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 war of attrition is now turning in opponents’ favor.
A solar developer’s defeat in Massachusetts last week reveals just how much stronger project opponents are on the battlefield after the de facto repeal of the Inflation Reduction Act.
Last week, solar developer PureSky pulled five projects under development around the western Massachusetts town of Shutesbury. PureSky’s facilities had been in the works for years and would together represent what the developer has claimed would be one of the state’s largest solar projects thus far. In a statement, the company laid blame on “broader policy and regulatory headwinds,” including the state’s existing renewables incentives not keeping pace with rising costs and “federal policy updates,” which PureSky said were “making it harder to finance projects like those proposed near Shutesbury.”
But tucked in its press release was an admission from the company’s vice president of development Derek Moretz: this was also about the town, which had enacted a bylaw significantly restricting solar development that the company was until recently fighting vigorously in court.
“There are very few areas in the Commonwealth that are feasible to reach its clean energy goals,” Moretz stated. “We respect the Town’s conservation go als, but it is clear that systemic reforms are needed for Massachusetts to source its own energy.”
This stems from a story that probably sounds familiar: after proposing the projects, PureSky began reckoning with a burgeoning opposition campaign centered around nature conservation. Led by a fresh opposition group, Smart Solar Shutesbury, activists successfully pushed the town to drastically curtail development in 2023, pointing to the amount of forest acreage that would potentially be cleared in order to construct the projects. The town had previously not permitted facilities larger than 15 acres, but the fresh change went further, essentially banning battery storage and solar projects in most areas.
When this first happened, the state Attorney General’s office actually had PureSky’s back, challenging the legality of the bylaw that would block construction. And PureSky filed a lawsuit that was, until recently, ongoing with no signs of stopping. But last week, shortly after the Treasury Department unveiled its rules for implementing Trump’s new tax and spending law, which basically repealed the Inflation Reduction Act, PureSky settled with the town and dropped the lawsuit – and the projects went away along with the court fight.
What does this tell us? Well, things out in the country must be getting quite bleak for solar developers in areas with strident and locked-in opposition that could be costly to fight. Where before project developers might have been able to stomach the struggle, money talks – and the dollars are starting to tell executives to lay down their arms.
The picture gets worse on the macro level: On Monday, the Solar Energy Industries Association released a report declaring that federal policy changes brought about by phasing out federal tax incentives would put the U.S. at risk of losing upwards of 55 gigawatts of solar project development by 2030, representing a loss of more than 20 percent of the project pipeline.
But the trade group said most of that total – 44 gigawatts – was linked specifically to the Trump administration’s decision to halt federal permitting for renewable energy facilities, a decision that may impact generation out west but has little-to-know bearing on most large solar projects because those are almost always on private land.
Heatmap Pro can tell us how much is at stake here. To give you a sense of perspective, across the U.S., over 81 gigawatts worth of renewable energy projects are being contested right now, with non-Western states – the Northeast, South and Midwest – making up almost 60% of that potential capacity.
If historical trends hold, you’d expect a staggering 49% of those projects to be canceled. That would be on top of the totals SEIA suggests could be at risk from new Trump permitting policies.
I suspect the rate of cancellations in the face of project opposition will increase. And if this policy landscape is helping activists kill projects in blue states in desperate need of power, like Massachusetts, then the future may be more difficult to swallow than we can imagine at the moment.
And more on the week’s most important conflicts around renewables.
1. Wells County, Indiana – One of the nation’s most at-risk solar projects may now be prompting a full on moratorium.
2. Clark County, Ohio – Another Ohio county has significantly restricted renewable energy development, this time with big political implications.
3. Daviess County, Kentucky – NextEra’s having some problems getting past this county’s setbacks.
4. Columbia County, Georgia – Sometimes the wealthy will just say no to a solar farm.
5. Ottawa County, Michigan – A proposed battery storage facility in the Mitten State looks like it is about to test the state’s new permitting primacy law.
A conversation with Jeff Seidman, a professor at Vassar College.
This week’s conversation is with Jeff Seidman, a professor at Vassar College and an avid Heatmap News reader. Last week Seidman claimed a personal victory: he successfully led an effort to overturn a moratorium on battery storage development in the town of Poughkeepsie in Hudson Valley, New York. After reading a thread about the effort he posted to BlueSky, I reached out to chat about what my readers might learn from his endeavors – and how they could replicate them, should they want to.
The following conversation was lightly edited for clarity.
So how did you decide to fight against a battery storage ban? What was your process here?
First of all, I’m not a professional in this area, but I’ve been learning about climate stuff for a long time. I date my education back to when Vox started and I read my first David Roberts column there. But I just happened to hear from someone I know that in the town of Poughkeepsie where I live that a developer made a proposal and local residents who live nearby were up in arms about it. And I heard the town was about to impose a moratorium – this was back in March 2024.
I actually personally know some of the town board members, and we have a Democratic majority who absolutely care about climate change but didn’t particularly know that battery power was important to the energy transition and decarbonizing the grid. So I organized five or six people to go to the town board meeting, wrote a letter, and in that initial board meeting we characterized the reason we were there as being about climate.
There were a lot more people on the other side. They were very angry. So we said do a short moratorium because every day we’re delaying this, peaker plants nearby are spewing SOx and NOx into the air. The status quo has a cost.
But then the other side, they were clearly triggered by the climate stuff and said renewables make the grid more expensive. We’d clearly pressed a button in the culture wars. And then we realized the mistake, because we lost that one.
When you were approaching getting this overturned, what considerations did you make?
After that initial meeting and seeing how those mentions of climate or even renewables had triggered a portion of the board, and the audience, I really course-corrected. I realized we had to make this all about local benefits. So that’s what I tried to do going forward.
Even for people who were climate concerned, it was really clear that what they perceived as a present risk in their neighborhood was way more salient than an abstract thing like contributing to the fight against climate change globally. So even for people potentially on your side, you have to make it about local benefits.
The other thing we did was we called a two-hour forum for the county supervisors and mayor’s association because we realized talking to them in a polarized environment was not a way to have a conversation. I spoke and so did Paul Rogers, a former New York Fire Department lieutenant who is now in fire safety consulting – he sounds like a firefighter and can speak with a credibility that I could never match in front of, for example, local fire chiefs. Winning them over was important. And we took more than an hour of questions.
Stage one was to convince them of why batteries were important. Stage two was to show that a large number of constituents were angry about the moratorium, but that Republicans were putting on a unified front against this – an issue to win votes. So there was a period where Democrats on the Poughkeepsie board were convinced but it was politically difficult for them.
But stage three became helping them do the right thing, even with the risk of there being a political cost.
What would you say to those in other parts of the country who want to do what you did?
If possible, get a zoning law in place before there is any developer with a specific proposal because all of the opposition to this project came from people directly next to the proposed project. Get in there before there’s a specific project site.
Even if you’re in a very blue city, don’t make it primarily about climate. Abstract climate loses to non-abstract perceived risk every time. Make it about local benefits.
To the extent you can, read and educate yourself about what good batteries provide to the grid. There’s a lot of local economic benefits there.
I am trying to put together some of the resources I used into a packet, a tool kit, so that people elsewhere can learn from it and draw from those resources.
Also, the more you know, the better. All those years of reading David Roberts and Heatmap gave me enough knowledge to actually answer questions here. It works especially when you have board members who may be sympathetic but need to be reassured.