Hyperscalers are turning to renewables to power AI data centers

In This Article:

According to research from the University of California, the amount of power needed for ChatGPT to generate a single 100-word email is 0.14 kilowatt-hours, roughly equivalent to running 14 LED lightbulbs for an hour. Angie Gildea, KPMG U.S. Energy Leader, tells Catalysts Hosts Seana Smith and Madison Mills that both hyperscalers and energy providers are exploring renewables like wind and solar to fulfill the robust demand generated by AI data centers. "There's about 8,000 data centers worldwide today, and that number is growing every day."

Tech companies are “willing to pay extra, additional money to get that renewable power,” Gildea says. “But speed to market is also important, and they need to have reliability, and they need to have redundancy.”

And while Meta (META) is renewing its call for a nuclear partner to expand its capabilities in the 2030s, KPMG found that just 16% of hyperscalers are doing the same. According to Gildea, “there's also some uncertainties around nuclear, particularly as it relates to public perception, some regulatory permitting challenges, and it does take several years to get those nuclear facilities up online.”

To watch more expert insights and analysis on the latest market action, check out more Catalysts here.

This post was written by Kayla Hawkins.