SemiAnalysis: The power gap in AI data centers is widening, and reciprocating engines may become the leading technology for behind-the-meter power supply.

Independent semiconductor and AI research firm SemiAnalysis reports that the rapid growth in AI computing power demand is transforming power supply models for data centers. Reciprocating engines, historically used primarily as backup power during grid outages, are being repositioned as baseload power sources operating around the clock. This year, reciprocating engine manufacturers have signed contracts for roughly 1GW of behind-the-meter (BTM) power projects, with annual new supply volumes projected to exceed 4GW in 2027 and 2028. After modeling U.S. grid capacity, SemiAnalysis notes that existing power reserves are expected to be exhausted between 2027 and 2028, and planned additions to utility-scale power generation capacity through 2030 remain insufficient to meet the new load demand from data centers. Combining its data center model, SemiAnalysis estimates that roughly 140GW of potential data center projects have not yet finalized power supply contracts, and many of these will likely adopt behind-the-meter power models to bypass grid expansion bottlenecks. Among behind-the-meter power technologies including reciprocating engines, aeroderivative gas turbines, and fuel cells, SemiAnalysis projects reciprocating engines will capture the largest market share. The firm cites their combination of low cost, rapid deployment, modular scalability, and stronger financing capabilities as key advantages, while equipment manufacturers including Caterpillar, INNIO, and Cummins are expanding production capacity to support large-scale deployments in the coming years. As AI data centers continue to expand, on-site self-generated power is evolving from a traditional backup resource to critical energy infrastructure, and reciprocating engines are poised to become a key solution for bridging power gaps in the computing power era.