The environment the grid AI runs on has to survive
Data-centre siting arguments treat generation as a number. This scores every station in the register for the environment its outdoor hardware actually stands in.
Key figures
42,094 stations scored, 10,679.7 GW of capacity.
3,669 stations, 1,800.6 GW, fall in the two highest exposure classes — 16.9% of scored capacity.
779.8 GW of that is gas, the capacity most often named as the answer to AI demand.
What the index measures
The Climate Exposure Screening Index scores each station from CES1 to CESX on the environment its outdoor equipment stands in: annual mean temperature, summer maximum, seasonal swing as a proxy for thermal cycling, distance to coast and Köppen-Geiger class. It covers 42,094 stations and 10,679.7 GW.
It is a screening heuristic for prioritising inspection, not a corrosion-rate model. It is not an ISO 9223 dose-response calculation and should not be quoted as one.
How capacity is spread across classes
| Class | Stations | Capacity (GW) |
|---|---|---|
| CES1 | 6,060 | 1,679.7 |
| CES2 | 10,839 | 2,192.2 |
| CES3 | 13,669 | 3,081.8 |
| CES4 | 7,857 | 1,925.5 |
| CES5 | 3,337 | 1,664.5 |
| CESX | 332 | 136.1 |
What actually attacks the hardware
| Dominant stressor | Stations | Capacity (GW) |
|---|---|---|
| humidity / wetness | 23,684 | 5,250.6 |
| thermal cycling | 7,987 | 2,053.5 |
| dust abrasion | 4,939 | 1,302.5 |
| marine salt corrosion | 3,669 | 1,800.6 |
| marine corrosion | 1,787 | 269.1 |
| heat / UV | 28 | 3.5 |
Humidity and wetness lead on both counts. Thermal cycling is second, and it is the one that matters most for anything bolted to a hot surface, because it is the expansion and contraction of the pipework that works fixings loose.
An honest word about the top classes
Every one of the 3,669 stations in CES5 and CESX has marine salt corrosion as its dominant stressor, and every marine-salt station is in CES5 or CESX. That is not a discovery about the world. It is how the index is weighted: proximity to the coast dominates the composite score.
Stated plainly, the top two classes are a coastal filter. That is useful if the question is where salt-driven degradation should be inspected first. It is misleading if read as a general ranking of harsh sites, because an inland station with severe thermal cycling can never reach the top class however hard its duty is.
Where the exposed capacity is
| Country | Capacity in CES5 and CESX (GW) |
|---|---|
| China | 306.2 |
| India | 252.7 |
| United States | 155.1 |
| Vietnam | 135.3 |
| Indonesia | 104.1 |
| Saudi Arabia | 91.3 |
| Brazil | 69.3 |
| United Arab Emirates | 64.2 |
Sources and method
- Scores from the power plant climate exposure dataset, https://doi.org/10.5281/zenodo.22172589, counted on 2026-09-25.
- Station locations, fuels and capacities from PowerAtlas.
- The index is a screening heuristic built from open climate grids. It is not an ISO 9223 dose-response calculation and carries no corrosion rate.
- WorldClim climate grids and the Köppen-Geiger classification, the inputs behind the exposure index’s temperature, seasonal swing and climate class fields.
- Global Power Plant Database, World Resources Institute, the largest single upstream source for the station records used here.