Energy Project Budget Overruns: Per-Source Data
Energy projects show the widest variance in overrun rates of any infrastructure category. Nuclear and frontier oil-and-gas plants are catastrophic. Solar and wind are remarkably well-behaved.
| Energy source | Mean cost overrun | Worst-case examples |
|---|---|---|
| Nuclear (large gen-III/III+) | +120% | Olkiluoto-3 (+300%), Flamanville (+450%), Hinkley Point C (+150%) |
| Frontier oil and gas (Arctic, deep water) | +100% to +400% | Kashagan (+400%), Sakhalin-2 (+100%) |
| Conventional oil and gas | +34% | Various LNG facilities |
| Hydroelectric and large dams | +96% | Itaipu (+240%), Three Gorges |
| Coal-fired | +27% | Kemper IGCC (+400%) |
| Onshore wind | +8% | Block Island (+40% offshore) |
| Offshore wind | +13% | Vineyard Wind 1 |
| Utility-scale solar PV | +1% | Cestas (+25%) |
| Power transmission | +8% | HVDC inter-connectors |
Sources: Flyvbjerg (2024), Sovacool & Cooper (2013, UEA), and project-specific regulatory filings (EDF, OL3 owners' consortium, EPA Kemper).
Why solar and wind behave so much better than nuclear
Solar panels and wind turbines are modular products built on a factory line. Each new project deploys roughly the same units as the last. The marginal project benefits from the learning curve of every previous one, and the design risk is borne once at the OEM, not project by project.
Nuclear plants are the opposite. Almost every gen-III reactor built in the last 25 years has been a first-of-a-kind or first-in-country, with bespoke site engineering, bespoke regulatory negotiation, and a supply chain that lost most of its expertise during the 1990s-2000s nuclear hiatus. There is no learning curve when every plant is a one-off.
This is the strongest argument for small modular reactors (SMRs): if they can hit a true factory-built, repeatable design, overrun risk should converge with solar rather than with traditional nuclear.