Experts Debate Whether AI Energy Demand Is a New Crisis or History Repeating
Stanford economist downplays energy demand projections as Google executive warns of unprecedented strain on the grid.
ASPEN, Colo., August 18, 2025 – Is energy demand from artificial intelligence something new—or simply a repeat of past electricity booms?
That was the subject of debate at the Technology Policy Institute in here on Monday. Experts sparred over just how much energy AI would require in the coming decade, and what should be done to build up electrical capacity.
Frank Wolak, Holbrook Working Professor of Commodity Price Studies at Stanford University, argued that AI demand fits into a long historical pattern. “Certainly my view is that [AI demand] is fundamentally not much different,” Wolak said, pointing out that U.S. electricity use grew by about 7 percent annually from 1900 to 1970. He maintained that the bigger challenge lies in “cleaning the grid,” not expanding it.
Others disagreed. Marty Hopkins, partner at Wilkinson Barker Knauer, stressed the unprecedented scale of current projections. “I think it is different than anything we’ve seen before, just because of the scale,” she said. In Texas, peak demand reached 85.5 gigawatts (GW) in August 2023, and forecasts—even after adjustments—still predict 145 GW by 2031, nearly double current levels.
Marsden Hanna, head of sustainability and energy policy at Google, echoed that concern. “Now what we’re seeing is very large potential additions to the system to meet economic demand,” he said. Hanna noted that the first five-year load forecast in 2024 “tripled from the prior year,” projecting 128 gigawatts of additional demand nationally by the decade’s end—a 15 percent increase over current infrastructure.
Wolak pushed back, cautioning that “everybody forecasts a lot of demand,” but that many projects never materialize.
Offering a middle view, Tanya Das, director of AI and energy technology policy at the Bipartisan Policy Center, said AI is only part of the story. “At the national level, it is true that data centers are not the primary source of electricity load growth,” she said.
BPC estimates that at most 25 percent of new demand by 2030 will come from data centers, with the rest driven by electric vehicles, manufacturing, and electrification of heating. But because data centers cluster in states like Virginia and Texas, Das said their impact is highly concentrated and requires strong local policies.
Panelists also weighed who should pay for costly grid upgrades. Hopkins warned that residential ratepayers could end up shouldering costs. Hanna emphasized that Google supports paying its “fair share” through long-term contracts, but cautioned against sector-specific tariffs. Das noted that state regulators ultimately decide how costs are allocated, and while consumer bills have not yet risen in Virginia, risks remain.
The discussion closed with a reminder of how quickly the issue has evolved: when Aspen first hosted a panel on AI in 2017, the word “energy” never came up, noted Scott Wallsten, President of TPI.
Member discussion