The United States leads the world in artificial intelligence, but it’s not guaranteed to stay there. The bottleneck isn’t talent, ideas, or capital—it’s electricity.
Electricity is the binding constraint for building and using hyperscale data centers, essential for training today’s advanced AI models. The Department of Energy projects that data center electricity demand will nearly triple by 2028. Meeting that growing demand will require building generation and transmission capacity at unprecedented speed. Yet some AI developers are struggling to obtain the electricity they need: Recently, the Federal Energy Regulatory Commission (FERC) blocked Amazon from directly connecting a data center to a Talen Energy nuclear power plant.
One major roadblock for expanding power availability is today’s permitting process. A recent Brookings Institution study made the case for permitting reform. Focusing on wind and solar power, the authors concluded:
[Energy] projects can and do get delayed at any stage of the process—local communities may oppose projects that take over their property or obstruct their views; states may impose high environmental review burdens or block unpopular projects; interstate grid operators are backlogged with connection requests; federal authorities may take years to grant the full range of permits a project requires.
Even building a conventional natural gas plant requires navigating a maze of federal, state, and local approvals—from the Environmental Protection Agency, FERC, local zoning boards, state environmental agencies and utility regulators, and sometimes the US Army Corps of Engineers. The challenges don’t stop with generation: Leading AI providers find the electric transmission grid to be their most difficult challenge, where construction and interconnection require several federal and state approvals and negotiations with utilities and electric grid operators.
The timelines can be staggering. A utility-scale solar or wind project can spend up to three years just negotiating permitting and grid connection. The American Public Power Association (APPA), which represents city-owned utilities, notes that even though environmental regulatory agencies often have deadlines for decision making, opponents have learned to sidestep the deadlines by triggering new proceedings, resulting in an “open-ended, ever-changing process in which a permit never gets approved.”
Meanwhile, the global AI race is heating up, demanding more from our energy resources. US companies released 40 notable AI models last year, compared to China’s 15. China has its own data center challenges, but the country is catching up in patents and research output, and strong AI models are emerging from countries like Canada, France, and South Korea. America’s innovation advantage depends on staying ahead in both computing power and the electricity that fuels it.
There have been efforts to streamline permitting by reducing administrative bloat and simplifying certain processes. A recent FERC decision seeks to streamline transmission planning. The APPA suggests enforcing decision deadlines, eliminating duplicative environmental studies and redundancies when multiple agencies are reviewing a project. The Institute for Progress and the R Street Institute have additional proposals. But policy changes need to stick: Environmental requirements changed from Obama, to Trump I, to Biden, and to Trump II. Such flip flopping drives up business costs.
Iowa’s recent improvements in government efficiency can serve as a template for permitting reforms. Gov. Kim Reynolds (R., Iowa) drove what amounted to an all-of-government cost-benefit review, requiring each agency to identify its core mission, the effectiveness of its programs, and where there was duplication and misalignment. The result was the elimination of 21 agencies from her cabinet, consolidation of licensing functions, and cutting the time taken for some regulatory decision-making by 90 percent. Lawmakers could follow this model for testing the usefulness and effectiveness of permitting and licensing procedures.
Texas provides another model. The 1977 Texas Sunset Act mandates that state agencies be reviewed every 12 years. The process has completely abolished 42 agencies and abolished another 53 by consolidating or transferring functions. Congress and state legislatures could sunset problematic permitting procedures absent the relevant agencies demonstrating that their requirements and procedures deliver economic benefit. This should be no problem for truly valuable regulations.
If the US wants to lead the future of AI, it needs to clear the brush for building electric infrastructure. Fast.
The post America’s AI Future Needs Faster Permitting appeared first on American Enterprise Institute – AEI.
