Global power emissions went flat. America's went up 4.3% — and the reason isn't what most people assume

What the numbers say. In the IEA's Electricity 2026 analysis, global CO₂ emissions from electricity generation were flat in 2025. Underneath that flat line, the major economies moved in different directions. India fell 3.3%. The European Union fell 2.2%. China fell about 1%. The United States rose 4.3%.

Four of the world's largest power systems, three going one way and one going the other.

What the IEA attributes it to. The analysis is specific, and short: the American increase came from higher coal-fired generation. Not from a collapse in renewables, not from nuclear retirements, not from a demand shock — from coal running more hours.

The distinction that most coverage collapses. It is worth separating four things that get treated as one:

Emissions changed. Electricity demand changed. Coal dispatch changed. And data-centre load grew. These are related but they are not the same measurement, and only the first two are in the IEA's headline figures.

On the second: the IEA forecasts electricity demand growing at an average of 3.6% a year through 2026–2030, which it calls "a sharp acceleration from the 2.8% pace of the past decade." That is a real and large change.

On the fourth: the IEA's emissions analysis does not mention data centres. That is worth stating plainly, because a great deal of commentary now treats AI data centres as the explanation for rising American power emissions, and this particular source does not make that claim. Data-centre growth is well documented elsewhere and is plainly part of why demand is accelerating. But "demand is rising partly because of data centres" and "the 4.3% emissions increase was caused by data centres" are different statements, and only the first is supported here.

Why coal is the mechanism. When demand on a grid rises, the system dispatches whatever additional generation is available and economic. If the marginal unit called on is coal, emissions rise roughly in proportion to how much more it runs — coal is the most carbon-intensive common source. So an emissions increase driven by coal dispatch tells you something specific: it was not that America built new coal, it was that existing coal plants ran more, because something had to meet the additional load and that was what the system reached for.

That also explains why the same demand growth produces different results elsewhere. India's demand is growing fast too, and its power emissions still fell — because what met the additional load was overwhelmingly new solar and wind rather than existing coal.

What the IEA expects next, and its reasoning. The forecast is that global power-sector emissions plateau across 2026–2030, despite that 3.6% annual demand growth. The IEA's own summary of why: "Even with strong gains in electricity demand, growth in power sector emissions is showing marked signs of slowing down as fossil-fired generation is constrained by the rapid deployment of renewables and rising nuclear power generation."

Read carefully, that is a claim about a race, not a victory. Clean capacity is being added fast enough to absorb most of the growth in demand. Where it isn't, fossil generation fills the gap. The plateau holds only as long as the first term keeps up with the second.

Why it matters. The American figure is a preview of what happens when load growth outruns clean capacity additions in a system that still has dispatchable coal available. Every large economy is about to test the same arithmetic, because demand growth is accelerating almost everywhere. The interesting question is not whether emissions rise or fall globally — flat is the current answer — but which side of that race each grid lands on. In 2025, three of the four largest landed on one side and the United States landed on the other.

What happens next, and it is checkable. The 2026 figures, when the IEA publishes them, will show whether the American increase was a single year or a trend. US coal generation is reported monthly by the EIA, so the mechanism can be watched directly rather than inferred. And the plateau forecast is falsifiable in the most useful way: if global power emissions rise materially in 2026 or 2027, the race is being lost, and the reason will be visible in exactly this kind of regional breakdown.