Texas faces new risks this summer as an increase in artificial intelligence and computer data centers further strain the state’s power grid.
Since winter storm Uri and the subsequent blackout in 2021, state officials have worked to improve the power grid, including developing new power plants and upgrading transmission lines. However, amid rising temperatures, aging power plants and the challenges of integrating renewable energy, the state’s power grid remains fragile.
Dan Stanzione, the associate vice president for research and director of the Texas Advanced Computing Center, said artificial intelligence and cryptocurrency operations require significantly more computational power compared to traditional data centers.
High-performance computing centers have been handling AI and cryptocurrency operations for a long time, Stanzione said. They were designed to handle much more power than standard IT centers, which use about 4,000 to 5,000 watts per rack, and commonly store 42 servers. However, AI data centers use up to 100 kilowatts per cabinet, similar to a rack.
“Worldwide, not just in Texas, … there’s this explosion of demand for AI,” Stanzione said. “There are ways to improve AI to use less power, and we are certainly working on those, but at the same time, decarbonizing the power grid is really the most important thing.”
A recent report by the American Council for an Energy-Efficient Economy suggests Texas could strengthen its power grid by making appliances and buildings more energy-efficient, lowering energy needs for businesses and homes. This approach could cut energy use during peak summer and winter months, making the grid less vulnerable to outages.
Ross Baldick, an emeritus electrical and computer engineering professor, said discussions on energy demand are increasing, as well as concerns over the impact of new AI standards in Texas that could add excessive strain to the grid. These standards could rapidly increase consumption within ERCOT, posing significant challenges to the grid’s stability and consumer energy costs.
“We’ve got an amount of growth that could appear over a very short time, that would be as much or more as the natural demand growth in our cost over that same period from other consumers,” Baldick said. “It’s not unprecedented, but I think it’s pretty unusual, particularly if we get multiples of these in a relatively short time.”
Currently, Texas has the highest energy consumption in the U.S. and Stanzione said there is a critical need for regulatory focus on grid investment and ethical AI standards.
“A more robust investment in the power grid would not just help AI but renewables (too),” Stanzione said. “Having intermittent, small sources (and) rooftop solar everywhere could all be integrated a lot better if we could invest more heavily in the grid itself, rather than just in the power plants.”
