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:
Why the new “reasoning” models might gobble up more electricity — at least in the short term
What happens when artificial intelligence takes some time to think?
The newest set of models from OpenAI, o1-mini and o1-preview, exhibit more “reasoning” than existing large language models and associated interfaces, which spit out answers to prompts almost instantaneously.
Instead, the new model will sometimes “think” for as long as a minute or two. “Through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes,” OpenAI announced in a blog post last week. The company said these models perform better than their existing ones on some tasks, especially related to math and science. “This is a significant advancement and represents a new level of AI capability,” the company said.
But is it also a significant advancement in energy usage?
In the short run at least, almost certainly, as spending more time “thinking” and generating more text will require more computing power. As Erik Johannes Husom, a researcher at SINTEF Digital, a Norwegian research organization, told me, “It looks like we’re going to get another acceleration of generative AI’s carbon footprint.”
Discussion of energy use and large language models has been dominated by the gargantuan requirements for “training,” essentially running a massive set of equations through a corpus of text from the internet. This requires hardware on the scale of tens of thousands of graphical processing units and an estimated 50 gigawatt-hours of electricity to run.
Training GPT-4 cost “more than” $100 million OpenAI chief executive Sam Altman has said; the next generation models will likely cost around $1 billion, according to Anthropic chief executive Dario Amodei, a figure that might balloon to $100 billion for further generation models, according to Oracle founder Larry Ellison.
While a huge portion of these costs are hardware, the energy consumption is considerable as well. (Meta reported that when training its Llama 3 models, power would sometimes fluctuate by “tens of megawatts,” enough to power thousands of homes). It’s no wonder that OpenAI’s chief executive Sam Altman has put hundreds of millions of dollars into a fusion company.
But the models are not simply trained, they're used out in the world, generating outputs (think of what ChatGPT spits back at you). This process tends to be comparable to other common activities like streaming Netflix or using a lightbulb. This can be done with different hardware and the process is more distributed and less energy intensive.
As large language models are being developed, most computational power — and therefore most electricity — is used on training, Charlie Snell, a PhD student at University of California at Berkeley who studies artificial intelligence, told me. “For a long time training was the dominant term in computing because people weren’t using models much.” But as these models become more popular, that balance could shift.
“There will be a tipping point depending on the user load, when the total energy consumed by the inference requests is larger than the training,” said Jovan Stojkovic, a graduate student at the University of Illinois who has written about optimizing inference in large language models.
And these new reasoning models could bring that tipping point forward because of how computationally intensive they are.
“The more output a model produces, the more computations it has performed. So, long chain-of-thoughts leads to more energy consumption,” Husom of SINTEF Digital told me.
OpenAI staffers have been downright enthusiastic about the possibilities of having more time to think, seeing it as another breakthrough in artificial intelligence that could lead to subsequent breakthroughs on a range of scientific and mathematical problems. “o1 thinks for seconds, but we aim for future versions to think for hours, days, even weeks. Inference costs will be higher, but what cost would you pay for a new cancer drug? For breakthrough batteries? For a proof of the Riemann Hypothesis? AI can be more than chatbots,” OpenAI researcher Noam Brown tweeted.
But those “hours, days, even weeks” will mean more computation and “there is no doubt that the increased performance requires a lot of computation,” Husom said, along with more carbon emissions.
But Snell told me that might not be the end of the story. It’s possible that over the long term, the overall computing demands for constructing and operating large language models will remain fixed or possibly even decline.
While “the default is that as capabilities increase, demand will increase and there will be more inference,” Snell told me, “maybe we can squeeze reasoning capability into a small model ... Maybe we spend more on inference but it’s a much smaller model.”
OpenAI hints at this possibility, describing their o1-mini as “a smaller model optimized for STEM reasoning,” in contrast to other, larger models that “are pre-trained on vast datasets” and “have broad world knowledge,” which can make them “expensive and slow for real-world applications.” OpenAI is suggesting that a model can know less but think more and deliver comparable or better results to larger models — which might mean more efficient and less energy hungry large language models.
In short, thinking might use less brain power than remembering, even if you think for a very long time.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
Current conditions: In the Atlantic, the tropical storm that could, as it develops, take the name Jerry is making its way westward toward the U.S. • In the Pacific, Hurricane Priscilla strengthened into a Category 2 storm en route to Arizona and the Southwest • China broke an October temperature record with thermometers surging near 104 degrees Fahrenheit in the southeastern province of Fujian.
The Department of Energy appears poised to revoke awards to two major Direct Air Capture Hubs funded by the Infrastructure Investment and Jobs Act in Louisiana and Texas, Heatmap’s Emily Pontecorvo reported Tuesday. She got her hands on an internal agency project list that designated nearly $24 billion worth of grants as “terminated,” including Occidental Petroleum’s South Texas DAC Hub and Louisiana's Project Cypress, a joint venture between the DAC startups Heirloom and Climeworks. An Energy Department spokesperson told Emily that he was “unable to verify” the list of canceled grants and said that “no further determinations have been made at this time other than those previously announced,”referring to the canceled grants the department announced last week. Christoph Gebald, the CEO of Climeworks, acknowledged “market rumors” in an email, but said that the company is “prepared for all scenarios.” Heirloom’s head of policy, Vikrum Aiyer, said the company wasn’t aware of any decision the Energy Department had yet made.
While the list floated last week showed the Trump administration’s plans to cancel the two regional hydrogen hubs on the West Coast, the new list indicated that the Energy Department planned to rescind grants for all seven hubs, Emily reported. “If the program is dismantled, it could undermine the development of the domestic hydrogen industry,” Rachel Starr, the senior U.S. policy manager for hydrogen and transportation at Clean Air Task Force told her. “The U.S. will risk its leadership position on the global stage, both in terms of exporting a variety of transportation fuels that rely on hydrogen as a feedstock and in terms of technological development as other countries continue to fund and make progress on a variety of hydrogen production pathways and end uses.”
Remember the Tesla announcement I teased in yesterday’s newsletter? The predictions proved half right: The electric automaker did, indeed, release a cheaper version of its midsize SUV, the Model Y, with a starting price just $10 shy of $40,000. Rather than a new Roadster or potential vacuum cleaner, as the cryptic videos the company posted on CEO Elon Musk’s social media site hinted, the second announcement was a cheaper version of the Model 3, already the lower-end sedan offering. Starting at $36,990, InsideEVs called it “one of the most affordable cars Tesla has ever sold, and the cheapest in 2025.” But it’s still a far cry from Musk’s erstwhile promise to roll out a Tesla for less than $30,000.
That may be part of why the company is losing market share. As Heatmap’s Matthew Zeitlin reported, Tesla’s slice of the U.S. electric vehicle sales sank to its lowest-ever level in August despite Americans’ record scramble to use the federal tax credits before the September 30 deadline President Donald Trump’s new tax law set. General Motors, which sold more electric vehicles in the third quarter of this year than in all of 2024, offers the cheapest battery-powered passenger vehicle on the market today, the Chevrolet Equinox, which starts at $35,100.
Get Heatmap AM directly in your inbox every morning:
Trump’s pledge to revive the United States’ declining coal industry was always a gamble — even though, as Matthew reported in July, global coal demand is rising. Three separate stories published Tuesday show just how stacked the odds are against a major resurgence:
As you may recall from two consecutive newsletters last month, Secretary of Energy Chris Wright said “permitting reform” was “the biggest remaining thing” in the administration’s agenda. Yet Republican leaders in Congress expressed skepticism about tacking energy policy into the next reconciliation bill. This week, however, Utah Senator Mike Lee, the chairman of the Senate Committee on Energy and Natural Resources, called for a legislative overhaul of the National Environmental Policy Act. On Monday, the pro-development social media account Yimbyland — short for Yes In My Back Yard — posted on X: “Reminder that we built the Golden Gate Bridge in 4.5 years. Today, we wouldn’t even be able to finish the environmental review in 4.5 years.” In response, Lee said: “It’s time for NEPA reform. And permitting reform more broadly.”
Last month, a bipartisan permitting reform bill got a hearing in the House of Representatives. But that was before the government shutdown. And sources familiar with Democrats’ thinking have in recent months suggested to me that the administration’s gutting of so many clean energy policies has left Republicans with little to bargain with ahead of next year’s midterm elections.
Soon-to-be Japanese prime minister Sanae Takaichi.Yuichi Yamazaki - Pool/Getty Images
On Saturday, Japan’s long-ruling Liberal Democratic Party elected its former economic minister, Sanae Takaichi, as its new leader, putting her one step away from becoming the country’s first woman prime minister. Under previous administrations, Japan was already on track to restart the reactors idled after the 2011 Fukushima disaster. But Takaichi, a hardline conservative and nationalist who also vowed to re-militarize the nation, has pushed to speed up deployment of new reactors and technologies such as fusion in hopes of making the country 100% self-sufficient on energy.
“She wants energy security over climate ambition, nuclear over renewables, and national industry over global corporations,” Mika Ohbayashi, director at the pro-clean-energy Renewable Energy Institute, told Bloomberg. Shares of nuclear reactor operators surged by nearly 7% on Monday on the Tokyo Stock Exchange, while renewable energy developers’ stock prices dropped by as much as 15%
Researchers at the United Arab Emirates’ University of Sharjah just outlined a new method to transform spent coffee grounds and a commonly used type of plastic used in packaging into a form of activated carbon that can be used for chemical engineering, food processing, and water and air treatments. By repurposing the waste, it avoids carbon emitting from landfills into the atmosphere and reduces the need for new sources of carbon for industrial processes. “What begins with a Starbucks coffee cup and a discarded plastic water bottle can become a powerful tool in the fight against climate change through the production of activated carbon,” Dr. Haif Aljomard, lead inventor of the newly patented technology, said in a press release.
Last week’s Energy Department grant cancellations included funding for a backup energy system at Valley Children’s Hospital in Madera, California
When the Department of Energy canceled more than 321 grants in an act of apparent retribution against Democrats over the government shutdown, Russ Vought, President Trump’s budget czar, declared that the money represented “Green New Scam funding to fuel the Left's climate agenda.”
At least one of the grants zeroed out last week, however, was supposed to help keep the lights on at a children’s hospital.
The $29 million grant was intended to build a 3.3-megawatt long-duration energy storage system at Valley Children’s Hospital, a large pediatric hospital in Madera, California. The system would “power critical hospital operations during outage events,” such as when the California grid shuts down to avoid starting wildfires, according to project documents.
“The U.S. Department of Energy’s cancellation of funding for [the] long-duration energy storage demonstration grant is disappointing,” Zara Arboleda, a spokesperson for the hospital, told me.
Valley Children’s Hospital is a 358-bed hospital that says it serves more than 1.3 million children across California’s Central Valley. It has 116 neonatal intensive care unit beds and nationally ranked specialties in pediatric neurology, orthopedics, and lung surgery, among others.
Energy Secretary Chris Wright has characterized the more than $7.5 billion in grants canceled last week as part of an ongoing review of financial awards made by the Biden administration. But the timing of the cancellations — and Vought’s gleeful tweets about them — suggests a more vindictive purpose. Republican lawmakers and President Trump himself threatened to unleash Vought as a kind of rogue budget cutter before the federal government shut down last week.
“We don’t control what he’s going to do,” Senator John Thune told Politico last week. “I have a meeting today with Russ Vought, he of PROJECT 2025 Fame, to determine which of the many Democrat Agencies, most of which are a political SCAM, he recommends to be cut,” Trump posted on the same day.
Up until this year, canceling funding that is already under contract with a private party would have been thought to be straightforwardly illegal under federal law. But the Supreme Court’s conservative majority has allowed the Trump administration to act with previously unimaginable freedom while it considers ruling on similar cases.
Faraday Microgrids, the contractor that was due to receive the funding, is already building a microgrid for the hospital. The proposed backup power system — which the grant stipulated should be “non-lithium-ion” — was supposed to be funded by the Energy Department’s Office of Clean Energy Demonstrations, with the goal of finding new ways of storing electricity without using lithium-ion batteries, and was meant to work in concert with that new microgrid and snap on in times of high stress.
That microgrid project is still moving forward, Arboleda, the hospital’s spokesperson, told me. “Valley Children’s Hospital continues to build and soon will operate its microgrid announced in 2023 to ensure our facilities have access to reliable and sustainable energy every minute of every day for our patients and our care providers,” she added. That grid will contain some storage, but not the long-term storage system discussed in the official plan.
Faraday Microgrids, formerly known as Charge Bliss, didn’t respond to a request for comment, but its website touts its ability to secure grants and other government funding for energy projects.
In a statement, a spokesman for the Energy Department said that the grant was canceled because the project wasn’t feasible. “Following an in-depth review of the financial award, it was determined, among other reasons, that the viability of the project was not adequate to warrant further disbursements,” Ben Dietderich, a spokesman for the Energy Department, told me.
The children’s hospital, at least, is in good company. On Tuesday, a Trump administration document obtained by Heatmap News suggested the Energy Department is moving to kill bipartisan-backed funding for two direct air capture hubs in Texas and Louisiana. And although California has lost the most grants of any state, the Energy Department has also sought to terminate funding for new factories and industrial facilities across Republican-governed states.
Editor’s note: This story initially misstated the number of neonatal intensive care unit beds at Valley Children’s Hospital. It has been corrected.
Rob and Jesse break down China’s electricity generation with UC San Diego’s Michael Davidson.
China announced a new climate commitment under the Paris Agreement at last month’s United Nations General Assembly meeting, pledging to cut its emissions by 7% to 10% by 2035. Many observers were disappointed by the promise, which may not go far enough to forestall 2 degrees Celsius of warming. But the pledge’s conservatism reveals the delicate and shifting politics of China’s grid — and how the country’s central government and its provinces fight over keeping the lights on.
On this week’s episode of Shift Key, Rob and Jesse talk to Michael Davidson, an expert on Chinese electricity and climate policy. He is a professor at the University of California, San Diego, where he holds a joint faculty appointment at the School of Global Policy and Strategy and the Jacobs School of Engineering. He is also a senior associate at the Center for Strategic and International Studies, and he was previously the U.S.-China policy coordinator for the Natural Resources Defense Council.
Shift Key is hosted by Robinson Meyer, the founding executive editor of Heatmap, and Jesse Jenkins, a professor of energy systems engineering at Princeton University.
Subscribe to “Shift Key” and find this episode on Apple Podcasts, Spotify, Amazon, YouTube, or wherever you get your podcasts.
You can also add the show’s RSS feed to your podcast app to follow us directly.
Here is an excerpt from our conversation:
Robinson Meyer: Your research and other people’s research has revealed that basically, when China started making capacity payments to coal plants, in some cases, it didn’t have the effect on the bottom line of these plants that was hoped for, and also we didn’t really see coal generation go down or change in the year that it happened. It wasn’t like they were paying these plants to stick around and not run. They were basically paying these plants, it seems like, to do the exact same thing they did the year before, but now they also got paid. And maybe that was needed for their economics, we can talk about it.
Why did coal get those payments and not, say, batteries or other sources of spare capacity, like pumped hydro storage, like nuclear? Why did coal, specifically, get payments for capacity? And does it have to do with spinning reserve? Or does it have to do with the political economy of coal in China?
Michael Davidson: When it came out, we said exactly the same thing. We said, okay, this should be a technology neutral payment scheme, and it should be a market, not a payment, right? But China’s building these things up little by little. Over time we’ve seen, historically, actually, a number of systems internationally started with payments before they move to markets because they realize that you could get a lot more competitive pressure with markets.
The capacity payment scheme for coal is extremely simple, right? It says, okay, for each province, we’re going to say what percentage of our benchmark coal investment costs are we going to subsidize. It’s extremely simple. It does not account for how much you’re using it at a plant by plant level. It does not account for other factors, renewables, etc. It’s a very coarse metric. But I wouldn’t say that it had had some, you know, perverse negative effect on the outcome of what coal generation is. Probably more likely is that these payments were seen, for some, as extra support. But then for some that are really hurting, they’re saying, okay, well then we will maybe put up less obstacles to market reforms.
But then on top of that, you have to put in the hourly energy demand growth story and say, okay, well you have all these renewables, but you don’t have enough storage to shift to evening peaks. You are going to rely on coal to meet that given the current rigid dispatch system. And so you’re dispatching them kind of regardless of whether or not you have the payment schemes.
I will say that I was a skeptic, right? Because when people told me that China should put in place a capacity market, I said, China has overcapacity. So if you have an overcapacity situation, you put in place a market, the prices should be zero. So what’s the point? But actually, when you’re looking out ahead with all of this surplus coal capacity that you’re trying to push down, you’re trying to push those capacity factors of those coal plans from 50%, 60%, down to 20% or even lower, they need to have other revenue schemes if you’re not going to dramatically open up your spot markets, which China is very hesitant to do — very risk averse when it comes to the openness of spot markets, in terms of price gaps. So that’s a necessary part of this transition. But it can be done more efficiently, and it should done technology neutral.
And by the way that is happening in certain places. That’s a national scheme, but we actually see that the implementation — for example, Shaanxi province, we have a technology neutral scheme that would include other resources, not just coal.
Mentioned:
China’s new pledge to cut its emissions by 2035
What an ‘ambitious’ 2035 electricity target looks like for China
China’s Clean Energy Pledge is Clouded by Coal, The Wire China
Jesse’s upshift; Rob’s upshift.
This episode of Shift Key is sponsored by …
Hydrostor is building the future of energy with Advanced Compressed Air Energy Storage. Delivering clean, reliable power with 500-megawatt facilities sited on 100 acres, Hydrostor’s energy storage projects are transforming the grid and creating thousands of American jobs. Learn more at hydrostor.ca.
A warmer world is here. Now what? Listen to Shocked, from the University of Chicago’s Institute for Climate and Sustainable Growth, and hear journalist Amy Harder and economist Michael Greenstone share new ways of thinking about climate change and cutting-edge solutions. Find it here.
Music for Shift Key is by Adam Kromelow.