ICF analysis warns that grid deliverability constraints may limit US AI-era power growth despite sufficient generation capacity.
US electricity demand could rise 39% by 2035 from 2026 levels, driven by data centers, industrial expansion, transportation electrification, and electric heating. A new report from consulting firm ICF projects electricity demand will rise 21% by 2030 and 39% by 2035, with peak demand increasing 25% by 2035. Yet ICF argues that demand growth is colliding with a more immediate problem: whether transmission networks can deliver power to emerging load centers.
"The central question is no longer just how much electricity demand is forecast, but how much demand can realistically be served – and where," the report states. Rob Gramlich, president of Grid Strategies, agrees. "I think we mainly have a grid delivery problem," he told Data Center Knowledge. "It's not the only issue, but expanding transmission delivery capacity would help meet demand growth more than any other single action."
For data center developers, the consequences are already visible in site selection and power procurement. Developers are widening site searches, prioritizing locations with available transmission capacity, and negotiating earlier with utilities over energization schedules and interconnection timelines. In some markets, access to transmission infrastructure now shapes project schedules as much as land, fiber, or capital availability.
ICF estimates the US has about 26 GW of generating capacity above minimum reliability requirements, roughly 3% of installed capacity—a margin expected to shrink to about 20 GW by 2030. In PJM and ERCOT, the cushion has largely disappeared. "As load growth has outpaced resource additions, any pre-existing excess capacity has quickly been absorbed," the report notes.
The firm projects 445 GW of new generation additions through 2030, including solar, storage, wind, and natural gas resources. Yet much of that capacity remains in development, creating a mismatch between when power is needed and when infrastructure can be deployed. US investor-owned utilities are expected to spend roughly $178 billion on transmission projects between 2025 and 2028, and the Department of Energy has backed plans for approximately 7,500 miles of new transmission lines by 2030.
However, the report warns that capital spending alone will not resolve near-term constraints. "Large capital plans will not translate into timely relief in the highest-growth areas without execution that navigates supply chain, siting, permitting, labor, and other challenges," ICF wrote. Gramlich notes the challenge extends far beyond a handful of data center hotspots: "It is acute everywhere that we have significant power demand including Texas, the Mid-Atlantic, Midwest, Southeast, Great Plains, and much of the West."
Jigar Shah, former director of the US Department of Energy's Loan Programs Office, agrees that transmission has become a central constraint. "This is what everyone is saying," he told Data Center Knowledge. "We don't have a generation problem, we have a wires problem." Shah argues that utilities can unlock additional capacity before major new transmission projects come online. "The best way to solve it is to get more out of the wires we have already paid for with batteries, demand flexibility, and grid enhancing technologies," he said.
Utilities and grid operators are exploring phased energization plans, flexible interconnection agreements, and grid-enhancing technologies that could bring large loads online before major transmission upgrades are completed. For data centers, this pressure is particularly acute. ICF projects that ERCOT demand will rise 53% from 2025 to 2035, while PJM faces mounting transmission constraints around Northern Virginia and emerging data center markets in Illinois and Ohio—regions expected to experience load growth exceeding 5% annually over the next decade.
Because transmission development moves more slowly than many large-load projects, utilities and grid operators may need to rely on alternative approaches, including grid-enhancing technologies, flexible interconnection arrangements, and conditional service agreements to bridge the gap. For data center developers, access to power is becoming less a question of how much generation exists nationally than whether local infrastructure can deliver it.