Fossil-fuel power plants planned in the United States could account for about 70% of the new capacity that begins operating in 2030, if they come online as scheduled. This article, however, argues that fossil plants slated for 2030 face a significant risk of starting operations later than planned. On October 13, 2026, the US environmental NGOs Environmental Defense Fund (EDF) and Atlas Public Policy reported that fossil capacity scheduled for completion in 2026–30 has reached 68 GW, up 45% from 47 GW six months earlier. Meanwhile, GridLab's September 2025 tally found that most new gas combined-cycle plants cannot begin operating until 2030–31 at the earliest, and that construction costs for some projects have risen above $2,000 per kW.
What Is the "70% of New Capacity" Figure Based On?
According to the EDF and Atlas Public Policy report, about 285 GW of generating capacity was under development in the US (planned or under construction) as of the second quarter of 2026. Planned fossil capacity has reached its highest level since 2016 and makes up roughly 30% of all capacity under development. Fossil capacity expected to be completed in 2026–30 totals about 68 GW, nearly all of it natural gas.
The report points to policy as one factor behind the tilt toward fossil plants. Some regions have introduced rules that fix the share of each generation type that counts toward reliable supply, and utilities are citing those rules to justify more gas plants. In addition, the US Department of Energy has used emergency orders to direct coal plants totaling 3.2 GW to keep operating.
It is important to understand what the "about 70%" figure means. It is not fossil's share of all capacity under development. It is fossil's share of the rated capacity of new facilities that would begin operating in 2030 if everything proceeds as planned. Fossil's share of all capacity under development is about 30%. According to the report, fossil's share of newly operating capacity was about 70% in 2010 and stayed above 30% every year from 2010 to 2019, but fell to 5% in 2024. In other words, a 70% share in 2030 would mean a return to the 2010 level if the plans materialize. Note, however, that the 2010 figure is an actual result, while the 2030 figure is a projection based on plans.
The tally uses the planned operating dates that developers reported to the US Energy Information Administration (EIA) in its monthly dataset, EIA-860M. The report itself acknowledges that this data cannot show whether projects scheduled to start in 2030 have already secured the gas turbines they need. Behind-the-meter captive generation not connected to the grid is excluded, and whether the plans are realized also depends on turbine supply constraints and on whether power demand from large customers grows as forecast. According to the EIA, new capacity scheduled for 2026 is a record 86 GW, of which natural gas accounts for only 6.3 GW. The periods differ from the 68 GW of fossil capacity covering 2026–30, so the two cannot be compared directly, but the figures hint that new fossil capacity may be concentrated in the late 2020s.
Mitsubishi Heavy's New Orders Deliver in 2028–30, and Construction Costs Top $2,000 per kW
Securing a large gas turbine does not mean a plant can start operating right away. Manufacturers sometimes sign reservation agreements to hold future production slots ahead of firm orders. A project then proceeds through firm ordering, manufacturing, delivery, installation, and plant construction before it finally begins operating. The year a turbine is delivered therefore differs from the year the plant starts up, and startup comes after delivery.
GridLab's September 2025 tally found that most new gas combined-cycle plants cannot begin operating until 2030–31 at the earliest. The year 2030, when EDF and Atlas Public Policy project fossil plants making up about 70% of new capacity, is thus the earliest point at which such plants can start up.
Mitsubishi Heavy Industries says it will deliver the 10 large gas turbines it ordered in April–June 2026 (4 for the US and 6 for Japan) roughly in 2028–30. For gas turbine combined cycle (GTCC) projects still under negotiation, it gives delivery timing of 2030–31. Because installation and construction are needed after delivery, projects whose turbines arrive in 2030 or later will have difficulty starting operations that same year. However, these are the delivery schedules for orders already booked and projects under negotiation that Mitsubishi Heavy described in its August 2026 Q&A, and they do not apply to every plant planned in the US. GE Vernova's disclosed production capacity also covers the world as a whole, and the share allocated to the US has not been disclosed. Organizing the major manufacturers' production plans and plant startup outlooks gives the following picture.
| Period | Disclosure / plan | Source |
|---|---|---|
| Q3 2026 | GE Vernova's annual gas turbine production capacity expected to reach 20 GW | GE Vernova earnings (July 22, 2026) |
| 2028 | GE Vernova plans to expand annual capacity to 24 GW | Same as above |
| 2028–30 | Scheduled delivery of 10 large gas turbines ordered by Mitsubishi Heavy in April–June 2026 | Mitsubishi Heavy (August 2026) |
| 2030 | If plans hold, about 70% of new capacity would be fossil. GE Vernova is taking steps to expand annual capacity to 30 GW | Atlas Public Policy / EDF (October 13, 2026), GE Vernova |
| 2030–31 | Earliest period in which many new gas combined-cycle plants can start operating; scheduled delivery for projects Mitsubishi Heavy says are under negotiation | GridLab (September 2025), Mitsubishi Heavy |
Laid out chronologically, the 2030–31 delivery timing Mitsubishi Heavy has set for projects under negotiation is unlikely to allow startup in 2030. If fossil plants starting operations in 2030 do materialize at a scale close to current plans, they would likely be supported mainly by projects that have already secured turbine production slots.
In addition to supply constraints, construction costs are rising. GridLab's September 2025 tally found that gas combined-cycle plants scheduled to start in 2026–27 had construction costs of $1,116–1,427 per kW, whereas recent combined-cycle projects have reached $2,000 or more. By comparison, that implies a cost increase of at least about 40–79%. Specifically, 2,000 ÷ 1,427 − 1 gives 40.2%, and 2,000 ÷ 1,116 − 1 gives 79.2%; because $2,000 is stated as a floor ("or more"), the increase should also be read as a lower bound.
This, however, compares projects close to startup with recent projects; it does not mean the cost of the same plant has risen. The figures are also in nominal dollars of each project's startup year, so the increase partly reflects differences in startup years. If a 1 GW plant were built at $2,000 per kW, the total would be $2 billion. Converted at the ECB reference rate of October 16, 2026 (¥158.02 per dollar), that is about ¥316 billion. Higher construction costs increase pressure on developers to revisit plans or push back startup dates, but there are currently no figures showing that cancellations are rising.
Data Center Demand Forecasts May Be Overstated by About 25 GW
There is uncertainty not only on the supply side but also in the outlook for the power demand that plants are meant to serve. Grid Strategies' 2025 report, compiled mainly from data that utilities submitted to the US Federal Energy Regulatory Commission (FERC) and adjusted for some regions using transmission planning figures, forecast that peak power demand will rise by 166 GW over the next five years. Data center demand accounts for about 90 GW of that. However, compared with market analysts' forecasts, utilities' estimates of data center demand may be overstated by about 25 GW, or roughly 40%.
The EDF and Atlas Public Policy report also cites forecasts of rising power demand from data centers and other sources as one reason new plant plans have tilted toward gas. If actual demand falls short of forecasts, fossil construction plans built on those forecasts could be forced to shrink or be postponed. However, the 25 GW of overstated demand identified by Grid Strategies and the 68 GW of fossil capacity tallied by Atlas Public Policy differ in scope and method, so one cannot simply be subtracted from the other to derive required generating capacity.
Other data also raise questions about how feasible the plans are. A Global Energy Monitor analysis from August 2026 found that gas-fired capacity under development in the US rose 50% in six months to 378 GW. Meanwhile, the turbine or engine manufacturer has not been disclosed for about two-thirds of gas projects under development worldwide and more than half of data center projects. Jenny Martos of the organization noted that it is now nearly impossible to tell which proposals are pie in the sky and which have a real chance of being built (original: "It is nearly impossible nowadays to guess what is a pie in the sky proposal, and what has a real chance of getting built."). Note that Global Energy Monitor's 378 GW covers all gas capacity under development, so its scope differs from the 68 GW that Atlas Public Policy tallied for facilities scheduled for completion in 2026–30.
Will Turbine Expansion and a Shift to Gas Engines Resolve the Supply Constraint?
There is a counterargument to the view that fossil startups in 2030 will be delayed: gas turbine production capacity is expanding rapidly. GE Vernova's backlog of gas power equipment plus slot reservation agreements rose from 100 GW to 116 GW in the most recent quarter, and the company expects it to reach 125 GW or more by the end of 2026. In May 2026, Mitsubishi Heavy president Ito also outlined a plan to raise shipments of GTCC equipment by more than 30% over current capacity in fiscal 2028, and then by a further 50% or more, expanding to about double the fiscal 2024 level.
Among developers, there is also a growing move to adopt reciprocating (gas) engines rather than wait for large gas turbines. According to Global Energy Monitor, reciprocating-engine capacity under development for data centers more than tripled in six months to 45 GW. Furthermore, because captive generation not connected to the grid is not included in the EIA's tally, the scale of new fossil capacity actually built could exceed the scale reflected in the roughly 70% projection for 2030.
Still, even as production capacity expands, orders and slot reservations are piling up faster. Dividing GE Vernova's combined backlog and reservations of 116 GW by the annual capacity of 20 GW expected in Q3 2026 gives 5.8 years' worth, and dividing by the 24 GW per year planned for 2028 gives about 4.8 years. In other words, a simple calculation suggests that orders already placed and reserved production slots alone would require about five years of production capacity. This is, however, a rough estimate that uses worldwide figures without distinguishing firm orders from reservation agreements, and it does not indicate actual delivery waiting times. Nor has the amount of capacity allocated to the US been disclosed.
These numbers suggest that equipment newly ordered from GE Vernova is likely to be delivered after 2030. For projects Mitsubishi Heavy says are under negotiation, by contrast, delivery is 2030–31, which includes 2030, so the same judgment cannot be applied as for GE Vernova. And while a shift to reciprocating engines may increase fossil supply, it is a response using a different equipment configuration from the large gas combined-cycle plants in the EIA's planning data.
Another counterargument is that solar and batteries have short construction times, so projects starting in 2030 are not yet fully reflected in the planning data, making fossil's share look higher than it will be. EDF and Atlas Public Policy reject this view, explaining that development times do not differ greatly by generation type and that many clean-energy projects also expect 2030–31 startup. The development times behind that claim, however, are estimates from industry groups and national laboratories.
By cancellation rate, too, fossil plants currently fare better. Of the 51.6 GW of fossil capacity planned in early 2025, 1.5 GW (2.9%) has been cancelled, versus 13.1 GW (4.7%) of the 281.7 GW of clean capacity. At least on current results, fossil plans are not especially prone to cancellation. But the question here is not whether planned gas plants will ultimately be built; it is whether they can start operating in their scheduled year. Even if construction plans are maintained, startup can still be pushed back by turbine delivery or construction delays.
The most important thing to watch in testing this view is the manufacturers' earnings disclosures during 2027. In particular, if Mitsubishi Heavy moves up the delivery timing for GTCC projects under negotiation, now set at 2030–31, that would signal that supply constraints are easing more than expected. In that case, fossil plants starting operations in 2030 would be more likely to materialize at a scale close to current plans. For GE Vernova, a key indicator will be how far the number of years obtained by dividing combined backlog and reservations by annual capacity falls from the current figure of about five.
The implications for Japan also deserve attention. Six of the 10 large gas turbines Mitsubishi Heavy ordered in April–June 2026 are for Japan, meaning US and Japanese plant plans draw on the same manufacturer's capacity. However, current company disclosures do not show whether rising US demand is delaying deliveries of equipment for Japan. Mitsubishi Heavy plans to raise GTCC equipment shipment capacity by more than 30% in fiscal 2028, and GE Vernova is taking steps to expand to 30 GW of annual production in 2030. How far these expansion measures can actually pull forward delivery timing, and whether they can support plant startup plans for 2030, is the focus to check in the 2027 earnings reports.