Sign In or Create an Account.

By continuing, you agree to the Terms of Service and acknowledge our Privacy Policy

Technology

What Does OpenAI’s New Breakthrough Mean for Energy Consumption?

Why the new “reasoning” models might gobble up more electricity — at least in the short term

A robot with a smokestack coming out of its head.
Heatmap Illustration/Getty Images

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.

Blue

You’re out of free articles.

Subscribe today to experience Heatmap’s expert analysis 
of climate change, clean energy, and sustainability.
To continue reading
Create a free account or sign in to unlock more free articles.
or
Please enter an email address
By continuing, you agree to the Terms of Service and acknowledge our Privacy Policy
Energy

Who Will Pay for Data Centers’ Energy? Not You, Utilities Say.

Utilities are bending over backward to convince even their own investors that ratepayers won’t be on the hook for the cost of AI.

A winking salesman.
Heatmap Illustration/Getty Images, Related Digital

Utilities want you to know how little data centers will cost anyone.

With electricity prices rising faster than inflation and public backlash against data centers brewing, developers and the utilities that serve them are trying to convince the public that increasing numbers of gargantuan new projects won’t lead to higher bills. Case in point is the latest project from OpenAI’s Stargate, a $7-plus-billion, more-than-1-gigawatt data center due to be built outside Detroit.

Keep reading...Show less
Green
Spotlight

Is North Dakota Turning on Wind?

The state formerly led by Interior Secretary Doug Burgum does not have a history of rejecting wind farms – which makes some recent difficulties especially noteworthy.

Doug Burgum.
Heatmap Illustration/Getty Images, Library of Congress

A wind farm in North Dakota – the former home of Interior Secretary Doug Burgum – is becoming a bellwether for the future of the sector in one of the most popular states for wind development.

At issue is Allete’s Longspur project, which would see 45 turbines span hundreds of acres in Morton County, west of Bismarck, the rural state’s most populous city.

Keep reading...Show less
Yellow
Hotspots

Two Fights Go Solar’s Way, But More Battery and Wind Woes

And more of the week’s top news about renewable energy conflicts.

The United States.
Heatmap Illustration/Getty Images

1. Staten Island, New York – New York’s largest battery project, Swiftsure, is dead after fervent opposition from locals in what would’ve been its host community, Staten Island.

  • Earlier this week I broke the news that Swiftsure’s application for permission to build was withdrawn quietly earlier this year amid opposition from GOP mayoral candidate Curtis Sliwa and other local politicians.
  • Swiftsure was permitted by the state last year and given a deadline of this spring to submit paperwork demonstrating compliance with the permit conditions. The papers never came, and local officials including Sliwa called on New York regulators to reject any attempt by the developer to get more time. In August, the New York Department of Public Service gave the developer until October 11 to do so – but it withdrew Swiftsure’s application instead.
  • Since I broke the story, storage developer Fullmark – formerly Hecate Grid – has gone out of its way to distance itself from the now-defunct project.
  • At the time of publication, Swiftsure’s website stated that the project was being developed by Hecate Grid, a spin-off of Hecate Energy that renamed itself to Fullmark earlier this year.
  • In a statement sent to me after the story’s publication, a media representative for Fullmark claimed that the company actually withdrew from the project in late 2022, and that it was instead being managed by Hecate Energy. This information about Fullmark stepping away from the project was not previously public.
  • After I pointed Fullmark’s representatives to the Swiftsure website, the link went dead and the webpage now simply says “access denied.” Fullmark’s representatives did not answer my questions about why, up until the day my story broke, the project’s website said Hecate Grid was developing the project.

2. Barren County, Kentucky – Do you remember Wood Duck, the solar farm being fought by the National Park Service? Geenex, the solar developer, claims the Park Service has actually given it the all-clear.

Keep reading...Show less
Yellow