. The total investment at risk exceeds $121 billion.

Most coverage frames this as an environmental story — a fight over clean energy policy and climate commitments. That frame misses the actual stakes. This is an industrial competitiveness story. Electricity demand in the United States is climbing sharply after two decades of flat growth, driven directly by the explosion of AI data centers. The power infrastructure required to sustain American AI leadership isn’t a future concern. It’s a present construction problem, and the federal government is actively blocking the solution.

The administration didn’t stumble into these delays through neglect or bureaucratic inertia. It created them through deliberate policy choices: new review requirements that slow project approvals and funding withdrawals that make power projects financially unviable before they break ground. That’s not deregulation. That’s the federal government adding friction to private energy investment at the worst possible moment for U.S. grid capacity.

An administration that campaigned on getting government out of the way of American industry is now the single largest source of delays threatening the energy backbone that American AI dominance depends on. The irony isn’t subtle — and the consequences aren’t theoretical.

The AI Collision Course: Demand Surging While Supply Gets Blocked

American AI ambitions run on electricity. That simple fact is now colliding head-on with a supply crisis the Trump administration is actively making worse.

Data center electricity demand has broken a two-decade streak of flat consumption, climbing sharply as AI model training, inference, and cloud infrastructure scale across the country. The expansion shows no sign of slowing. Power grids that coasted through the 2000s and 2010s without meaningful load growth are now facing sustained, aggressive demand from a single industry that the same administration publicly champions as a national priority.

On the supply side, the picture is deteriorating fast. A 2025 Wood Mackenzie study found that permitting changes and federal funding withdrawals have already canceled 7 gigawatts of generating capacity on federal land this year alone. The damage does not stop there. Additional regulatory scrutiny threatens to eliminate another 12 gigawatts on federal land and 80 gigawatts on private property — a combined 92 gigawatts of new electricity generation that may never reach the grid. That capacity loss represents more than $121 billion in at-risk investment across the U.S. energy sector.

To put 92 gigawatts in perspective: that is not a rounding error. That is the difference between a grid that can absorb explosive AI-driven power demand and one that physically cannot keep up.

The U.S. AI industry’s competitive position globally depends on access to cheap, reliable, abundant power. Data center operators site their facilities where electricity is plentiful and affordable. When domestic supply gets constrained through permitting delays and canceled projects, the pressure to build capacity elsewhere — in jurisdictions with fewer regulatory obstacles — grows. American AI leadership does not exist in a vacuum separate from American energy infrastructure. They are the same problem.

Most coverage of these permitting rollbacks frames the story as an environmental setback. That framing is incomplete. The real stakes are whether the U.S. electrical grid can physically support the scale of AI compute infrastructure the country is betting its technological future on. Right now, the administration is blocking the supply needed to win that bet.

Federal Land vs. Private Land: Two Different Problems, Both Getting Worse

The 92 gigawatts of threatened electricity capacity breaks into two distinct categories, and conflating them obscures how difficult this problem actually is to fix.

On federal land, the administration holds direct control. Permitting rule changes and federal funding withdrawals are policy levers that Washington pulled deliberately. Those decisions have already killed 7 gigawatts of generating capacity in 2025 alone, according to Wood Mackenzie. Another 12 gigawatts on federal land sits under active threat. Because these are executive-branch decisions, they could theoretically reverse with a policy shift — a new directive, a restored funding stream, a revised permitting framework. The damage is real, but the mechanism is legible.

Private land is a different and underreported story. The 80 gigawatts at risk there — the overwhelming majority of threatened capacity — isn’t being blocked by a single federal permit denial. It’s being strangled by regulatory uncertainty that ripples outward from federal policy signals. Developers financing large-scale power generation projects, whether solar farms, wind installations, or battery storage facilities, require predictable regulatory environments to secure capital. When the federal government introduces volatility into energy permitting at scale, lenders reprice risk and developers shelve projects on land the federal government doesn’t even touch. That chilling effect on clean power investment doesn’t show up in a single headline-grabbing cancellation — it shows up in projects that quietly never break ground.

This distinction carries a critical implication. Federal land problems have a policy antidote. Private land problems do not reset automatically when Washington reverses course. Investor confidence, once eroded, rebuilds on its own timeline. Wood Mackenzie puts the total investment exposure across both categories at more than $121 billion. That capital, tied directly to the electricity infrastructure that AI data centers require to expand, doesn’t wait for political cycles to stabilize. The grid capacity gap widens in real time, regardless of whatever permitting reform might eventually arrive.

What Comes Next: Grid Stress, Higher Costs, and Who Bears the Risk

The math here is unforgiving. With 92 gigawatts of clean generating capacity now at risk — 7 gigawatts already cancelled in 2025 alone — utilities and grid operators face a shrinking menu of options as electricity demand accelerates. The gap between supply and demand doesn’t stay empty for long. Grid operators will fill it with whatever generation is available: aging natural gas peakers, coal plants that were already scheduled for retirement, and diesel backup systems that nobody wanted to lean on. That means dirtier power, higher operating costs, and a grid running closer to its stress limits.

Those costs land somewhere. Businesses and residential customers will absorb them through higher electricity prices as competition for a constrained power supply intensifies. The current policy debate in Washington has largely skipped over this consequence. Lawmakers focused on cutting what they frame as regulatory bloat are not publicly accounting for the price spike that follows when $121 billion in planned energy investment gets frozen or cancelled outright.

The private sector is not waiting for that conversation to happen. Wood Mackenzie’s analysis makes clear that project cancellations are not a forecast — they are a present-tense reality. Developers pulling permits on federal land and walking away from projects on private property are making hard financial decisions based on the regulatory environment as it exists right now, not as it might improve later.

For AI infrastructure, the exposure is direct. Data centers require firm, reliable, long-term power commitments. Hyperscalers and colocation operators building out capacity to meet surging artificial intelligence workloads need electricity supply locked in years in advance. A grid under stress, drawing from expensive and unreliable backup generation, is not a foundation anyone can build an AI strategy on. Every gigawatt of cancelled clean power capacity is capacity that won’t be available when the next wave of data center demand comes online — and that wave is already forming.