For most of the last decade, the slowest step in standing up new dispatchable generation in the US was the interconnection study queue. A combined-cycle gas plant developer could line up land, gas supply, and a power purchase agreement in roughly the same window it took the RTO to clear the cluster studies, sign the interconnection agreement, and assign the network upgrade cost. The queue reforms FERC pushed through Order 2023 in 2023, and the cluster study cycle reforms PJM, MISO, and CAISO implemented across 2024 and 2025, were aimed at cutting that timeline.
The reforms have worked, mostly. Cluster study cycles that used to drag four to six years are now closing in 24 to 36 months for projects that withdraw on schedule. The interconnection queue is no longer the binding constraint. The new binding constraint is sitting one level upstream of the project, in the order books of the three OEMs that make heavy-duty gas turbines: GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries.
What the order books actually show
GE Vernova disclosed in its Q1 2026 earnings call that its heavy-duty gas turbine backlog had reached $42 billion, a roughly four-times increase over the equivalent backlog two years earlier. Slot availability for new orders of HA-class machines (the 7HA.03 and 9HA.02 frames that anchor most new US combined-cycle builds) is now quoted into 2029 and 2030 depending on configuration. Reservation agreements, in which a buyer pays a non-refundable fee to hold a delivery slot before signing a full equipment purchase order, are the standard practice on every recent transaction.
Siemens Energy reported a gas services and large gas turbine backlog north of 130 billion euros at the end of fiscal 2025, with the company describing the order intake as having been pulled forward by data center load growth in North America and Asia. The HL-class machines (the 9000HL and 7000HL frames) carry similar 2029 to 2030 lead times.
Mitsubishi Heavy Industries has the JAC and M501JAC frames serving the same large combined-cycle market. The MHI Power Systems segment reported a roughly tripling of orders in the year through March 2026 compared with two years earlier, and has explicitly described the backlog as constraining acceptance of new orders in the heavy-duty class.
The aggregate effect across the three OEMs is that the global heavy-duty gas turbine order book is now booked through 2029 in core configurations, and any new buyer joining the queue in mid-2026 is realistically signing for first power around 2030.
Why the backlogs grew
The cleanest single-line explanation is data center load growth, but that flattens a more layered story.
Through the early 2020s, the heavy-duty gas turbine market was structurally oversupplied. The OEMs had built capacity for a forecast of mid-cycle utility-scale combined-cycle build-outs that did not arrive. Coal retirements were happening, but most of the replacement megawatts were going to wind, solar, and storage, not to new gas. The OEMs trimmed plant capacity, deferred capacity expansions, and ran their factories at reduced utilization. GE separated its energy business into GE Vernova in part because the underlying gas turbine business needed a different operating cadence than the rest of the GE portfolio.
The order intake started turning in 2023 as the first wave of data center hyperscaler load grew visible in utility planning. Through 2024 and 2025, the order book filled in two distinct layers. The first layer was utility-scale combined-cycle additions tied to data center load: large vertically integrated utilities in the Southeast, ERCOT-area independent power producers, and PJM-area independent power producers signing for one to four units each. The second layer was data center colocation projects in which the hyperscaler or developer was signing directly with the OEM for behind-the-meter or front-of-the-meter generation, often paired with a power purchase agreement with the local utility.
The combination loaded the order books faster than the OEMs’ planned capacity ramps could absorb. The OEMs have responded with capacity additions, but heavy-duty gas turbine manufacturing capacity does not expand quickly. A new turbine assembly facility takes three to five years from announcement to first unit out the door. Supplier capacity for hot section castings, single-crystal blades, and large forgings is the binding constraint inside the OEMs’ own constraint. GE Vernova has publicly committed to expanding its Greenville, South Carolina assembly capacity. Siemens Energy has announced expansions in Berlin and in Charlotte. Mitsubishi Heavy has announced capacity additions in Japan and is studying additional US capacity. None of the announced expansions deliver meaningful incremental output before 2028.
How utilities are responding
The utility planning shops have not stopped writing IRPs that include new combined-cycle gas plants. They have changed how those plants are described in the integrated resource plan, how the equipment procurement is sequenced relative to the rest of the project, and what the contingency planning around delivery risk looks like.
The new pattern, visible in IRP filings across Duke Energy, Southern Company, Dominion Energy, NextEra subsidiaries, Entergy, and TVA over the last 18 months, is that the IRP identifies a 2028 or 2029 in-service combined-cycle plant. The equipment procurement narrative is then separated from the project narrative, with the IRP describing a turbine procurement either already executed under reservation or actively in negotiation, with the project schedule contingent on the turbine delivery date.
A handful of utilities have started talking publicly about a reservation strategy in which the utility holds turbine slots speculatively, ahead of any specific project assignment, to maintain optionality for the load growth they are forecasting but have not yet contracted. The reservation fees are a small fraction of total project cost but a large absolute number, and the regulatory treatment of speculative reservation fees has become a topic in state utility commission proceedings in Virginia, Georgia, Florida, and Texas.
The independent power producers running merchant capacity in PJM and ERCOT have moved faster than the regulated utilities on reservation strategy, in part because they do not have to defend the reservation cost in front of a state commission. The visible projects from Vistra, Constellation, NRG, Calpine, and Talen Energy include reservation-based turbine procurements that were committed in 2024 and early 2025, before the slot availability moved fully out to 2029 to 2030.
The asymmetry between IPP procurement speed and utility procurement speed is showing up in capacity market outcomes. The PJM capacity auction in mid-2025 for the 2026 to 2027 delivery year cleared at the regional cap on most zones and revealed a structural shortage of new entrants. The auction result was as much a statement about gas turbine procurement timing as it was a statement about generation economics. New entry in the 2027 to 2029 window is not constrained by the auction clearing price. It is constrained by whether the developer holds a turbine reservation that delivers in time.
What this means for the reliability window
The earlier posts in this series identified four levers for the 2027 to 2029 reliability window: storage build-out, coal life-extension, demand-side reliability, and grid-enhancing technologies as a delivery mechanism. The turbine OEM constraint is upstream of all four in different ways.
Storage is the lever least affected by the turbine constraint. Lithium-ion battery storage has different supply chain dynamics, with cell capacity in China and Korea responsive to demand on a 12 to 24 month timeline. The storage build-out is constrained by interconnection queue timing and by transformer and switchgear supply, not by anything that resembles the turbine constraint. The 2025 to 2027 storage additions in CAISO and ERCOT have continued to scale roughly as planned.
Coal life-extension is the lever most directly tied to the turbine constraint. Every gigawatt of coal capacity kept online into the 2027 to 2029 window is a gigawatt of new gas combined-cycle capacity that does not have to arrive on schedule. The utilities that have announced coal retirement deferrals in the last 18 months, including the larger MISO and SPP coal fleets, have in several cases cited replacement generation timing as the rationale. The replacement generation timing question is, beneath the surface, a turbine procurement question.
Demand-side reliability and the data center curtailment contracts that are now being negotiated in MISO, PJM, and ERCOT are also coupled to the turbine constraint. The economics of paying a hyperscaler to curtail at peak hours look different in a world where the replacement generation can be built in 30 months than in a world where it cannot be built before 2030. The reliability planners writing the curtailment contracts are doing so against an assumed generation backstop. The generation backstop is the constraint that is moving.
Grid-enhancing technologies, the topic of the previous post in this series, sit one layer down from the turbine constraint and partially compensate for it. The GETs deployment ramp delivers megawatts of incremental transfer capability without requiring new generation. Where the local capacity need is satisfiable through better utilization of the existing generation stack, GETs are a partial substitute. Where the local capacity need is satisfiable only through new dispatchable generation, GETs cannot substitute for the turbine.
The honest summary is that the OEM constraint is now the binding constraint, with the IPPs better positioned than the regulated utilities, and the coal life-extension decisions being made over the next 12 to 18 months effectively setting the floor for how much risk the reliability window carries. The IRPs that were written in 2024 assuming 2028 in-service dates for new combined-cycle capacity are quietly being rewritten in 2026 around the turbine delivery date. The capacity auctions are pricing the rewrite. The state utility commissions are beginning to ask about the reservation strategy. The next 18 months of OEM capacity expansion announcements and reservation policy decisions are where the 2027 to 2029 window gets settled.