AI is spreading faster than any other technology in history, driven by competing hyperscalers and Millions of users. As a result, the data centers built to enable this growth have become increasingly popular Control by communities and across party linesincluding concerns about higher electricity bills.
But there is also an imbalance emerging: OpenAI, Google and Meta The US can finance and build AI infrastructure faster than the US can build the power generation and transmission needed to support it.
“There’s a real disconnect,” said Rob Gramlich, president of consulting firm Grid Strategies Assets. He said technology companies are known to move quickly, while utilities are “known to move very slowly” because they have to ensure that thousands of parts of an interconnected network work together, which requires “deliberate research and planning” that can take years.
This creates a “timing mismatch,” Gramlich said, which is why the power grid may not be ready to meet data center power needs. They are estimated to consume nearly 12% of all U.S. electricity by 2030, nearly six times the share before the AI boom in 2018 Lawrence Berkeley National Laba government-funded scientific research center with a focus on energy.
Electricity demand will also grow faster than expected in the next decade due to AI. NERC network monitor Grid reliability assessment in 2025 Projects: Summer electricity demand in North America will grow by more than 224 GW over the next 10 years – 69% above forecast growth last year – driven by new AI data centers. In the Western US grid region, planned data centers account for an average of 10% of demand forecasts and can be as high as 40% in some areas.
The grid was already limited before the AI boom. Gramlich said the U.S. electric industry has “fallen out of the practice of building new infrastructure” after 25 years of low growth in electricity demand. He said that’s because major grid expansions in the 1980s and 1990s gave utilities additional capacity and that electricity demand didn’t grow as much between 2000 and 2023 as in previous decades, when kitchen appliances and air conditioning were on the rise.
“But now we have electric vehicles, we have electric space heating, we have new manufacturing, most of which uses more electricity than traditional manufacturing, and then we have data centers,” Gramlich said. “The growth is about half of data centers and half of these other new things, but these new, fast-growing uses of electricity are all happening at once.”
Access to electricity as a bottleneck for data centers
Kathryn Burkewho leads U.S. specialty energy and power growth at insurance company Marsh, said Assets that access to electricity is “probably the fifth, if not fifth, bottleneck to data center development in the U.S. right now,” but that this demand is now colliding with an energy grid that still needs to be modernized.
She said companies are demanding more power and faster delivery speeds than the grid can handle. A gigawatt of electricity – what can power 750,000 US homes— can take “much more time” than the two years companies typically require to generate, Burke said. For example, for U.S. energy projects that came online in 2025, it took an average of five years from requesting a grid connection to commercial operation, the study found separate Berkeley Lab research.
Providing electricity also requires infrastructure that the country does not have. The Department of Energy said in July that there was an “urgent need” for more of them “due to growing data center utilization” and other factors such as a growing economy.
The most immediate consequence may not be widespread power outages. Gramlich said utilities generally don’t connect new customers if they can’t reliably serve them. Instead, data centers may have to wait years for full operations or accept temporary connections that can have their power interrupted if the network becomes overwhelmed.
“There is a power shortage,” said Gramlich. “Not everyone will receive the full level of service they want, at least until the system meets these new requirements.”
Burke said that because utilities need to invest in new generation capacity and grid upgrades to power data centers, they are demanding more money from data center developers so they don’t get “stranded” if the data center doesn’t meet requirements. It predicts that about 50% to 60% of data center projects will be delayed and will not be operational within the one to two years that companies are hoping for.
“It’s hard to predict how many of these data centers will actually get built at the end of the day, if they get the funding, if they meet all the different constraints, and that’s an area that’s still a question mark, but we’re seeing a big slowdown,” Burke said.