Yves here. So much for saving the planet. Many experts had warned that AI, particularly power-hungry US implementations, would greatly increase energy demand. It’s already taken off even before Trump repudiated reducing carbon emissions and other environmental protections.
By Wolf Richter, editor at Wolf Street. Originally published at Wolf Street
The quantity of electricity generated in the US by all sources, from natural gas to rooftop solar, rose by 3.1% in 2024 from 2023 to a record of 4,304,039 gigawatt-hours (GWh), according to data from the EIA today.
This is now clearly a breakout in demand, after 14 years of stagnation, from 2007 through 2021, when electricity users, to reduce their costs, invested in more efficient equipment – lights, appliances, electronic equipment, industrial equipment, heating and air-conditioning, etc. – and in better building insulation, shading, etc., to reduce their power costs. This relentless drive for greater efficiency kept demand roughly stable for years despite the growing economy and population. And it mired many power generators and electric utilities in a no-growth business where it was difficult to justify investment.
Now the scenario has changed, largely due to the growth in demand from data centers (AI, cloud, crypto) and the increasing penetration of EVs in the national vehicle fleet – EVs accounted for over 10% of US vehicle sales in 2024.
The Share of Total Electricity Generated by Source:
Natural gas rules. Power generation from natural gas rose by 3.3% to a record of 1,864,874 GWh in 2024.
The share of natural gas as source for power generation remained roughly unchanged in 2024, matching the record of 42.7% of 2023, about double its share in 2007. Natural gas had surpassed nuclear in 2006 and coal in 2016 (blue in the chart below).
The US is the largest natural gas producer and the largest LNG exporter in the word. Production has oversupplied the US market and has caused the price of natural gas to collapse since 2009.
The modern combined-cycle gas turbine powerplants have a thermal efficiency of around 65%, nearly double that of older coal powerplants. These two – low price of US natural gas and the high efficiency of the combined-cycle plants – made natural gas immensely attractive for power generators.
Coal power generation fell by 3.3% to 652,760 GWh in 2024. Its share dropped to a record low of 14.9% of total power generation, down from 51% in 2001 (black in the chart below).
Coal cannot compete with cheap natural gas and the efficiency of a combined-cycle gas turbine powerplant. More recently, wind power became more cost-efficient than coal. It all boils down to costs. Gas is cheap. With renewables, the “fuel” is free; and all methods of power generation require costly plants, equipment, maintenance.
Power generators have not built any new coal-fired power plants over the past decade. They’re too inefficient and expensive to operate.
And they’ve been retiring their old inefficient and expensive-to-operate coal-fired power plants. In 2025, of the 12.3 Gigawatts (GW) of capacity that power generators plan to retire, 66% are old coal-fired plants, 21% are old natural-gas-fired plants, 13% are old petroleum-fired plants.
Generation from all renewables combined – wind, solar, hydro, geothermal, and biomass – rose by 3.1% to a record 1,061,258 GWh, driven by surging generation from wind (+7.7%) and solar (+26.9%).
The share of all renewables combined increased to 24.2% of total power generation (red). More on them separately in a moment.
Nuclear power generation edged up 0.9% to 781,979 GWh, and its share edged down to 17.8% of total generation (green).
The share of petroleum liquids and petroleum coke declined to 0.3%, having nearly vanished as source of power generation. The planned retirements this year will further reduce generation (purple).
Power Generation from Renewables
Wind power generation jumped by 7.7% in 2024, to a record 453,454 GWh. Its share grew to 10.3% of total power generation (red in the chart below).
The Big Five states for utility-scale wind-power generation in 2023, according to separate EIA data, in GWh and % share of US wind power generation:
- Texas: 119,836 GWh, 28%
- Iowa: 41,869 GWh, 10%
- Oklahoma: 37,731 GWh, 9%
- Kansas: 27,462 GWh, 6.5%
- Illinois: 22,054 GWh, 5%
Solar power generation – utility scale and rooftop solar – surged 26.9% to 303,167 GWh. Its share ballooned to 6.9% of the total power generated, surpassing hydropower (yellow).
Wind and solar combined had a share of 17.2% of total power production in the US, a higher share than coal, and close to nuclear.
Power generation from small-scale solar – such as rooftop systems on homes, retail stores, parking garages, etc. – jumped 15.3% to 84,630 GWh, for a share of 1.9% of total power generated in 2024.
Hydropower generation dipped 1.1% to 242,226 GWh. Its share declined to 5.5% of total power generation (blue).
Biomass power generation declined 1.0% to 46,740 GWh, and its share eased to 1.1% of total power generated. Biomass includes wood and wood-derived fuels, landfill gas, and other waste biomass (black).
Geothermal remained at a minuscule share of 0.4% of total power generated. Most geothermal plants were built in the 1970s in California (green).
Just a point about electricity demand – by far the biggest user of electricity (residential and commercial) is space heating and air con. Annual demand is therefore very weather driven, hence it is unwise to read too much into year by year fluctuations. Deeper demand shifts take several years to become apparent. The ‘flattening out’ of demand seen since the 2008 crash is something that’s been seen in nearly all advanced economies – most likely driven by much greater efficiency in lighting and other plant. It should also be highlighted that increased electricity demand by EV’s represents a significant drop in overall energy use due to the inherently greater efficiency of electric drives. Its not clear (all the major agencies disagree on this), whether increased electrification and data centres will result in a greater level of demand to off-set what seems a long term decline in usage due to efficiency gains and demographic flattening out. Most likely we are at a plateau in the advanced economies.
Increased demand by data centres and crypto, etc., has been rapid and intense but mostly been in regional grids – internationally its not a particularly big driver of demand, and is for the foreseeable future likely to be still a relatively small overall component relative to the ‘big’ users – i.e. heating, cooling, lighting. Getting too hung up on it is a major distraction from the necessity of making these sectors far more efficient (most progress has been on lighting).
A point on natural gas generation – the big driver for natural gas generation was not fracking or ‘efficiency’ as Richter puts it, or even fuel cost. The increase in natural gas use was seen in other countries too, including those without much natural gas. It was an accidental by product of increasing deregulation of grids in the 1990’s, which favoured investments which were relatively low capital input, but which could take advantage of high peak costs. In the UK in the 1990’s it was called the ‘dash for gas’ – entirely accidental, but it devastated the economics of the big thermal plants. The big advantage gas has in a grid over coal (or pretty much anything else), is that its relatively cheap to build a CCGT plant, but even on low usage and high gas prices they can be very profitable as they can supply the highest priced electricity during peak periods, something you can’t do with big thermal plants or (mostly) with renewables. This drove the huge increase in CCGT capacity, which then, when gas prices dropped, meant that it drove the thermal plants out of the market as they were then able to eat into overall demand. This has also had the accidental by-product of making wind and solar more viable, as gas is very effective at load balancing. In many grids now, a combination of renewables and natural gas has made traditional base load thermal plants unnecessary. Thermal plants are now only being built at any scale in countries (like China or Vietnam), where the market has been deliberately unbalanced to encourage investment in excess capacity.