The power grid queue isn't moving fast enough for the AI industry. Interconnection wait times in the United States now average more than four years, and the hyperscalers building the next generation of GPU clusters can't afford to wait. So they are doing something that energy analysts would have found strange two years ago: pairing battery storage directly with natural gas turbines, building their own off-grid power plants, and skipping the utility entirely. BloombergNEF has tracked 4.9 gigawatts of battery energy storage announcements co-located with fossil fuel generation at data centers worldwide. That figure represents roughly 32 percent of all announced global on-site data center battery capacity. The projects vary in scale and configuration, but the logic is consistent across all of them: gas provides the baseload, batteries handle the spikes, and together they deliver the kind of uninterrupted uptime that a server hall running continuous AI inference cannot live without. This is not a clean energy story. It is a reliability story, and the battery industry is finding itself at the center of it. The New Supplier Math Data center storage looks different from ordinary grid storage procurement. A utility battery usually competes on energy arbitrage, capacity value, and ancillary services. A data center battery has to satisfy uptime requirements first, then economics. That pushes buyers toward bankable suppliers, redundant controls, fast service contracts, and conservative warranty terms. Short duration Lithium-ion systems cover seconds-to-hours events, smooth gas turbine output, and help data centers manage peak grid draw. Multi-day backup Iron-air and other long-duration chemistries target extended outages, renewable firming, and campuses that cannot rely on diesel alone. Power electronics Inverters, switchgear, and control software are becoming as important as the cells because AI loads move faster than conventional industrial loads. That mix favors companies with complete systems rather than cell supply alone. Fluence, Tesla, Wärtsilä, Sungrow, Powin, and other integrators can package batteries with dispatch software and grid services expertise. Cell makers still benefit, but the margin pool moves toward whoever can guarantee a working power plant around the cells. The emissions debate also gets sharper. Batteries can make gas turbines cleaner by reducing inefficient ramping, but they can also make new gas capacity easier to justify. That means battery suppliers are being pulled into a politically sensitive role: enabling AI growth while utilities, regulators, and hyperscalers argue over how much new fossil generation is acceptable. May 19 Update: Battery Suppliers Are Now Reporting the Data Center Pipeline Directly The story moved again in mid-May. Battery storage suppliers are no longer treating AI power demand as a future category. They are quoting active pipelines tied to data centers, grid interconnection delays, and on-site generation projects. Fluence has been linked to more than 30 GWh of global data-center-related battery storage opportunities, with a large share in the United States. Tesla's Megapack deployments remain the visible benchmark, helped by sales into xAI's Memphis power buildout. Calibrant Energy's 31 MW and 62 MWh system for an Aligned data center campus in the Pacific Northwest shows that the market is not limited to mega-campuses. The near-term constraint is no longer whether batteries belong in the data center stack. It is whether suppliers can deliver enough cells, inverters, transformers, and software controls fast enough to match AI developers' speed-to-power demands. Why the Hybrid Model Makes Sense for AI An AI data center running large-scale model training pulls power in sharp, unpredictable bursts. A GPU cluster that idles at 60 percent load can spike to 100 percent in seconds when a new training job starts. Gas turbines are efficient at producing steady output but wear out faster when they ramp up and down constantly. Batteries solve that problem by absorbing the swings, letting turbines run at a stable operating point while the battery bank charges and discharges in response to whatever the servers demand. The uptime requirement matters too. Hyperscalers typically specify 99.999 percent availability for critical compute infrastructure, which translates to roughly five minutes of downtime per year. Gas-plus-battery configurations can meet that target without depending on a utility grid that may itself be unreliable during peak demand periods. When the grid fails, the turbine keeps spinning and the battery provides the bridge. Batteries projected to enable 9.8 gigawatt-hours of gas generation at data centers through 2030, according to BloombergNEF's tracking. That number is likely conservative given the pace of announcements in the first quarter of 2026 alone. The Projects Taking Shape The xAI Colossus supercomputer in Memphis, Tennessee is the most visible example. Elon Musk's AI company partnered with Tesla to install Megapack battery systems alongside gas turbines in what amounts to a 1.2 gigawatt off-grid power plant. The facility can operate independently of the local utility, a capability that has become a selling point rather than a quirk as grid congestion worsens in major metro areas. In West Texas, Pacifico Energy's GW Ranch project takes the concept to a different scale entirely. The plan calls for 1.8 gigawatts of battery storage paired with 7.65 gigawatts of gas-fired generation, creating an off-grid power hub purpose-built to serve a massive data center campus. Texas has the available land and gas infrastructure to make this viable, and the ERCOT grid's independent structure means there are fewer regulatory complications for behind-the-meter generation at this scale. Williams Companies, a natural gas infrastructure firm, has moved into data center power services by pairing Tesla battery systems with gas plants at multiple US locations. The company is essentially offering data center operators a turnkey power solution that doesn't require them to navigate utility interconnection queues. On the utility side, NIPSCO Generation is building two 1.3 gigawatt gas plants alongside 400 megawatts of battery storage for Amazon in northern Indiana. DTE Electric is deploying six energy storage systems to complement a 1.4 gigawatt data center serving Oracle in Michigan. These are utility-scale projects, but they are designed to serve a single customer's power needs rather than the broader grid. What This Means for Battery Suppliers Tesla's Megapack business is the clearest beneficiary of this trend. The product's capacity, form factor, and existing track record with hyperscalers gives it an advantage in data center sales cycles that prioritize proven reliability over cost per kilowatt-hour. Tesla's energy storage revenue grew significantly in 2025, and the data center pipeline suggests that growth continues regardless of how EV demand trends develop. Fluence Energy, the Siemens and AES joint venture focused on utility-scale storage, is in active talks with gas generation companies about data center power partnerships. The company's software stack for optimizing battery dispatch could be well-suited to the gas-plus-battery use case, where the dispatch logic needs to balance turbine efficiency, battery state of charge, and AI workload demand simultaneously. On the equipment side, GE Vernova has introduced new products designed specifically for data center gas-plus-battery configurations, and Caterpillar has expanded its industrial power systems offerings to address the same market. The convergence of power equipment makers and battery suppliers around data center customers reflects how large the opportunity has become. For the broader battery storage market, data center demand is becoming a meaningful counterweight to slower-than-expected EV growth. LG Energy Solution explicitly cited AI data center demand when e