You’re out of free articles.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
Sign In or Create an Account.
By continuing, you agree to the Terms of Service and acknowledge our Privacy Policy
Welcome to Heatmap
Thank you for registering with Heatmap. Climate change is one of the greatest challenges of our lives, a force reshaping our economy, our politics, and our culture. We hope to be your trusted, friendly, and insightful guide to that transformation. Please enjoy your free articles. You can check your profile here .
subscribe to get Unlimited access
Offer for a Heatmap News Unlimited Access subscription; please note that your subscription will renew automatically unless you cancel prior to renewal. Cancellation takes effect at the end of your current billing period. We will let you know in advance of any price changes. Taxes may apply. Offer terms are subject to change.
Subscribe to get unlimited Access
Hey, you are out of free articles but you are only a few clicks away from full access. Subscribe below and take advantage of our introductory offer.
subscribe to get Unlimited access
Offer for a Heatmap News Unlimited Access subscription; please note that your subscription will renew automatically unless you cancel prior to renewal. Cancellation takes effect at the end of your current billing period. We will let you know in advance of any price changes. Taxes may apply. Offer terms are subject to change.
Create Your Account
Please Enter Your Password
Forgot your password?
Please enter the email address you use for your account so we can send you a link to reset your password:
Fossil fuel plant retirements are slowing down, and projected load growth is to blame.

To fully decarbonize the electricity system will require more than just the rapid deployment of non-carbon-emitting generation capacity, plus the transmission necessary to get that electricity to where it needs to go. It will also require that our existing stock of electricity generation — which is largely natural gas- and coal-powered — get mostly mothballed. So far, this process has been proceeding briskly. Renewable deployment is on the way up and is projected to accelerate, and older electricity generation was sliding quickly but gracefully into retirement — until recently.
Retirements of existing generation have slowed down dramatically in the first half of this year, which is on pace to be the slowest for existing generation retirements since 2011, according to new data from the Energy Information Administration.
In the first half of the year, some 5.1 gigawatts of generating capacity have been retired, and another 2.4 gigawatts are scheduled to be retired by year’s end, for a projected total of 7.5 retired gigawatts. From 2004 to 2023, by contrast, just over 12 gigawatts of capacity were retired each year on average, with almost 15 gigawatts retired per year this decade. Since 2022, according to EIA data, over 90% of retired capacity has been coal or natural gas.
What’s behind the slowdown? “Reliability is threatened because the grid conditions are tightening,” Douglas Giuffre, executive director of gas, power and renewables analysis at S&P Global Commodity Insights, explained in an email. “This is partly due to the recent pace of coal and natural gas retirements in the U.S., which worked off some of the excess capacity in power markets. Now we are seeing tighter reserve margins, and a relatively thin pipeline of new gas-fired projects that can come online quickly.” That’s especially concerning for utilities at a time when projected electricity demand is way, way up.
The wave of retirements was a national phenomenon, often having nothing to do with state-level plans to decarbonize. Coal and gas were being retired so steadily over the past 20 years not just because plants were aging, but also because power use was essentially flat from the early 2000s through, essentially, yesterday. This meant that older plants — especially dirty coal plants — became uneconomic to run, especially as natural gas prices began to fall.
Now, we are in a completely different world. Electricity use is forecast to start growing again, thanks to a buildout of new data centers and manufacturing, plus the ongoing electrification of automobiles and home heating and cooling.
The Southeast offers an example of how these trends have played out on the ground. In December 2020, the Mississippi Public Service Commission determined that the state had “excess reserves … largely due to decreases in projected load” and ordered a 950 megawatt reduction in generating capacity by Mississippi Power by 2027. A consulting firm hired by the commission determined that Plant Daniel, a coal plant, was “relatively inefficient compared to other available resources;” a few months later, the utility said it would decommission Plant Daniel by 2027.
Then Georgia Power, the utility that covers most of the state (and, like Mississippi Power, a subsidiary of Southern Company), rushed out a new three-year plan for its future power usage less than a year after finalizing its old one. Its demand forecast through the end of the decade had jumped from 400 megawatts to 6,600 megawatts, the result of a projected boom in data center construction.
“They came in with a preselected list of ways it wanted to meet that power need,” including buying power from Plant Daniel and new gas, Bob Sherrier, a staff attorney at the Southern Environmental Law Center, told me. Georgia Power told the state’s utility commission that to respond to growing demand it would need to extend contracts with its sister utility in Mississippi — which meant not only that Daniel would remain open for at least another year — and build new new plants that could run on gas or diesel, plans for which regulators approved on Tuesday. The utility also hinted that its existing plans to euthanize, for the most part, its coal-fired generation fleet by the end of 2028 were likely to be revised.
“To meet that projected need, the utilities are reverting to what they know, which is fossil fuels,” Sherrier said.
In vertically integrated markets, where utilities own generating assets and sell power to customers, environmentalists have seen delayed retirements and the building of new fossil plants as examples of utilities slipping into their comfort zone, building and operating expensive projects instead of developing or procuring renewables to handle rising demand.
But it's not just in vertically integrated markets where fossil retirements are being delayed. In Maryland, for instance, Brandon Shores, a coal-fired power plant that was scheduled to close in 2025, is staying open because PJM Interconnection, the regional electricity market, determined that a plan to replace it with battery storage was not a “realistic option at present” nor “technically viable to resolve the reliability violations or avoid the need for an RMR agreement at this time,” PJM president Manu Asthana said in a letter to Paul Pinsky, the director of the Maryland Energy Administration. The transmission investments required to make up the difference, meanwhile, would take several years.
Along with the neighboring Wagner plant, which burns a mix of coal, oil, and natural gas, Brandon Shores will likely stay open more than three years past its planned retirement date thanks to what’s known as a “reliability must run” contract, which “would put Maryland ratepayers on the hook for over $600 million dollars in out-of-market payments,” according to a letter written by several Maryland congressional representatives to PJM.
Environmental advocates have blamed PJM for not doing enough proactive transmission planning to account for predictable and scheduled plant retirements.
The slowing retirements mean that emissions from the electricity sector, which have been falling since the mid-2000s (with occasional bumps up as the economy has recovered from downturns), are expected to plateau over the next year or so. EIA forecasts show carbon dioxide emissions from electricity as essentially flat from 2023 to 2025, with increased natural gas emissions essentially offsetting falling coal emissions.
There is a bright side to the data, however. So far this year, the U.S. has installed just over 20 gigawatts of new generation, 80% of which has been solar and battery storage, including a 600-plus megawatt projects in Nevada and Texas. If added generation comes on in the second half of this year as planned, the EIA projects we’ll have 15 gigawatts of battery storage by year’s end. Along with the large and growing solar generation in states like California, Nevada, and Texas, the U.S. is getting closer to a grid that can, at least, run without carbon emissions day or night.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
At this point, I think it’s clear that AI data centers are unpopular.
You probably know it, at least. I was preparing talk about data center opposition on a podcast today and I took the opportunity to dive back into our data, so I certainly know it. At this point, we’ve written about results from our polling that show Americans overwhelmingly oppose local data center construction, that majorities of Americans now support a national data center moratorium, and that the only group of Americans who feels more optimistic than pessimistic about artificial intelligence is … men older than 65 years old.
So I got curious: Given all that, who actually supports AI data centers?
One question from our recent Heatmap Pro poll, conducted by Embold Research, helps give us a sense. This is the profile of someone our data says would support a data center built in their local area:
A few facets stand out. These data center YIMBYs are more likely to be men, and more likely to be 2024 Trump voters, but they’re not locked into one age demographic or voting cohort. A third are Harris supporters, and roughly a third are women. Data center YIMBYs are more likely to be older than 50, but the majority isn’t overwhelming.
Get Rob in your inbox daily.
Perhaps more surprising: The group has many more people who voted third-party in the 2024 election (8%) than the general population (just under 2%), although that response could also include people who didn’t vote. (Alas, the data can’t quite confirm how many in this group are libertarian.)
What’s perhaps most interesting: This group overwhelmingly believes that artificial intelligence will make their lives better. And in heartening news for climate advocates, they are even more likely to support a given data center project if it is powered by renewables.
I was going to joke that the profile is essentially a newly retired engineering dad — except that, to my surprise, these data center YIMBYs are far less gender imbalanced than the American engineering profession. (They’re also less gender-imbalanced than American Tesla owners.) So I’ll leave it at that.
Five takeaways from the latest Lazard Levelized Cost of Energy report.
It’s all getting more expensive.
That’s the conclusion of the investment bank Lazard’s latest report on the levelized cost of energy, one of the most closely watched and cited energy reports of the year.
Levelized cost of energy measures the dollars per megawatt-hour a power plant needs to earn in revenue to break even over the course of its lifetime in present-value terms.
What makes LCOE so alluring is that it’s a way to compare any type of generator, whether it requires a large upfront investment but has few operating costs, like a utility-scale solar project, or whether its expenses are largely fuel costs incurred in the future, like a combined cycle natural gas plant. This is also why LCOE has its critics, who point out that a solar panel that only runs during certain times of day has a different value to the electricity system than a natural gas plant that can ramp up and down quickly or a nuclear plant that provides steady baseload power.
Anyway, here’s what we can learn from this year’s Lazard report.
Curves that were once gently sloping downward are starting to look like incipient U’s. While longterm LCOE falls are still dramatic and impressive for some technologies — utility solar has fallen from $359 per megawatt-hour in 2009 to $69 in 2026 — the short term rises are worrisome. That $69 per megawatt hour represents a nearly 10% increase from 2025, when utility-scale solar had a LCOE of $58. And it’s not just renewables — the LCOE for a combined cycle natural gas plant rose from $78 per megawatt-hour to $90 in the past year. Gas plant LCOE got as low as $60 in 2021. That’s a 50% price hike in just five years.
Lazard attributed the increase in solar and wind LCOE to “higher capital costs, sustained interest rates, tariff pass-through and supply chain repricing.” These technologies are also the most “sensitive” to subsidies by way of the tax code, with federal tax tax credits taking the low end cost of utility solar to as low as $16 per megawatt hour. To the extent those tax credits are no longer available or weren’t accessible due to strict eligibility rules, that could be a source of future upward pressure on costs.
That being said, renewables “maintain their relative cost advantage despite facing the same cost pressures affecting the rest of the generation stack,” the Lazard analysts concluded.
Natural gas, meanwhile, is seeing prices spiral upward on huge and growing customer demand.
“Continuous upward revisions to demand projections have driven a sharp increase in announced new-build gas generation despite a 15-year high LCOE and historically long development lead times,” according to Lazard.
The report hints at what LCOE is not always able to capture, namely that generators like gas have attributes besides low cost that make them attractive. “New gas combined cycle plants offer the lowest-cost dispatchable power in high-demand and low-cost-gas environments,” the analysts point out.
Anyone building a new combined cycle gas plant, however, will have to deal with the high cost and low availability for turbines, which is “extending development timelines well beyond historical norms.” That provides an opening for renewables that can be deployed quickly and cheaply, like solar and accompanied by battery storage.
In 2019, the low end of LCOE for onshore end was $28 per megawatt-hour, according to Lazard’s figures, and the high end was $54. In 2026, however, the low end costs sits a bit higher at $37 per megawatt-hour, but the high end cost rose to $99. There’s a similar story for utility solar: in 2019, the spread between low and high was a snug $8 per megawatt-hour, while this year it’s ballooned to $58.
The broadening range is “likely reflecting that some project developers have been better able to mitigate broader cost pressures across supply chain and project-level economics than others,” the Lazard analysts wrote.
The Lazard report doesn’t just look at the discounted cost of individual generators over their lifetimes. It also tries to figure how much they cost on certain grids. One way of doing this is to look at what Lazard calls the “cost of firming intermittency” or “levelized firming costs.” This is essentially looking at what it costs to bring solar, solar and storage, and wind and storage onto actual grids considering their ability to perform when the grid is most stressed.
This measure tries to refine LCOE to give a sense of how various forms of energy generation compare to gas plants in real world circumstances, not just as a financial construct. This is not a perfect, real-world comparison — gas capacity needs to be “firmed” as well, as it’s not always entirely available at times of peak need — but at least it gives an idea of how these resources actually function in a real-world grid.
Even with firming costs, “renewables remain broadly cost-competitive,” the report concludes.
Not surprisingly, some of the most dramatic costs are in America’s most troubled electricity market, PJM Interconnection. The unsubsidized cost of firming intermittency for solar and storage is $167 per megawatt-hour, compared to $150 in Texas or $115 in California. That’s also compared to a $129 per megawatt-hour at the high end for conventional combined cycle gas plants in PJM.
PJM is notorious for its inability to bring on new resources quickly and its strict standards for accrediting the contribution of storage and renewables to grid stability.
While the Lazard authors explicitly caution that it doesn’t measure what the“total system costs are for 1 MWh of incremental electricity” and can’t say “the optimal mix of renewables, conventional generation and storage,” it does conclude that “firming costs and dispatchability are increasingly critical for comparing resources on a more complex grid.”
In short, no matter what ends up on the grid, grid planners will have to think carefully about how to make sure it’s reliable and works in concert with what’s already there.
Timber companies think of them as pests, but new research indicates that stands of the slender tree can act as barriers against raging flames.
Colorado’s Aspen Acres Fire is named after a quiet RV campground located high in the San Isabel Mountains, about a five-hour drive due southeast of the state’s better-known Aspen. Both places, however, are named after the iconic deciduous tree known for its golden leaves in the fall. While the start of monsoon season may yet prevent the Aspen Acres Fire — the seventh-largest in Colorado’s history — from joining Utah’s Babylon Fire as the second 100,000-acre “megafire” of the season, the conflagration has been aided in its rampage not by aspens, but rather by dead, downed, and blighted ponderosa pines, spruce, and Douglas firs. The wildfire has now burned over 98,000 acres and nearly 300 homes, and is only 36% contained due to steep terrain that has hampered firefighting efforts, along with extreme drought conditions and beetle infestations that have greatly degraded the forest health of the region.
But what about its aspens? Though the extent of the damage at the campground remains unknown, according to a recent study of Populus tremuloides, Colorado’s iconic golden trees could be one of the keys to more wildfire-resistant forests in the future.
Flavie Pelletier, a recent PhD graduate of McGill University’s Natural Resource Sciences program, told me she first became interested in aspens while working as a tree planter in British Columbia. “The historical assumption on aspen is that stands are very good at stopping fire progression. But the paradox is that if you take an aspen by itself, it’s going to burn at high severity,” Pelletier, who published her findings in Forest Ecology and Management, told me.
By creating near-real-time maps of fires using satellites and comparing them against the Canadian Forest Service’s newly available maps of dominant tree species in the boreal, Pelletier and her colleagues discovered that aspen were almost two and a half times more common at the perimeter of a burned area than inside it. The finding suggests that despite the flammability of a single aspen with its thin bark, stands of aspen act as a kind of barrier when wildfire ran up against them, likely because they lack the flammable resins of conifers and their high foliage helps force running crown fires back toward the ground. Pine and spruce, by contrast, showed a near-zero or even negative effect.
When aspen stands did burn, Pelletier found they did so more slowly: A tree cover of 50% aspen burned at about 224 hectares per day, compared to 717 hectares per day in areas where aspen made up less than 10% of the cover. That’s the equivalent of about 1,000 FIFA-regulation soccer pitches per day in places where aspen are sparser — like Aspen Acres.
Even more surprising, though, was that the pattern held true in the early season, when the trees are still twiggy and have yet to grow their moisture-filled leaves, and despite the severity of fire weather. “Aspen still showed resilience even when the fire weather was very intense, [like in 2023, when] we had all the fires,” Pelletier said.
But she was also the first to admit that seasons are getting more extreme, and that there’s no guarantee the pattern will hold for the next 10 or 20 years.
Pelletier was reluctant to make a policy recommendation based on her research, noting that she’s not a forest manager. But in Alberta and British Columbia, timber companies spray hundreds of thousands of acres of timber with glyphosate, an herbicide, to kill off aspens because the trees outcompete the more commercially valuable conifers. Her findings are “a big argument to stop the spreading of herbicides because you’re increasing the risk of fire in your forest by removing aspen,” Pelletier said.
Despite her hesitation, Pelletier is explicit in her paper about one thing: that aspens “should be encouraged — specifically around key landscape positions, such as population centers” — given that they are a proven means of hardening the wildland-urban interface against wildfires. It might be too late for the idyllically named Aspen Acres, of course; any of the aspens that once drew tourists to the area are likely now ash.
But this not be Colorado’s last fire, either.