The race to find power for data centers set up a heated debate between Fermi America and Google executives on Wednesday over issues like going off grid and renewable energy’s ability to meet massive load growth.
Fermi, which was co-founded by former Energy Secretary Rick Perry, has secured more than two gigawatts of gas for a behind-the-meter power island in Texas, while major hyperscalers have so far prioritized grid connections where possible for its data centers.
Fermi CEO Toby Neugebauer, during a panel at CERAWeek in Houston, argued that only gas and nuclear can provide enough power at scale for AI computing. He added that building an off-grid site shields residential customers from the costs at a time when affordability is a major concern for Americans — and a major focus at the conference.
Google’s global head of data center energy Amanda Peterson disagreed, touting the hyperscaler’s recent deal with Xcel Energy to power a data center with a mix of wind, solar, and long-duration battery storage. She warned, in stark contrast, that islanded data centers could actually lead to “unintended consequences” including higher costs for ratepayers.
“When you’re building islands, you have to overbuild the system for the same amount of reliability,” she said. “That’s a lot of investment in gas that is used just a few hours a year and otherwise is sitting idle. That generation could be fed back into the grid at a cost completely paid for by data centers. That’s a missed opportunity.”
Off-grid data centers also won’t pay for grid upgrades that benefit the overall system, which in turn could exacerbate the affordability issue. “We are focusing too narrowly on this moment and the bottlenecks, and not thinking collectively as a system in 10 years,” she said.
The division reflects what’s at stake for the grid and the broader U.S. economy. Decisions made today will determine how clean, reliable, and affordable the energy system is for decades to come and whether the U.S. can lead in the global AI race.
“If we say we’re not going to unlock this opportunity if not for gas and nuclear, then we’re going to not unlock this opportunity,” Peterson said, referring to estimates that AI could unlock trillions of U.S. GDP by the end of the decade. “Even if we start building nuclear right now, it’s not going to show up for another 10 years. If you’re just relying on gas, we’re also screwed. All long-lead time on turbines has been scaled up through 2030.”
Karim Amin, an executive board member of Siemens Energy that is supplying Fermi with turbines, confirmed onstage that if a developer asks for turbines today, they won’t be delivered until 2030 or beyond. A 2025 report found that turbine costs are up across the board, and are unlikely to come down any time soon.
Fermi last year made headlines with its plans for a “HyperGrid” project, which could be one of world’s largest data center complexes if up to 11 GW of power is delivered through a mix of gas, solar, and a new nuclear facility. Last September, despite having yet to publicly disclose any customers, it filed for an initial public offering. Fermi hasn’t publicly disclosed any customers, yet.
Neugebauer said Fermi “would love to develop solar” given the abundant sun in Amarillo, Texas. But the data center customers he talks to are demanding baseload power so they can run with near-constant reliability. That’s why Fermi plans to start operating with a natural gas plant, and is also exploring the construction of traditional AP1000 nuclear plants and advanced small modular reactors. Fermi in October announced several deals to lock in nuclear equipment and start engineering work for an AP1000.
‘Unintended consequences’
Neugebauer said that by going behind-the-meter, Fermi can move faster because the project won’t immediately require transmission lines nor burden other ratepayers.
“Consumers are paying twice as much for electricity as the industrial users,” Neugebauer said. “That’s not sustainable, and I know it’s not the popular thing to say. I just think that’s breeding opposition.”
Neugebauer’s statement is backed up by data from the U.S. Energy Information Agency, which found that residential and commercial customers typically pay more due to higher distribution costs. Industrial customers use more electricity and can receive it at higher voltages, making delivery more efficient.
Hyperscalers including Amazon, Google, Meta, Microsoft, and OpenAi signed a “ratepayer protection pledge” at the White House in February, stating they will negotiate rate structures with utilities to cover their electricity costs, either build or buy new generation for their data centers, and pay for grid upgrades like transmission.
Ingmar Ritzenhofen, chief commercial officer of RWE Americas, said deploying a mix of solar, wind, storage, and gas peaker plants is the fastest way to deploy power in the U.S., and to do so affordably.
“The burden that you put on the ratepayer is minimized if you put that combination work,” he said. ”However, I think right now we’re still in a world where we tend to handcuff ourselves.”
