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.
Add it to the evidence that China’s greenhouse gas emissions may be peaking, if they haven’t already.
Exactly where China is in its energy transition remains somewhat fuzzy. Has the world’s largest emitter of greenhouse gases already hit peak emissions? Will it in 2025? That remains to be seen. But its import data for this year suggests an economy that’s in a rapid transition.
According to government trade data, in the first fourth months of this year, China imported $12.1 billion of coal, $100.4 billion of crude oil, and $18 billion of natural gas. In terms of value, that’s a 27% year over year decline in coal, a 8.5% decline in oil, and a 15.7% decline in natural gas. In terms of volume, it was a 5.3% decline, a slight 0.5% increase, and a 9.2% decline, respectively.
“Fossil fuel demand still trends down,” Lauri Myllyvirta, the co-founder of the Centre for Research on Energy and Clean Air, wrote on X in response to the news.
Morgan Stanley analysts predicted Friday in a note to clients that this “weak downstream demand” for coal in China would “continue to hinder coal import volume.”
Another piece of China’s emissions and coal usage puzzle came from Indonesia, which is a major coal exporter. Citing data from trade data service Kpler, Reuters reported Friday that Indonesia’s thermal coal exports “have dropped to their lowest in three years” thanks to “weak demand in China and India,” the world’s two biggest coal importers. Indonesia’s thermal coal exports dropped 12% annually to 150 million tons in the first third of the year, Reuters reported.
China’s official goal is to hit peak emissions by 2030 and reach “carbon neutrality” by 2060. The country’s electricity grid is largely fueled by coal (with hydropower coming in at number two), as is its prolific production of steel and cement, which is energy and, specifically, coal-intensive. For a few years in the 2010s, more cement was poured in China than in the whole 20th century in the United States. China also accounts for about half of the world’s steel production.
At the same time, China’s electricity demand growth is being largely met by renewables, implying that China can expand its economy without its economy-wide, annual emissions going up. This is in part due to a massive deployment of renewables. In 2023, China installed enough non-carbon-emitting electricity generation to meet the total electricity demand of all of France.
China’s productive capacity has shifted in a way that’s less carbon intensive, experts on the Chinese energy system and economy have told Heatmap. The economy isshifting more toward manufacturing and away from the steel-and-cement intensive breakneck urbanization of the past few decades, thanks to a dramatically slowing homebuilding sector.
Chinese urban residential construction was using almost 300 million tons of steel per year at its peak in 2019, according to research by the Reserve Bank of Australia, about a third of the country’s total steel usage. (Steel consumption for residential construction would fall by about half by 2023.) By contrast, the whole United States economy consumes less than 100 million tons of steel per year.
To the extent the overall Chinese economy slows down due to the trade war with the United States, coal usage — and thus greenhouse gas emissions — would slow as well. Although that hasn’t happened yet — China also released export data on Friday that showed sustained growth, in spite of the tariff barriers thrown up by the Trump administration.
All of the awesome earth-moving and none of the planet- or lung-harming emissions.
Construction is a dirty business, literally and figuratively. Mud and gunk and tar come with the territory for those who erect buildings and pave roads for a living. And the industrial machines that provide the muscle for the task run on hulking diesel engines that spew carbon and soot as they work.
Heavy equipment feels like an unlikely place to use all-electric power in order to ditch fossil fuels. The sheer size and intense workload of a loader or excavator means it has enormous energy needs. Yet the era of electric construction equipment has begun, with companies such as Volvo, Komatsu, and Bobcat all now marketing electric dirt movers and diggers. One big reason why: Full-size machines create the opportunity to make construction projects quieter and cleaner — a potentially huge benefit for those that happen in dense areas around lots of people.
Volvo, for example, appeared at last week’s Advanced Clean Transportation Expo in Anaheim, California, primarily to tout its efforts to reduce emissions in the trucking industry via hydrogen-powered semis, electric trucks, and technological refinements to reduce pollution such as nitrous oxide from traditional diesel. But the Swedish brand also trotted out its clean power dirt movers.
The L120 electric loader that is now taking reservations has a lifting capacity of 6 metric tons on pure electric power, making it useful for job sites such as recycling centers and ports. To see such a beast in person — and displayed on pristine convention-center carpet as if it were this year’s Ford Mustang, no less — is an odd and humbling experience that elicits a little-boy level of glee at beholding a big machine. Its bucket, large enough to carry a basketball team, seems to exist on a scale that is too big for battery power, yet Volvo claims the L120 can match the performance of its diesel brethren.
Volvo also brought an electric excavator, the machine used for shoveling out huge bucketfuls of earth. The EC230 Electric is based on the diesel-powered machine of the same name, but with a stack of batteries adding up to 450 kilowatt-hours of capacity and 650 volts of power give the excavator seven to eight hours of runtime on clean electric power.
“Going to the 600-volt battery packs with similar power density that we’re using in [semi] trucks allowed us to take that into the larger construction equipment,” Keith Brandis, VP of policy and regulatory affairs for Volvo North America, told me. “A big breakthrough for us was making sure that the duty cycle — the vibration, the harshness, the temperature extremes — was proven. We have coolant that runs throughout that battery pack, so we precondition the temperatures for very cold starts as well as during very hot temperatures.”
Indeed, the two big boys on display in Anaheim expand Volvo’s lineup of electric construction machines up to seven. The new full-size offerings also take battery power up to a scale needed for serious projects, where it could cut the noise and pollution that emanate from a site. Volvo says its e-machines are already at work on the restoration project in New York City’s Battery Park, at the southern end of Manhattan, where the local government made quiet and clean construction equipment a priority.
Volvo is not alone in this space. Komatsu builds a slate of electric excavators in a variety of sizes leading up to the 20-ton PC210LCE, which the Japanese brand introduced in 2023.
At the smaller end, Bobcat now builds battery-powered mini-loaders and compact excavators. Caterpillar made an EV dump truck a couple of years ago, and more heavy-duty electric machines for industries like mining are on the way.
Although electric loaders and excavators have begun to match the capability of their combustion-powered cousins and have reached a battery runtime that spans a full workday, Volvo and other heavy equipment manufacturers face a few hurdles in convincing more construction companies to go electric. Just like with passenger cars, there is the matter of price. Battery-powered equipment costs more up front, so companies must be convinced that the savings they’ll reap via reduced fuel and maintenance costs will make the electric equipment less expensive in the long run.
And just like with passenger cars, incentives play an outsized role in affordability. Brandis noted that municipalities often have fixed budgets for equipment replacement, which is inconvenient when clean, electric equipment costs substantially more. “We typically rely on purchase incentives or infrastructure incentives, grants, or vouchers that are available,” he said, such as California’s HVIP voucher for zero-emission heavy equipment.
Then there is the construction version of range anxiety, simply ensuring there is enough electricity at any job site to recharge a division of electric loaders. At locations where sufficient electrical infrastructure is already in place, Volvo is helping electric buyers install switchgears, meters, and EV chargers built to talk to the big machines. “It eliminates one other problem point for the customer because we’ve already proven that the operability is there with the equipment,” Brandis told me.
The problem with construction, however, is that sometimes it takes place in remote locations far from easy connections. At ACT, Ray Gallant of Volvo construction equipment said this is the point at which the power has to come to the customer. Volvo recently acquired the battery production business of Proterra, which, among other things, would help the corporation develop battery electric storage solutions that it could deploy remotely — at a far-flung job site, say.
“When we’re in remote sites, we have to take the electrons to the electric machines,” he said.
The lawmakers from opposite parties discussed the Inflation Reduction Act and working together to pass legislation at Heatmap’s Energy Entrepreneurship 2025 event.
Will Republicans’ reconciliation bill successfully gut the Inflation Reduction Act?
A Democratic and Republican senator speaking last week at Heatmap’s Energy Entrepreneurship 2025 event predicted that it will not.
A proposal effectively killing the IRA “wouldn’t make it through the House,” Senator John Curtis of Utah, a Republican, said flatly at the event.
“If you believe that democracy does follow representation, those House members from those states are going to fight like hell to maintain those credits,” Senator John Hickenlooper, a Democrat of Colorado, agreed. He argued that 70% of the credits and benefits in Biden’s flagship climate law go to red states.
“I think you’re going to find enough Republicans push back on the value of these credits that there will be a thoughtful discussion and very careful review of each one. And as you know from the number of people that have spoken up on this, I think we’re in a good place, but that doesn’t mean they won’t be pushed and poked and prodded,” Curtis added, referencing the Republican signatories of letters sent to party leaders urging the preservation of the credits. Curtis and Hickenlooper both were optimistic about the chances of the credits surviving the budget reconciliation underway.
Consensus, not compromise, was the name of the game at Heatmap’s D.C. Climate Week event, which saw Heatmap executive editor Robinson Meyer sit down with the senators to discuss their approach to climate policy and bipartisan collaboration.
Robinson Meyer, Senator John Curtis, and Senator John Hickenlooper.Taylor Mickal Photography,
Curtis and Hickenlooper have worked together on the Co-Location Energy Act, which ensures that wind and solar projects can be developed on land already leased for other types of energy projects, and the Fix Our Forests Act, which emphasizes wildfire mitigation and forest health.
Thursday’s discussion also touched on working with the Trump administration on climate and energy policy. Curtis revealed that he spoke to all of Donald Trump’s nominees, including Chris Wright, about his work in the House on the Conservative Climate Caucus. “They all knew about it, and they all supported it,” he noted, adding that EPA administrator Lee Zeldin was a member of the Caucus when he served in the House.
“I think it's very important for me, for Coloradans, for me to have Chris Wright's cell phone number and be able to talk to him,” Hickenlooper stated, emphasizing that he’s willing to work with the Trump administration to achieve Colorado’s climate goals.
The Co-Location Energy Act was “common sense,” according to Curtis. The act was introduced back in December by himself and Congressman Mike Levin, a Democrat from California. “Two thirds of [Utah] is owned by the federal government, and if you say that’s off the table for development, that’s a huge problem,” he said.
Fix Our Forests, which passed the House in January after being introduced by Congressmen Scott Peters, a Democrat from California and Bruce Westerman, a Republican of Arizona, “is a case study in how we can get things done,” Curtis noted. The key to speaking to conservatives about climate change, he said, is avoiding divisive language, comparing the wrong approach to a coercive time-share presentation. “The salesman says to you, ‘do you love your kids?’ and you feel like you're backed into a corner,” he explained. “I think the way we approach this oftentimes puts Republicans on the defensive.”
Hickenlooper agreed, “You never persuade someone to change their mind about something that really matters by telling them why they’re wrong and why you’re right.”