Researchers say batteries should come before gas for AI power demand

A Georgetown and Duke report urges states to steer data centers toward battery storage, citing stranded-asset risk and cost advantages over gas.

Researchers at Georgetown University and Duke University have published a report arguing that states should push utilities and tech firms to choose battery storage over fossil fuels for data centers wherever possible. The backdrop is a gas buildout: U.S. utilities are planning more than 60 GW of new gas plants to serve data center demand, and about 90 GW of behind-the-meter gas generators for data centers have been proposed nationwide. Much of the AI industry favors gas because it delivers round-the-clock power in one package.

The report's case rests on risk and cost. If data center projects are cancelled or close early, other utility customers could be left paying for overbuilt gas capacity, a stranded-asset risk. Batteries are modular and can be scaled up or down if demand changes. The report says they are much cheaper than round-the-clock gas when a few hours of stored energy can meet near-term needs, and they can help avoid costly grid buildouts and cushion the grid against data centers' large, fast-changing loads.

This matters for energy abundance because the choice made now decides whether AI demand locks in more fossil capacity, with its costs spread across ratepayers, or pulls forward cheaper storage. BloombergNEF forecasts U.S. storage deployment will pass 11,000 GWh by 2036, a 30% increase over its forecast from six months earlier, and data center demand is driving much of that growth.

The report points to policy levers. States that regulate utilities directly can adopt "bring your own new clean energy" tariffs, which Google has signed onto in Nevada and Minnesota. In restructured markets, it suggests storage mandates, as in Illinois, Maryland and New Jersey, plus state planning with regional grid operators.

The caveats are real. Lithium-ion batteries cannot provide round-the-clock power, so they can meet large portions of data center needs, not all. The cost advantage applies where hours of storage solve near-term needs. The conclusions are the authors' own analysis, reported by Canary Media, a clean-energy outlet.

Source: Canary Media