Yves here. This humble site has repeatedly pointed out that “clean” energy often does not live up to its billing, due to use of materials that have large, nasty environmental effects, and supposedly green devices relying on dirty energy sources. In the US, there had been a push to shift electricity generation, away from coal and to some degree, other fossil fuel sources. That initiative has gone into reverse thanks to ginormous AI/data center power demands and Trump hearting fossil fuels. By contrast, China has also (in the first decade of this century) been building more coal-fired electrical plants. In 2024, coal still represented 59% of China’s electrical generation, versus 19.5% in the US. However, the US shift has been primarily from very dirty coal to less dirty but still CO2 producing gas. In addition, China has been working to change its profile, while the US is now pedal to the metal in reverse.
By Leonard S. Hyman, an economist and financial analyst specializing in the energy sector who headed utility equity research at a major brokerage house and William I. Tilles, a senior industry advisor and speaker on energy and finance who headed utility equity research at two major brokerage houses and then became a portfolio manager investing in long/short global utility equities. Originally published at OilPrice
- The U.S. is expanding electric vehicles, data centers, and heat pumps—yet the power feeding them increasingly comes from fossil fuels.
- Policymakers are approving new natural gas plants and extending coal operations.
- The U.S. is building out power capacity for electrification—but without meaningful carbon cuts, leading to a future of “clean tech” powered by dirty energy.
One of the biggest mistakes we made about electrification is that we assumed that the new technologies embracing electricity, like electric vehicles for example, would embrace decarbonization at the same time. We were wrong. In fact, in the US today, exactly the opposite has occurred. The paradox of electrification without decarbonization is simple. Take, for example, two convenient symbols of electrification, an electric car and a heat pump. Both displace fossil fuel usage, but both technologies also rely on electricity. We, and others, assumed that the large base load fossil-fired power generating stations favored by utilities would, due to growing public pressure, eventually be replaced by something more environmentally benign, like small modular reactors or extensive solar plus batteries. Instead, policymakers have decided to ignore environmental concerns and embrace more base load fossil fuel usage. In practical terms, this means permission to build new natural gas-fired power plants, while coal-fired power stations will be permitted or even encouraged to remain open.
Because power plants last forty years or more, due to inflation, a new plant always costs lots more than the unit it’s replacing. One of the frequent criticisms of renewable energy technologies like offshore wind or solar plus batteries is that they are very expensive. This is certainly true compared to the fully depreciated facility they are replacing. The simple fact is that all new power-generating facilities, regardless of technology, are far more costly than the units they replace. These are called capital expenses because the assets financed are expected to be in service for decades. But renewables have one outstanding advantage over fossil-fueled power generation. They have almost no big operating expenses like fuel. Unlike coal or gas-fired power plants, renewables only “consume” free stuff like wind or sunlight. Over a thirty or forty year lifetime operating period, the expense differentials are enormous. But in our intensely ideological times, these differences are often overlooked for partisan political purposes. And yes, we know they’re intermittent.
This idea of electrification without decarbonization has been adopted by technology executives such as Google’s former CEO, Eric Schmidt. Speaking at an AI conference last October, he pointed out that for him, and presumably other executives, attaining environmental goals like decarbonization is much less important than a vastly expanded network of data centers. His basic point was, “Forget about your climate targets, you won’t achieve them anyway.” Also he suggested that perhaps all these new data centers would actually solve the climate crisis. But as a former CEO, he also seemed to be asserting one very old corporate prerogative: the unlimited right of a corporation to pollute without interference from the government. As an aside, what’s so interesting to us is that no serious political figure has given such full-throated support for this view, probably since the McKinley era. But here we are.
People like Mr Schmidt and his former employer, Google, are among those leading the charge for much more rapid growth in domestic power generation resources, and they’ve stated quite plainly that they couldn’t care less about CO2 emissions and their adverse environmental implications. More fossil-fired power plants are just fine with them so long as the bulk electricity is affordable and available. Their “only” requirement is for an enormous increment to the existing generating fleet at a cost of at least several hundred billion dollars. These investments will either be “inside the fence” and paid for by the data centers themselves or they will be borne by local utilities with the large incremental expense passed along to consumers in the form of higher prices.
The question for us as analysts is to ask whether this view has become mainstream among senior policymakers in Washington and, if so, what are the implications? What would a pollution-indifferent buildout of the US power grid look like? In two words, more gas. There have been no new coal-fired power-generating plants built in the US in decades because their economics were undercut by domestically sourced natural gas. So a pollution-tolerant administration in Washington, for example, does not reverse the unattractive power plant economics of new coal. But the interesting thing for us is that even with a utility-friendly administration in Washington, the prospects for future gas-fired plants haven’t deviated much from long term historical trends. The US EIA expects about 18.7 gigawatts of new gas fired capacity to be added between now and 2028, with 2028 showing a big uplift relative to prior years. But the EIA’s historical data also show us what a big commitment to new gas-fired power generation truly looks like. In the five year period, 2000 to 2005, the US added over 160 gigawatts of new gas-fired power generation, with almost 50 gigawatts of new generation added in the year 2000 alone. The EIA data shows the US expecting to add about 11 gigawatts in 2028, a supposed breakout year. Like the man said, those are rookie numbers.
All we can conclude at this point is that policymakers in Washington have created an administrative permission structure for more pollution from fossil-fired power-generating stations. Whether this will lead to a vast increase in new fossil-fired power plant development remains to be seen. Despite what we and many others expected, the present goal seems to be more electrification, especially for AI, and without decarbonization.
I’m struggling to see on what basis the article can argue that the US grid is not decarbonising. It may be decarbonising too slow, but the trends are very clear. The carbon intensity of the US grid has declined by around 35% since 2007 and is still on a strong downward trajectory. There has been a total decline in carbon emissions from the US grid of 41% since 2005. The author of the article seems to be confusing overall capacity of the grid with real world emissions. Much of the new gas and coal is intended as balancing load and peaking load, not continual baseline since solar and wind are far cheaper when installed. It may be that a very significant increase in electricity demand due to data centres/AI and EV’s along with general electrification of industry could change this – but this impact can’t be seen in the national level data yet (impacts are mostly local and regional).
The situation in China is quite straightforward – there has been massive growth in electricity use, matched by a huge roll out of both renewables and coal/gas generation. The carbon intensity of each watt-hour of generation has reduced, but overall growth has ensured that emissions continue to rise. The overall impact has been to ensure the Chinese grid is still far ‘dirtier’ than most developed countries, and the situation is worsening. It is expected that there will be a strong decline in carbon intensity and (hopefully) overall emissions as demand eases off (as eventually it must) and renewables/nuclear take over as the main source of power. The bad news is that electricity demand in China is consistently increasing, even above the levels of economic growth. While we focus on the potential risks of AI/data centres, the real climate killer may be a newfound fondness of the Chinese for air conditioning.
Per capita CO2 emissions in China is about half of the USA. And yet you cut the USA a lot of slack while pointing the finger at China?
I’ve not found a study that would show the actual US energy consumption if we made all the stuff we import from China. And of course the corresponding drop in chinas energy use.
Carbon emissions embedded in trade tend not to be as high as people assume. The vast amount of carbon produced in economies tends to be in things like domestic heating, transport, agriculture and so on, not manufactured products. Studies into it tend to be plagued with definitional problems (such as who ‘owns’ the emissions if, say, China makes a product using energy from Australian coal and then sells it to the US). In reality, the total difference to overall emissions are not all that big unless there is an enormous disparity in trade, or a particular traded product happens to be very carbon intensive.
Per capita consumption-based CO₂ emissions that account for trade, according to the site you pointed at, was in 2022 16,5 t for the USA and 7,2 t for China. Could you elaborate why you think the real climate killer might be China increasing its energy consumption when its trade adjusted per capita consumption was 44% of that of the USA?
I like the link you have about trade emissions, very cool.
It appears to me if I understand it correctly, the US has a 564 million ton import. IE 564 tons of co2 is produced elsewhere which matches the idea of a large trade imbalance you mention of which the US has the single largest trade imbalance. China has a negative balance of 1 million tons because they have a huge export market.
You need to do WAY better than this.
Per capita GDP in China is less than one-sixth of that in the US. And China very much aspires to achieving US/Western lifestyles for its people.
You’ve actually proved PlutoniumKun’s point.
Please indulge me and explain what is the point that I apparently missed?
Also, I went to check, and China’s per capita GDP (PPP) for 2024 and according to the World Bank was about 1/3 (31%) that of the USA. Not 1/6.
Dear Mr Kun,
Two thoughts: 1) As Yves pointed out in her intro, the main driver of the decarbonization we’ve experienced has been to go from a very dirty base load fuel, coal, to a merely dirty one, gas; which emits about half as much CO2 as coal.
2) Where we disagree is your statement that “new gas and coal are intended as balancing loads.” First, all of the large turbine manufacturers are sold out for the next five years. These thousand megawatt CCGTs will begin come online in about two years according to the EIA. This is all base load. And more importantly, it’s a continuation of the coal to gas transition that we’ve seen for several decades. Homer City, for example, in Pennsylvania, a 1900 MW coal fired generation plant is being replaced by 4500 MWs of new CCGTs, for, of course, a new data center. Lastly, we won’t be able to use gas for balancing load longer term because it’s economics are being destroyed by batteries. The FT this week has an excellent chart on this.
News this week is that the tipping point for loss of coral reefs is close, with increased warnings of accelerating melt of both West Antarctica and Greenland Ice sheets.
Meanwhile on the Thwaite……….
Headlong into the abyss.
Can only suggest that all AI and data centres ought to be built less than 0.5m above sea level.
Here is a report that includes actual new generation by year and type.
https://bcse.org/market-trends/2025-key-trends/
Renewables have far outpaced everything else which really is just gas.
And now that battery prices have dramatically dropped in price they are making a much larger impact.
Eric Schmidt: “Forget about your climate targets, you won’t achieve them anyway.”
Comments like this from our corporate ‘leaders’ are among the reasons we are sleepwalking into a climate nightmare. They may be planning to ride it out in their luxury bunkers, but the melting glaciers would have it otherwise.
It’s a wicked hard problem, to accept that large load growth needs reliable industrial production, and intermittent (solar, wind) without storage is problematic at best. Hence, why natural gas is being built out as opposed to building large scale intermittent generation with corresponding large scale batteries or pumped storage hydro. I’d love to see AI flameout and hopefully see incremental intermittent growth to supplement with consumption grow with the population. I’m aware that utilities are incentivized to build big production to recoup their costs over a 40-50 year lifespan, and without specific policy incentives they won’t choose to RFP or build intermittent generation.
People would rather be told that AI will solve all their problems than be asked to rely less on cars and air conditioning. In that sense, companies and consumers are colluding.
Projections about additional gas turbine electrical generation are missing an important “fly(s) in the ointment.”
According to many observers, the current glut of low cost natural gas because of oversupply due to fracking is about to came to a screeching halt for two reasons. First, the fracking industry is about to enter a sharp decline due to the depletion of most/all of the major tight oil plays, resulting in a reduction of gas production. Second, after destroying Nordstream & committing Europe to purchasing wildly expensive liquid compressed natural gas from the U.S., the result will drive NG prices through the roof.
This likely scenario will have profound implications for AI in particular, and the economy in general.
CNG and to a greater extent LNG isn’t very friendly to transport or consume. It takes a lot of energy to compress (or liquify, natch) natural gas and you get penalized for that energy source on a weight basis compared to oil. However, compared to batteries natural gas is a clean rock star. I think that Germany or other Euro folks that are setup to consume ng will likely stick with imported gas from Russia, India, Scandinavia or even the US if they want to continue with their infrastructure. This may be a moot point if modular nukes get a running start in 3-5 years, but I’m a bit skeptical on that front. NC had a great article a week or so ago about small nukes being unable to bridge the anticipated growth from large load customers.