The race to power artificial intelligence has data centers chasing every available energy solution. One emerging approach pairs batteries — long championed as a renewable energy tool — with fossil fuel generation.
BloombergNEF has tracked 4.9 gigawatts of energy storage projects co-located with on-site fossil fuel generation at data centers. That accounts for roughly 32 percent of announced global on-site data center battery capacity. Projects include some of the largest AI complexes underway, such as Elon Musk’s Colossus supercomputer in Memphis, Tennessee. Manufacturers have taken notice: Caterpillar Inc. and GE Vernova Inc. have both announced products pairing energy storage with gas generation.
Batteries typically absorb excess solar and wind power and discharge it when conditions are unfavorable for renewable generation. But declining battery costs have opened a new use case — pairing storage with natural gas to deliver reliable power on demand.
Gas turbines can provide continuous electricity, but many do not operate continuously and cannot ramp quickly enough to match computing demand spikes. Batteries fill those gaps by releasing power instantly. They also reduce wear on gas turbines that are not designed for frequent cycling.
“I assumed batteries would be a tool for decarbonization,” said Michael Thomas, founder of clean energy research firm Cleanview, who has tracked the pairing trend. “What we are learning in this new AI era is that they can also be used as a tool for fossil fuel power because their technological advantages make it possible to build and operate an off-grid power plant.”
Grid connections now average four years to arrange, pushing data center operators toward self-generated power as a bridge. Gas generators paired with batteries serve that need while utilities work through interconnection queues that stretch years long.
Batteries also smooth the sharp demand surges that come with model training and other compute-intensive workloads. BNEF projects energy storage will support 9.8 gigawatt-hours of gas generation at data centers through 2030.
Major Projects Taking Shape
The xAI Colossus facility in Memphis is installing rows of Tesla Megapacks alongside gas turbines as part of a 1.2 gigawatt off-grid power plant. In West Texas, Pacifico Energy’s GW Ranch data center will host 1.8 gigawatts of battery storage alongside 7.65 gigawatts of gas generation — the project recently received the largest air pollution permit ever issued in the United States. Musk’s Memphis complex has faced multiple lawsuits alleging the gas turbines are worsening air quality in surrounding communities.
Williams Cos., a natural gas pipeline operator, plans to deploy Tesla batteries alongside gas plants it is constructing for several data center clients. “Batteries really help support the turbines and give us the 99.999 percent reliability,” said Rob Wingo, executive vice president, at the S&P Global Power Markets Conference.
Utilities are building storage systems on the grid as well. NIPSCO in northern Indiana is constructing two 1.3 gigawatt gas plants alongside 400 megawatts of energy storage to serve Amazon data centers. In Michigan, DTE Electric received regulatory approval to build six storage systems supporting a 1.4 gigawatt Oracle data center — a move that increases total generating capacity by 25 percent.
Fluence Energy, a storage provider, is in negotiations with natural gas companies to deliver batteries that can bring data centers online before turbines are even installed. “We are seeing massive demand coming out of the hyperscalers and data center operators,” said Jeff Monday, chief growth officer.
The Tradeoffs
Pairing batteries with gas extends the operational life of fossil fuel infrastructure — and delays the clean energy transition. It also increases emissions and contributes to local air pollution.
“There is nothing about batteries that are inherently clean,” Thomas said. “Batteries are just a technology.” Whether that technology accelerates decarbonization or prolongs dependence on fossil fuels depends entirely on how it is deployed.