HOUSTON, March 23, 2026, 11:59 CDT
Google warned on Monday that the United States is not building electricity supply fast enough for the next wave of AI data centers, underscoring how power is starting to rival chips as the industry’s main physical constraint. “We are concerned that we are not full throttle on energy,” Alphabet and Google President Ruth Porat said at CERAWeek in Houston. 1
The problem matters now because big cloud companies are adding computing capacity faster than grids, substations and new generation can be built. The U.S. Energy Information Administration said electricity demand has been rising 1.7% a year since 2020 after more than a decade of flat use, driven chiefly by large computing facilities, and expects load to rise another 1.9% in 2026 and 2.5% in 2027. Google last week said it can make up to 1 gigawatt of data-center load available for demand response — temporary cutbacks when the grid is tight — roughly enough to power 750,000 homes, and Michael Terrell, the company’s head of advanced energy, called that “a really important tool for meeting future demand.” 2
Nvidia and startup Emerald AI said on Monday they are working with AES, Constellation, Invenergy, NextEra, Nscale and Vistra on “flexible” AI factories. The design leans on co-located generation and storage, meaning power built on or beside the data-center site, to speed grid connections; Nvidia CEO Jensen Huang said energy, compute, networking and cooling must be designed together, while Constellation CEO Joe Dominguez said, “We don’t have a supply problem — we have a peak problem.” 3
The scramble is already changing project design. On March 20, the U.S. Department of Energy said SoftBank and AEP Ohio plan to redevelop federal land in southern Ohio with a 10-gigawatt data-center campus and 10 GW of new power generation, including 9.2 GW of gas, in a buildout that officials said is expected to start this year. 4
Washington has been nudging tech groups in the same direction. A White House fact sheet published on March 4 said Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI agreed to build, bring or buy new power for their data centers and pay for the delivery upgrades needed to serve them, instead of shifting those costs to households. 5
Google is not alone. Amazon Web Services told Reuters in February that securing grid connections has become one of the biggest factors in where it expands in Europe, and Pamela MacDougall, AWS’s head of energy markets and regulation in EMEA, said connecting to the transmission network there can take up to seven years, versus roughly two years to develop a data center. 6
By late morning in New York, Alphabet was up $0.61 at $301.61 and Nvidia had gained $2.87 to $175.57. Power names tied to the AI buildout were stronger: Constellation added $7.68 to $289.67, Vistra rose $5.73 to $151.75 and Talen Energy climbed $13.09 to $316.06.
But the bottleneck is not easily fixed. Analysts said PJM, the largest U.S. grid operator, could push a burst of direct power deals with data-center owners under its February proposal for large new users, yet James West of Melius Research and Rick Pederson of Bow River Capital both cautioned that permitting, state approvals and long interconnection queues could still choke off any “flurry” of transactions. 7
The pressure is not likely to ease. The Electric Power Research Institute estimates data centers could account for 9% of U.S. electricity demand by 2030, up from 4% in 2025, while S&P Global said last week that worldwide data-center power demand could grow 12% to 16% a year from 2025 through 2030. 8
