Svmuu News: SemiAnalysis posted on X stating that reciprocating engines—traditional backup power equipment for data centers—are shifting from an “emergency backup” role to becoming a primary, round-the-clock power source to meet the massive electricity demand driven by the expansion of AI computing power.
SemiAnalysis noted that in the past, data centers typically activated on-site reciprocating engine backup power only during grid outages, resulting in annual operating times of just a few hours. However, as grid capacity becomes constrained, an increasing number of data centers are repurposing these units as prime power sources to ensure continuous power supply.
According to SemiAnalysis’ energy model projections, reciprocating engine manufacturers—including Caterpillar, INNIO, and Cummins—are expected to have secured approximately 1 GW in orders for behind-the-meter (BTM) projects this year; the volume of related orders in both 2027 and 2028 is projected to exceed 4 GW.
SemiAnalysis notes that the current order volume remains very limited compared to future potential demand. After modeling U.S. grid constraints, the firm found that existing power supply reserves are projected to be exhausted between 2027 and 2028, and that planned new utility-scale generation capacity through 2030 will be insufficient to meet the growing demand from data centers.
Based on its data center model, SemiAnalysis estimates that there are still approximately 140 GW of identified data center projects that have not yet signed power purchase agreements and may adopt a behind-the-meter power supply model.
Among the technologies available for off-grid power supply—including reciprocating engines, aeroderivative gas turbines, and fuel cells—SemiAnalysis expects reciprocating engines to account for the largest share.
The firm believes that reciprocating engines offer the optimal balance between cost and deployment speed, while their modular design supports phased expansion of data centers; compared to fuel cells, they demonstrate greater commercial maturity and financing feasibility. As manufacturers expand production capacity in the coming years, reciprocating engines are expected to become a critical infrastructure component for resolving power bottlenecks in AI data centers.
SemiAnalysis notes that the growth in AI computing power is driving data centers to shift from relying solely on traditional grid power to a hybrid “grid + on-site power” model, and that energy infrastructure is becoming the new core constraint on the expansion of the AI industry.