US power companies are locked in a scramble for the heavy electrical gear that keeps the lights on, as a wave of artificial-intelligence data centres pushes demand for transformers, turbines and switchgear well beyond what factories can supply.
Lead times once measured in months now run into years, and developers are fighting over the same scarce production slots as the build-out strains data-centre power and reshapes local grids.
The tightest bottleneck is in large power transformers, the substation workhorses that raise and lower voltage across the network. Delivery windows that sat at roughly 24 to 30 months before the boom have stretched to three to five years for the biggest units, industry trackers say, with average US lead times now running near 128 weeks.
The strain does not stop there. Switchgear, the protective equipment that isolates faults and routes power, is effectively spoken for through 2028 at some suppliers, and analysts warn that medium-voltage kit and generator step-up units face similar multi-year queues.
Turbines are scarcer still. GE Vernova’s gas-turbine backlog hit 100 gigawatts in the first quarter of 2026, and the company expects to reach at least 110GW of orders and slot reservations by year-end, with delivery windows at the major makers booked years out and some frames sold into the next decade.
Siemens Energy has reported a record order backlog of roughly €136bn, and rivals including Mitsubishi are carrying books that stretch about five years. The surge, executives say, is being driven above all by data centres racing to lock in power before they can switch anything on.
Part of the problem sits upstream. Transformers depend on grain-oriented electrical steel and skilled assembly labour, both in short supply, so production has been slow to respond even as orders pile up.
The numbers behind the crunch are stark. Demand for generator step-up transformers rose 274% between 2019 and 2025, and substation transformer demand climbed 116% over the same period, according to industry data, as AI computing, electrified transport and a manufacturing revival all pulled on the grid at once.
One market forecast has US data-centre electrical-equipment spending jumping from around $20bn in 2025 to $65bn by 2030. The bottleneck is already bending build schedules, with some analyses suggesting that up to half of planned US data-centre projects have slipped or been cancelled, partly because operators cannot secure gear and grid connections in time.
For utilities, the mismatch is forcing a change in habit. Some are placing orders speculatively and paying to reserve factory slots long before a project is approved, reversing the just-in-time buying that shaped grid procurement for decades.
Analysts caution that part of the backlog may be inflated by double-ordering, as buyers book capacity with several suppliers at once, making the true depth of the shortage hard to read.
Regulators have taken notice. The Federal Energy Regulatory Commission has moved to fast-track data-centre connections, but a faster queue does not conjure a transformer out of thin air, and the underlying shortage of hardware and generating capacity remains.
The delays feed straight into what households pay. Utilities near the biggest hubs have warned of rising bills, and connection queues in several major markets now stretch beyond four years, with scarcity pushing up equipment prices as well as wait times.
To close the gap, manufacturers are pouring money into domestic capacity. Hitachi Energy has committed more than $1bn to US production, including a new plant in South Boston, Virginia, due online in 2028, while Siemens is expanding transformer manufacturing in North Carolina.
Even so, new factories take years to build and certify, and few in the industry expect the shortage to ease before the end of the decade. For now, securing a transformer or a turbine slot has become almost as central to an AI project as the chips inside it.
Get the TNW newsletter
Get the most important tech news in your inbox each week.