Omdia sees data centers deploying 35 GW of behind-the-meter power by 2030
Summary
Data center developers are increasingly generating their own electricity on site rather than waiting for utility connections, a shift that is benefiting suppliers of gas turbines and battery systems, Power Engineering reports. Research firm Omdia projects that data centers could install 35 GW of behind-the-meter (BTM) power by 2030.
The driver is a lack of grid capacity in many locations, combined with interconnection queues that keep getting longer. Developers would generally rather purchase power from utilities or independent power producers (IPPs). With billions of dollars of AI infrastructure on the line, however, many are unwilling to sit out the delay. Omdia analyst Shen Wang said large AI data centers will proceed "one way or another" as demand for AI grows.
Aligned Data Centers, a developer and operator, is cited as an example of how large the buildout has become. The company says it has nearly 100 facilities under construction and about 50 sites already operating. The source excerpt ends before giving further detail, and it does not break down the 35 GW projection by technology or region.
Why it matters
A 35 GW projection for on-site generation signals that grid access, not land or customer demand, is now setting the pace for many AI campuses. For utilities and IPPs, it means a share of large-load demand may bypass traditional supply arrangements. For developers and investors, it shifts capital and procurement toward turbines, batteries and fuel supply, and puts equipment vendors in a stronger position as builders try to bring capacity online ahead of interconnection dates.
Expert Lens
On-site generation solves the queue problem but swaps it for others: turbine delivery slots, gas pipeline capacity, air permits and the cost of carrying generation assets that a utility would otherwise own. These plants are likely to be designed as bridges until a grid connection arrives, which makes the eventual handoff, and whether the on-site assets become stranded or serve as backup, a central question in project finance. Batteries may increasingly be paired with turbines to handle the sharp load swings of AI training clusters. Lenders will probably price these projects differently from grid-fed campuses because fuel and equipment risk sits with the developer. The next signal to watch is whether utilities and regulators begin offering faster or conditional interconnection terms to win back large loads that are otherwise building their own power.
Sources
This brief was drafted with AI assistance from public sources; every number is attributed to the sources above. See the AI-assisted Disclosure and Methodology.