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.
Lower borrowing costs aren’t enough to erase the threat of tariffs and Trump.
It won’t rescue the renewables industry, but at least it’s something.
The Federal Reserve announced today that it will cut the federal funds rate by 0.25 percentage points, bringing it down to between 4% and 4.25%. Fed officials also projected quarter-point rate cuts at the last two meetings of the Federal Open Markets Committee this year.
This may provide some relief to renewables developers and investors, who are especially sensitive to financing costs. “On the financing side, high rates are never going to be exactly a good thing,” Advait Arun, a climate and infrastructure analyst at the Center for Public Enterprise, told me. “I think in this case, it’s going to be good that we’re finally seeing cuts.”
Because the fuel for solar and wind energy is essentially free, the lion’s share of the cost to develop these energy sources comes up front, meaning that interest rates can have a disproportionate effect on how projects pencil out. Renewable projects also tend to carry more debt than fossil fuel projects, according to energy consultancy Wood Mackenzie. When interest rates rise by 2 percentage points, the consultancy estimated, the levelized cost of electricity for renewables rises by 20%, compared to 11% for a gas-fired power plant, which might have higher operating costs but less need to borrow.
But the challenges for the renewables industry go well beyond financing. Developers are still wondering how they will be able to use Chinese-linked components without losing eligibility for clean energy tax credits. Those tax credits now come with a ticking clock after the passage of this summer’s One Big Beautiful Bill Act, which shortened the eligibility period for wind and solar projects. The Treasury Department also tightened the definition of what it means to “start construction,” making qualification even more of a race. All the while, the Trump administration’s regulatory assault on the sector, especially wind, has led to project cancellations across the industry.
“High interest rates obviously impact the business, but there are a lot of other headwinds and other things going wrong, as well,” Gautam Jain, a senior research scholar at the Center on Global Energy Policy at Columbia University, told me. “If anything, compared to the beginning of the year, rates have come down quite a bit.”
Maheep Mandloi, an analyst at the investment bank Mizuho Securities, wrote in a note to clients that renewable stocks rose last week in part because investors saw yields falling on 10-year government bonds. Ten-year Treasuries are a widely used benchmark for corporate debt, and when they get cheaper, it often means that companies can access financing more cheaply.
Falling 10-year yields are also a sign that the market anticipates a Fed rate cut. So far this year, the 10-year Treasury bond yield has fallen from 4.57% to 4.00% as of Wednesday afternoon after the rate cut was announced.
Lower borrowing costs are a welcome transition for the industry. Borrowing costs started to rise dramatically in 2022, as the Fed hiked interest rates to combat the worst inflation the U.S. had seen since the early 1980s. Annual price increases had been bouncing around or even below 2% since the 2008 recession before climbing to as high as 9% in the summer of 2022, following Russia’s invasion of Ukraine, which led to an energy price shock. The uneven and stimulus-fueled economic recovery from Covid-19 also created price instability throughout the economy, including the renewable energy industry.
Renewable energy businesses in particular were hammered by higher interest rates, as well as higher costs for commodities like steel and for final products like solar panels.
Even as unprecedented government support flowed into the renewables industry from the Inflation Reduction Act, signed in August 2022, clean energy stocks continued to stagnate, with the iShares Clean Energy ETF falling over 30% from the beginning of the Biden administration through the end of 2023. (Despite the assault from the Trump administration, the index has actually risen about 30% so far this year after falling in the fall and winter of 2024, as uncertainty around the IRA’s tax credits has dissipated.)
One of the poster children for renewables dysfunction is the Danish wind developer Orsted, which has been a victim of just about every brickbat thrown at the industry. In its most recent financial statement, the company said that its future earnings estimates were imperiled by “assumptions with major uncertainty,” which included “investment tax credits, interest rates, imposed tariffs in the U.S., and the supply chain.”
Home solar giant Sunrun, too, has cited financing stresses. In its most recent quarterly report, the company disclosed that “rising interest rates, including recent historic increases starting in 2021 … [are] reducing the proceeds we receive from certain Funds.” It also acknowledged that “because our financing structure is sensitive to volatility in interest rates, higher rates increase our cost of capital and may decrease the amount of capital available to us to finance the deployment of new solar energy systems.” High rates, the company disclosed, “have impacted and may continue to impact our business and financial results.”
Even as rates come down, the renewable industry still has the Trump administration to contend with. The various agencies of the executive branch have shown little hesitation about getting in the way of renewable energy development, even for projects that are already nearly complete. The Treasury Department also has yet to issue guidance on complying with OBBBA’s rules about sourcing from Chinese suppliers, prolonging uncertainty for many in the industry. Trump’s tariff policy, too, remains a potential wildcard, as developers await a Supreme Court ruling on the legality of the president’s efforts thus far.
“In terms of being able to build more supply with the benefit of lower financing costs,” Arun told me, “I think this is where we’re running into all of the issues with delays in procuring components — the uncertainty regarding whether the tariffs will be struck down or not, and of course, changes to the inflation Reduction Act through the OBBBA.”
Last week, analysts at Rhodium Group projected that Trump’s policies could slow U.S. progress on reducing emissions by more than half.
For renewables developers, the rate cuts may be welcome, but everything else — and there’s a lot of everything else — may be what really matters, Jain told me. “All those things add additional uncertainty, and anybody who’s in the space will be aware that more could come,” he said. “Of course, lower rates will help, but it’s a combination of the two.”
On Democrats’ AI blueprint, more nationalized minerals, and the GOP’s anti-geoengineering push
Current conditions: Tropical Storm Mario is lashing the southwestern U.S. with rainstorms and potential flash flooding • The drought in the Northeast and the Ohio Valley is worsening, with rain deficits in major cities 15% below average • Tropical Cyclone Mirasol is bringing heavy rains to the Philippine island of Luzon.
The Trump administration announced a lawsuit Tuesday aimed at tanking Vermont’s Climate Superfund Act, which set up the nation’s first program to force fossil fuel companies to pay for adaptations to deal with the effects of warming temperatures. The Department of Justice said the legislation “will likely” impose “billions of dollars in liability on foreign and domestic energy companies for their alleged past contributions to climate change.” The motion, filed on Monday, comes months after the Justice Department filed an initial complaint in May targeting the law and similar legislation in New York, Hawaii, and Michigan.
“Like New York, Vermont is usurping the federal government’s exclusive authority over nationwide and global greenhouse gas emissions,” Acting Assistant Attorney General Adam Gustafson said in a press release. “More than that, Vermont’s flagrantly unconstitutional statute threatens to throttle energy production, despite this administration’s efforts to unleash American energy. It’s high time for the courts to put a stop to this crippling state overreach.”
Arizona Senator Mark Kelly. Chip Somodevilla/Getty Images
Arizona Senator Mark Kelly released a proposal Wednesday morning designed to give Democrats a roadmap to back the buildout of data centers to support the boom in artificial intelligence. The 16-page pitch makes no mention of novel tools grid operators are considering to force data centers to dial back electricity consumption when power supply is low, known as demand response. But the proposal does call for establishing a pipeline of projects to support large-scale clean electricity production from 24/7 sources. “While solar and battery storage dominate today’s pipeline, they alone can’t reliably power the AI,” the blueprint reads. “We must build an innovation pipeline for geothermal, nuclear, and other clean dependable sources, while also deploying near-term solutions that advance and strengthen our energy systems for the demands ahead.”
The value of finding ways to add more data centers before that large new power output is available is the big reason for supporting the curtailment of electricity usage at big server farms, Heatmap’s Matthew Zeitlin wrote last month. “Creating a system where data centers can connect to the grid sooner if they promise to be flexible about power consumption would require immense institutional change for states, utilities, regulators, and power markets.”
Get Heatmap AM directly in your inbox every morning:
The U.S. government is in talks to set up a multibillion-dollar fund for overseas mining projects to help counter China’s grip over the world’s critical mineral supply, the Financial Times reported. The Trump administration is discussing the effort with the New York investment firm Orion Resource Partners, and looking to establish the fund under the U.S. International Development Finance Corporation. The fund would invest in projects to produce minerals such as copper and rare earths. “These talks really show that the [Donald] Trump administration is trying to align its financial tools with its broader mineral ambitions,” Gracelin Baskaran, director of the critical minerals security programme at the Center for Strategic and International Studies in Washington, told the newspaper. “This public-private partnership stands to catalyze a significant amount of capital.”
The move is the latest effort by the Trump administration to take on a bigger role in the mining industry, which requires high upfront costs and years-long development timelines that pose problems for companies beholden to quarterly shareholder updates. In July, the Department of Defense took an ownership stake in MP Materials, the only active rare earths producer in the U.S., marking the most significant federal intervention in the private sector since Washington nationalized railways during World War I. In a sign of the dealmaking environment, Heatmap’s Katie Brigham wrote this month that “everybody wants to invest in critical mineral startups.”
The House of Representatives held a hearing Tuesday on the risks posed by weather modification and geoengineering technologies. Led by Georgia Republican Representative Marjorie Taylor Greene, the hearing — entitled “Playing God with the Weather — a Disastrous Forecast” — examined the idea of manipulating the makeup of the atmosphere to artificially cool the planet, which is an emerging, if hotly contested, idea among some commercial startups. GOP officials such as Greene and Secretary of Health and Human Services Robert F. Kennedy, Jr., have raised concerns over what such technology could do. The issue took on a new partisan valence after the flash flooding that killed more than 135 people in Texas this summer, which Fox News suggested could be linked to cloud-seeding experiments underway in the region.
In his testimony, Christopher Martz, a meteorologist and policy analyst at the Committee for a Constructive Tomorrow, warned that there were still major uncertainties about the potential deployment of geoengineering technologies. At times, however, the questioning devolved into debates over the reality of settled science about the effects of fossil fuel emission on warming itself.
“Did man create the Ice Age?” Greene asked Martz at one point.
“No,” he responded.
“Yeah, right, so none of us were alive back then to know for sure,” she said.
Solar developer PosiGen is planning to pull out of three of its projects in Connecticut. The company told state officials late last month it would need to shut down its facilities, eliminating 78 jobs, as financing dried up for the projects. The move highlighted the challenges ahead for the solar industry as federal tax credits barrel toward next year’s phaseout deadline. In 2015, the Connecticut Green Bank helped fund low-and moderate-income homeowners’ purchase of solar panels through PosiGen. But the federal program backing the effort, known as Solar for All, is set to unwind under the Trump administration. The company expects to start laying off workers in Connecticut next week, according to the news site CT Insider.
Robert Redford died Tuesday at 89 years old. During his lengthy career and filmography, the actor fashioned himself as an activist voice for a number of causes, including the U.S. effort to decarbonize its electrical sector. In February 2016, after the Supreme Court paused the Obama administration’s Clean Power Plan, Redford accused the conservative justices of rendering a verdict “on the wrong side of history” in an op-ed in Time magazine. “It was a clear departure from how our courts normally handle government oversight. And I cringe at how we will have to answer to history. When our children and their children ask, ‘When the majority of Earth’s citizens — its scientists, military professionals, industrialists, and more — realized the threat of climate change was real, why didn’t you do more? Why did you delay?’”
Rob talks with Sarah Kapnick about our new era of energy insecurity.
We live in a new energy era — one in which the inputs and technologies key to clean electricity production are at the heart of international politics. What will that mean for decarbonization? And how should climate tech companies prepare?
On this week’s episode of Shift Key, Rob chats about those questions and more with Dr. Sarah Kapnick. She is the Global Head of Climate Advisory at J.P. Morgan, where she advises the bank's clients on climate, energy, biodiversity and sustainability topics. She was the former chief scientist at the National Oceanic and Atmospheric Administration from 2022 to 2024, and was previously a research scientist at NOAA’s Geophysical Fluid Dynamics Laboratory in Princeton, New Jersey.
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. Jesse is off this week.
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: When companies come to you looking for help navigating this particular moment — where federal policy is quite up in the air, where rates are coming down but kind of high, AI capex is surging — what advice do you give them for navigating this moment?
Sarah Kapnick: The advice that I give them is looking to some of those things that strategically are likely to have more consistency over time, and that they’re looking for those places of more consistency, and that they feel that they can invest in, that they will have support ongoing — particularly if it’s a project that lasts beyond administrations.
They’re really concerned with what they think is going to last. And then for the stuff that doesn’t, that there may be more volatility, they want to identify that volatility, and they want to think through, okay, how can I take opportunity now if I think there’s a small window for it? Or how do I plan for taking opportunity when the opportunity presents itself down the line?
And so, it’s a mixture of long-term planning and thinking through, strategically, where the world is headed and where they can fit in over time, yet also taking opportunities that either present themselves now or they have conviction that will present themselves soon, and then being ready to be the first when that opportunity presents themselves so that they can run with it.
Mentioned:
The New Map of Energy and Geopolitics
Previously on Shift Key: How China’s Industrial Policy Really Works
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.
Music for Shift Key is by Adam Kromelow.