AI's rapid expansion and power needs
Artificial intelligence is growing at a pace that outstrips any previous technology, spurred by the competition among hyperscalers and a flood of users. The data centres that host AI workloads are now a focal point for community concerns and political debate, especially over rising electricity bills.
Companies such as OpenAI, Google and Meta can finance and erect AI infrastructure far more quickly than the United States can build the generation and transmission capacity required to power it.
Grid capacity struggles
Rob Gramlich, president of consulting firm Grid Strategies, told EuroHerald that the speed of tech firms contrasts sharply with the deliberate, years-long planning needed for utilities. "There is a real disconnect there," he said.
"There is a real disconnect there," Gramlich said.
He explained that the mismatch in timing means the grid may not be ready for the projected electricity demand of AI data centres. The Lawrence Berkeley National Lab estimates that these facilities could consume almost 12% of all US electricity by 2030, six times the share recorded in 2018.
The North American Electric Reliability Corporation (NERC) projects that summer peak demand in North America will rise by more than 224 GW over the next decade, a 69% increase over the growth forecast a year earlier, driven largely by new AI data centres. In the Western US grid region, planned data centres account for about 10% of demand forecasts and can reach as high as 40% in certain locales.
Industry perspectives
According to Kathryn Burke, head of US specialty energy and power growth at Marsh, access to power is now one of the top bottlenecks for data centre development. Companies are asking for gigawatts of capacity on a two-year timeline, yet a gigawatt, enough to power roughly 750,000 US homes, can take considerably longer to generate.
Berkeley Lab research shows that projects coming online in 2025 took a median of five years from grid-connection request to commercial operation. The US Department of Energy (Department of Energy) highlighted a "pressing need" for additional generation due to load growth from data centres and other factors such as electric vehicles and space heating.
Potential delays and future outlook
Utilities typically refrain from connecting new customers if they cannot guarantee reliable service. As a result, data centres may face years of waiting or provisional connections that can be interrupted when the grid is strained. "There is scarcity on the grid," Gramlich warned, adding that not every developer will receive the full level of service they desire until the system catches up.
Burke predicts that 50 to 60% of data centre projects could be delayed beyond the one-to-two-year window that developers hope for. She cautioned that the ultimate number of completed facilities remains uncertain, dependent on financing, regulatory constraints and the ability of the grid to expand.
In the coming months, utilities are expected to seek higher payments from developers to offset the risk of stranded assets, while policymakers may need to accelerate approval processes for new generation and transmission projects to avoid a widening gap between AI demand and electricity supply.