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.
A new Data for Progress poll provided exclusively to Heatmap shows steep declines in support for the CEO and his business.
Nearly half of likely U.S. voters say that Elon Musk’s behavior has made them less likely to buy or lease a Tesla, a much higher figure than similar polls have found in the past, according to a new Data for Progress poll provided exclusively to Heatmap.
The new poll, which surveyed a national sample of voters over the President’s Day weekend, shows a deteriorating public relations situation for Musk, who has become one of the most powerful individuals in President Donald Trump’s new administration.
Exactly half of likely voters now hold an unfavorable view of Musk, a significant increase since Trump’s election. Democrats and independents are particularly sour on the Tesla CEO, with 81% of Democrats and 51% of independents reporting unfavorable views.
By comparison, 42% of likely voters — and 71% of Republicans — report a favorable opinion of Musk. The billionaire is now eight points underwater with Americans, with 39% of likely voters reporting “very” unfavorable views. Musk is much more unpopular than President Donald Trump, who is only about 1.5 points underwater in FiveThirtyEight’s national polling average.
Perhaps more ominous for Musk is that many Americans seem to be turning away from Tesla, the EV manufacturer he leads. About 45% of likely U.S. voters say that they are less likely to buy or lease a Tesla because of Musk, according to the new poll.
That rejection is concentrated among Democrats and independents, who make up an overwhelming share of EV buyers in America. Two-thirds of Democrats now say that Musk has made them less likely to buy a Tesla, with the vast majority of that group saying they are “much less likely” to do so. Half of independents report that Musk has turned them off Teslas. Some 21% of Democrats and 38% of independents say that Musk hasn’t affected their Tesla buying decision one way or the other.
Republicans, who account for a much smaller share of the EV market, do not seem to be rushing in to fill the gap. More than half of Republicans, or 55%, say that Musk has had no impact on their decision to buy or lease a Tesla. While 23% of Republicans say that Musk has made them more likely to buy a Tesla, roughly the same share — 22% — say that he has made them less likely.
Tesla is the world’s most valuable automaker, worth more than the next dozen or so largest automakers combined. Musk’s stake in the company makes up more than a third of his wealth, according to Bloomberg.
Thanks in part to its aging vehicle line-up, Tesla’s total sales fell last year for the first time ever, although it reported record deliveries in the fourth quarter. The United States was Tesla’s largest market by revenue in 2024.
Musk hasn’t always been such a potential drag on Tesla’s reach. In February 2023, soon after Musk’s purchase of Twitter, Heatmap asked U.S. adults whether the billionaire had made them more or less likely to buy or lease a Tesla. Only about 29% of Americans reported that Musk had made them less likely, while 26% said that he made them more likely.
When Heatmap asked the question again in November 2023, the results did not change. The same 29% of U.S. adults said that Musk had made them less likely to buy a Tesla.
By comparison, 45% of likely U.S. voters now say that Musk makes them less likely to get a Tesla, and only 17% say that he has made them more likely to do so. (Note that this new result isn’t perfectly comparable with the old surveys, because while the new poll surveyed likely voters , the 2023 surveys asked all U.S. adults.)
Musk’s popularity has also tumbled in that time. As recently as September, Musk was eight points above water in Data for Progress’ polling of likely U.S. voters.
Since then, Musk has become a power player in Republican politics and been made de facto leader of the Department of Government Efficiency. He has overseen thousands of layoffs and sought to win access to computer networks at many federal agencies, including the Department of Energy, the Social Security Administration, and the IRS, leading some longtime officials to resign in protest.
Today, he is eight points underwater — a 16-point drop in five months.
“We definitely have seen a decline, which I think has mirrored other pollsters out there who have been asking this question, especially post-election,” Data for Progress spokesperson Abby Springs, told me .
The new Data for Progress poll surveyed more than 1,200 likely voters around the country on Friday, February 14, and Saturday, February 15. Its results were weighted by demographics, geography, and recalled presidential vote. The margin of error was 3 percentage points.
On Washington walk-outs, Climeworks, and HSBC’s net-zero goals
Current conditions: Severe storms in South Africa spawned a tornado that damaged hundreds of homes • Snow is falling on parts of Kentucky and Tennessee still recovering from recent deadly floods • It is minus 39 degrees Fahrenheit today in Bismarck, North Dakota, which breaks a daily record set back in 1910.
Denise Cheung, Washington’s top federal prosecutor, resigned yesterday after refusing the Trump administratin’s instructions to open a grand jury investigation of climate grants issued by the Environmental Protection Agency during the Biden administration. Last week EPA Administrator Lee Zeldin announced that the agency would be seeking to revoke $20 billion worth of grants issued to nonprofits through the Greenhouse Gas Reduction Fund for climate mitigation and adaptation initiatives, suggesting that the distribution of this money was rushed and wasteful of taxpayer dollars. In her resignation letter, Cheung said she didn’t believe there was enough evidence to support grand jury subpoenas.
Failed battery maker Northvolt will sell its industrial battery unit to Scania, a Swedish truckmaker. The company launched in 2016 and became Europe’s biggest and best-funded battery startup. But mismanagement, production delays, overreliance on Chinese equipment, and other issues led to its collapse. It filed for Chapter 11 bankruptcy protection in November and its CEO resigned. As Reutersreported, Northvolt’s industrial battery business was “one of its few profitable units,” and Scania was a customer. A spokesperson said the acquisition “will provide access to a highly skilled and experienced team and a strong portfolio of battery systems … for industrial segments, such as construction and mining, complementing Scania's current customer offering.”
TikTok is partnering with Climeworks to remove 5,100 tons of carbon dioxide from the air through 2030, the companies announced today. The short-video platform’s head of sustainability, Ian Gill, said the company had considered several carbon removal providers, but that “Climeworks provided a solution that meets our highest standards and aligns perfectly with our sustainability strategy as we work toward carbon neutrality by 2030.” The swiss carbon capture startup will rely on direct air capture technology, biochar, and reforestation for the removal. In a statement, Climeworks also announced a smaller partnership with a UK-based distillery, and said the deals “highlight the growing demand for carbon removal solutions across different industries.”
HSBC, Europe’s biggest bank, is abandoning its 2030 net-zero goal and pushing it back by 20 years. The 2030 target was for the bank’s own operations, travel, and supply chain, which, as The Guardiannoted, is “arguably a much easier goal than cutting the emissions of its loan portfolio and client base.” But in its annual report, HSBC said it’s been harder than expected to decarbonize supply chains, forcing it to reconsider. Back in October the bank removed its chief sustainability officer role from the executive board, which sparked concerns that it would walk back on its climate commitments. It’s also reviewing emissions targets linked to loans, and considering weakening the environmental goals in its CEO’s pay package.
A group of 27 research teams has been given £81 million (about $102 million) to look for signs of two key climate change tipping points and create an “early warning system” for the world. The tipping points in focus are the collapse of the Greenland ice sheet, and the collapse of north Atlantic ocean currents. The program, funded by the UK’s Advanced Research and Invention Agency, will last for five years. Researchers will use a variety of monitoring and measuring methods, from seismic instruments to artificial intelligence. “The fantastic range of teams tackling this challenge from different angles, yet working together in a coordinated fashion, makes this program a unique opportunity,” said Dr. Reinhard Schiemann, a climate scientist at the University of Reading.
In 2024, China alone invested almost as much in clean energy technologies as the entire world did in fossil fuels.
Editor’s note: This story has been updated to correct the name of the person serving as EPA administrator.
Rob and Jesse get real on energy prices with PowerLines’ Charles Hua.
The most important energy regulators in the United States aren’t all in the federal government. Each state has its own public utility commission, a set of elected or appointed officials who regulate local power companies. This set of 200 individuals wield an enormous amount of power — they oversee 1% of U.S. GDP — but they’re often outmatched by local utility lobbyists and overlooked in discussions from climate advocates.
Charles Hua wants to change that. He is the founder and executive director of PowerLines, a new nonprofit engaging with America’s public utility commissions about how to deliver economic growth while keeping electricity rates — and greenhouse gas emissions — low. Charles previously advised the U.S. Department of Energy on developing its grid modernization strategy and analyzed energy policy for the Lawrence Berkeley National Laboratory.
On this week’s episode of Shift Key, Rob and Jesse talk to Charles about why PUCs matter, why they might be a rare spot for progress over the next four years, and why (and how) normal people should talk to their local public utility commissioner. Shift Key is hosted by Jesse Jenkins, a professor of energy systems engineering at Princeton University, and Robinson Meyer, Heatmap’s executive editor.
Subscribe to “Shift Key” and find this episode on Apple Podcasts, Spotify, Amazon, 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: I want to pivot a bit and ask something that I think Jesse and I have talked about, something that you and I have talked about, Charles, is that the PUCs are going to be very important during the second Trump administration, and there’s a lot of possibilities, or there’s some possibilities for progress during the Trump administration, but there’s also some risks. So let’s start here: As you survey the state utility landscape, what are you worried about over the next four years or so? What should people be paying attention to at the PUC level?
Charle Hua: I think everything that we’re hearing around AI data centers, load growth, those are decisions that ultimately state public utility commissioners are going to make. And that’s because utilities are significantly revising their load forecasts.
Just take Georgia Power — which I know you talked about last episode at the end — which, in 2022, just two years ago, their projected load forecast for the end of the decade was about 400 megawatts. And then a year later, they increased that to 6,600 megawatts. So that’s a near 17x increase. And if you look at what happens with the 2023 Georgia Power IRP, I think the regulators were caught flat footed about just how much load would actually materialize from the data centers and what the impact on customer bills would be.
Meyer:And what’s an IRP? Can you just give us ...
Hua: Yes, sorry. So, integrated resource plan. So that’s the process by which utilities spell out how they’re proposing to make investments over a long term planning horizon, generally anywhere from 15 to 30 years. And if we look at, again, last year’s integrated resource plan in Georgia, there was significant proposed new fossil fuel infrastructure that was ultimately fully approved by the public service commission.
And there’s real questions about how consumer interests are or aren’t protected with decisions like that — in part because, if we look at what’s actually driving things like rising utility bills, which is a huge problem. I mean, one in three Americans can’t pay their utility bills, which have increased 20% over the last two years, two to three years. One of the biggest drivers of that is volatile gas prices that are exposed to international markets. And there’s real concern that if states are doubling down on gas investments and customers shoulder 100% of the risk of that gas price volatility that customers’ bills will only continue to grow.
And I think what’s going on in Georgia, for instance, is a harbinger of what’s to come nationally. In many ways, it’s the epitome of the U.S. clean energy transition, where there’s both a lot of clean energy investment that’s happening with all of the new growth in manufacturing facilities in Georgia, but if you actually peel beneath the layers and you see what’s going on internal to the state as it relates to its electricity mix, there’s a lot to be concerned about.
And the question is, are we going to have public utility commissions and regulatory bodies that can adequately protect the public interest in making these decisions going forward? And I think that’s the million dollar question.
This episode of Shift Key is sponsored by …
Download Heatmap Labs and Hydrostor’s free report to discover the crucial role of long duration energy storage in ensuring a reliable, clean future and stable grid. Learn more about Hydrostor here.
Music for Shift Key is by Adam Kromelow.