Power Plant Construction Cost per MW: 2026 Benchmarks

2026 build-cost benchmarks by technology

Each figure below is installed or overnight capital cost per kW of net capacity; multiply by 1,000 for cost per MW. The low end of each range generally reflects engineering-estimate datasets and favorable regions; the high end reflects recent signed or filed projects in today’s constrained equipment and labor market.

Sources: EIA / Sargent & Lundy, Capital Cost and Performance Characteristic Estimates (Dec 2023, 2023$, published for AEO2025); GridLab, Energy Futures Group & Halcyon, The New Reality of Power Generation (Sept 2025); Lazard, Levelized Cost of Energy+ (Version 19.0, July 2026); NREL Annual Technology Baseline 2024 and 2025 cost benchmarks; LBNL Utility-Scale Solar 2025 Data Update and Land-Based Wind Market Report; Utility Dive coverage of GE Vernova Q1-Q2 2026 results; DOE, Bechtel, and Georgia Power on Vogtle 3 & 4. All retrieved August 2026.

Per-kW versus per-MW, and the AC/DC caveat

Cost per kW and cost per MW are the same metric at different scales: $1,500/kW equals $1.5M per MW. Two traps distort comparisons. First, solar and battery figures may be quoted per kW-DC or per kW-AC; a plant with a 1.3 DC/AC ratio looks about 25% cheaper on a DC basis, so this guide quotes AC. Second, battery costs are more honestly stated per kWh of energy: a $334/kWh 4-hour system is $1,336/kW, but the same dollars per kW would buy only two hours of duration at twice the per-kWh price. Always confirm the basis before comparing bids.

What moves the number

The spread between a dataset benchmark and a signed EPC contract is driven by a handful of forces, most of which have pushed in the same direction since 2024. Ranked roughly by impact on gas-heavy portfolios:

The common thread: none of these drivers is cyclical demand alone. OEMs are deliberately protecting margin over volume, electrical-equipment supply chains remain tight, and the craft workforce is finite. Owners should budget to the filed-project end of the ranges, not the dataset end, for anything reaching service after 2028.

Regional and market variation

Identical scope can price 30-50% apart depending on where it is built. EIA applies regional multipliers for labor, logistics, and ambient conditions that run from 0.93x on the Gulf Coast to 1.39x in New York City, and GridLab’s filing data shows apparent regional spread on a small sample. Directionally, from cheapest to most expensive:

For owners, the regional lesson is that labor geography is now as decisive as equipment price. The same H-class island can carry a seven-figure-per-MW swing between a loose Gulf Coast craft market and a saturated PJM one, and multipliers shift year to year as mega-projects absorb regional workforces.

What changed from 2024 to 2026

In 2024, planning datasets still told a benign story: EIA’s Sargent & Lundy study (December 2023) put combined cycle at $868-921/kW, and NREL’s ATB projected roughly $1,638/kW for a 1x1 CCGT in 2030. By late 2025, GridLab’s review of actual IRP and CPCN filings showed new gas projects clearing $2,000-2,260/kW, and by mid-2026 trade coverage of GE Vernova’s results put new-order economics near $2,500/kW with the backlog at 116 GW, up from 100 GW a single quarter earlier. That is a doubling of market-clearing gas capex in roughly two years, driven by turbine scarcity, reservation fees, and electrical-equipment inflation rather than any one-off shock.

Renewables and storage moved far less, but not down. LBNL data shows standalone solar LCOE rising from $53/MWh (2023) to $60/MWh (2024) on higher capex and financing, and US battery pricing at $270-458/kWh installed remains well above the sub-$130/kWh systems built in China, with tariffs and domestic-content rules setting the floor. Lazard’s July 2026 edition still ranks utility solar ($40-98/MWh) and onshore wind ($37-99/MWh) as the cheapest new energy, with gas CCGT now at $51-129/MWh as turbine repricing flows through.

The net 2026 picture: every path to new capacity is more expensive than the 2024 planning books assumed, the cost ranking between technologies has widened rather than reshuffled, and delivery dates, not dollars, are increasingly the binding constraint owners are paying to secure.

The workforce behind every megawatt

Capital cost per MW is, to a large degree, labor cost per MW. Vogtle 3 and 4 peaked around 9,000 craft professionals on one site for 2,234 MW, roughly four workers per MW at peak, spanning welders, pipefitters, electricians, and ironworkers (DOE/Bechtel). A single large CCGT typically concentrates several hundred to over a thousand craft workers for a 30-40 month build, heaviest in pipe trades and electrical, while utility solar and BESS need shorter, wider peaks dominated by electricians and equipment operators. Multiply those curves across the 116 GW of backlogged gas turbines plus record solar and storage pipelines, and the same regional craft pools are being claimed several times over.

That is why labor now shows up in the cost data itself: GridLab lists skilled-worker shortages alongside turbine backlogs as a reason elevated prices will persist, and regional spreads of 30-50% track union density and craft availability as much as steel. For owners, the practical move is to treat workforce as a critical-path procurement item, locking key supervision, pipe, and electrical capacity on the same planning horizon as the turbine slot, because in 2026 a missing crew slips a COD just as surely as a missing transformer.

For the hiring side of this market, see the Power & Grid Construction practice.