Google and Fermi Executives Clash Over Off-Grid vs. Grid-Connected Power for AI Data Centers

Google and Fermi Executives Clash Over Off-Grid vs. Grid-Connected Power for AI Data Centers
March 25, 2026
A fundamental debate over how to power the explosive growth of artificial intelligence data centers took center stage at the CERAWeek energy conference in Houston this week, pitting the off-grid ambitions of new power developers against the grid-integration strategies of established tech giants. The high-stakes discussion underscores the critical challenge of scaling energy infrastructure to meet unprecedented demand while balancing reliability, cost, and environmental goals.

The debate featured a direct exchange between Toby Neugebauer, CEO of Fermi America, and Amanda Peterson, Google’s global head of data center energy. Fermi, co-founded by former Energy Secretary Rick Perry, is pursuing a "behind-the-meter" power island model, having already secured over two gigawatts of natural gas supply for a major project in Texas. In contrast, hyperscalers like Google have generally prioritized connecting their facilities to the existing electrical grid.

Neugebauer argued that only gas and nuclear power can provide the massive, reliable baseload required for AI computing at scale. He contended that building independent, off-grid power systems shields residential customers from shouldering the costs of new infrastructure, addressing a core affordability concern. “Consumers are paying twice as much for electricity as the industrial users,” Neugebauer stated, citing U.S. Energy Information Agency data on higher distribution costs for residential and commercial customers. “That’s not sustainable.”

Google’s Peterson sharply disagreed, warning that isolated power islands could create “unintended consequences,” including higher overall costs for ratepayers. She pointed to Google’s recent deal with utility Xcel Energy, which will power a data center using a combination of wind, solar, and long-duration battery storage, as a model for grid collaboration. “When you’re building islands, you have to overbuild the system for the same amount of reliability,” Peterson explained. “That’s a lot of investment in gas that is used just a few hours a year and otherwise is sitting idle.”

Peterson further cautioned that an over-reliance on gas and nuclear presents its own bottlenecks, potentially stifling AI-driven economic growth estimated to unlock trillions in U.S. GDP this decade. She noted that new gas turbine orders face lead times stretching to 2030, a point confirmed onstage by Siemens Energy executive Karim Amin, and that new nuclear plants take a decade or more to build.

The division highlights the broader strategic fork in the road for U.S. energy policy. Fermi’s vision is exemplified by its planned “HyperGrid” complex, which could scale to 11 gigawatts using gas, solar, and a new nuclear facility. The company, which filed for an IPO last September, is exploring both traditional AP1000 and small modular nuclear reactors to meet customer demands for constant reliability.

Meanwhile, major tech firms, including Amazon, Google, Meta, Microsoft, and OpenAI, signed a “ratepayer protection pledge” at the White House in February, committing to cover their own power costs and grid upgrade expenses. Industry experts like RWE Americas' Ingmar Ritzenhofen suggest a hybrid approach using renewables, storage, and gas peakers is the fastest and most affordable path forward, warning that the industry must avoid handcuffing itself to a single solution. The outcome of this debate will shape the reliability, cost, and environmental footprint of the nation's digital infrastructure for decades.

Source: latitudemedia