dotshock/Shutterstock Save for later Print Download Share LinkedIn Twitter The outlook for the artificial intelligence-led data center boom in the US has gotten a reality check as growth projections run headlong into power grid constraints, high costs and the limitations of deploying new energy capacity. While data centers surely will be a source of load growth in the coming years, the size of the opportunity for energy suppliers remains uncertain. Previous projections had cast data centers as an emerging sector with a near-unquenchable thirst for energy. Now the conversation is turning more toward the need for data centers to exercise flexibility in their demand for power from a grid system currently incapable of adding capacity as quickly as hyperscalers had hoped. Utilities "can't build data centers to drive America forward fast enough, because they can't get enough power," US Energy Secretary Chris Wright told a Department of Energy conference in San Diego this week. "We've got to unleash those constraints." He called for getting "more power out of our existing grid today" and building new infrastructure to "bring massively more power generation on." The constraints facing power developers are many, ranging from permitting and labor challenges to a multiyear backlog of gas turbine orders. "You don't want to leave any electrons on the table that aren't being utilized from what's already been built," warned Scott Strazik, CEO of turbine supplier GE Vernova. As developers race to build out more gas-fired power — seen by some as the most reliable option for what hyperscalers require — prices to do so are also on the rise. The average cost of building a gas plant increased from about $1,000 per kilowatt a decade ago to more than $3,000/kW today, Matt O'Connor, a managing director at investment firm Carlyle, told last month's CERAWeek by S&P Global conference.
