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:
Will the rise of machine learning and artificial intelligence break the climate system? In recent months, utilities and tech companies have argued that soaring use of AI will overwhelm electricity markets. Is that true — or is it a sales pitch meant to build more gas plants? And how much electricity do data centers and AI use today?
In this week’s episode, Rob and Jesse talk to Jonathan Koomey, an independent researcher, lecturer, and entrepreneur who studies the energy impacts of the internet and information technology. We discuss why AI may not break the electricity system and the long history of anxiety over computing’s energy use. Shift Key is hosted by Robinson Meyer, executive editor of Heatmap, and Jesse Jenkins, a Princeton professor of energy systems engineering.
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: Before we go any further — and I think you just hinted at your answer, here, but I want to tackle it directly — which is that I think people look at the hockey stick graphs for AI use, and they look at current energy use for AI, and they look at load growth data coming from the utilities, and they go, “Oh my gosh, AI is going to absolutely overrun our energy system. It’s going to cause emissions to shoot up,” because again, this is just extrapolating from what’s recent.
But of course, part of the whole AI mythos is like, once it starts, you can’t stop it. There is a story out there that, frankly, you see as much from folks who are worried about the climate as you do from AI boosters, which is that very soon, we'’e going to be using a huge amount of energy on AI. And I want to ask you this directly: Should we be worried about AI, number one, overrunning the energy system? Or number two, AI causing a massive spike in carbon emissions that dooms us to, let's say, pass 2.5C that uses up the rest of our carbon budget? Is that something you're worried about? And just how do you think about this?
Jonathan Koomey: Everyone needs to calm the heck down. So we talked about the original baseline, right? So the baseline, data centers are 1% of the world's electricity. And maybe AI now is 0.1%, right? For Google, it’s 0.15%, whatever. But 10% of the 1% is AI.
So let’s say that doubles — let’s say that triples in the next few years, or even goes up fivefold. That gets to about half a percent. So I think it will pale in comparison to the other growth drivers that Jesse was talking about in electrification. Because if you think about light vehicles, if you electrified all light vehicles in the U.S., that’s like a 20% or 25% increase in electricity consumption. And if you did that over 20 years, that’s like 1-ish% per year. Right? So that's, that to me is a very credible thing that’s likely to happen. And then when you add heat pumps, you add industrial electrification, a lot more.
I think there will be local impacts. There will be some places where AI and data centers more generally will be important and will drive load growth, but it is not a national story. It is a local story. And so a place like Ireland that has, I think at last count 17%, 18% of its load from data centers, if that grows, that could give them real challenges. Same thing, Loudoun County in Virginia. But you really do have to separate the national story or the global story from the local story.
Jesse Jenkins: I think it was just about a week ago, Nvidia which is the leading producer of the graphics processing units that have become now the main workhorse chips for generative AI computing, they released their new best-in-class chip. And as they revealed that chip, they — for the first time, it sounded like — started to emphasize the energy efficiency improvements of the GPU. And the basic story the CEO told is that it would take about 73% less electricity and a shorter period of time to train AIs on this new chip than it did on their previous best-in-class chip. So that’s just one generation of GPU with nearly three-quarters reduction in the amount of energy consumed per ... I don't know how you measure the units of large language model training, but per smarts trained into generative AI. So yeah, huge gains.
And one might say, well, can that continue forever? And I guess we should maybe get your thoughts on that. But it has continued at least for the last 10 to 20 years. And so there’s a lot of reason to believe that there’s continued gains to be made.
Koomey: Most people, when they think of efficiency, they think of Moore’s Law. They think of shrinking transistors. And anyone who follows this knows that every year or two, there’s another article about how Moore’s Law is ending, or slowing, or you know, it’s getting harder. And there’s no question about it, it’s absolutely getting harder and harder to shrink the transistors. But it turns out shrinking transistors is only one way to improve efficiency and performance. For a long time, the industry relied on that.
From the early days of microprocessors, starting in ’71, over time, they would ramp up the clock speed. And at the same time, they would ramp down the voltage of the chip. And that was called Dennard scaling. It allowed them to keep ramping up performance without getting to crazy levels of leakage current and heat and melting the chip and the whole thing. That worked for a long time, til the early 2000s. And then they hit the threshold voltage for silicon, which is like one volt. So once you hit that, you can no longer do that trick. And they needed new tricks.
So what they did was they, most of you remember who were around at that time, there was this big shift to multiple cores on a chip. That was an innovation in hardware architecture that allowed them, for a time, to improve efficiency by going to software that could run on multiple cores, so you could multiprocess various activities. So that’s one way you can improve things. You can also work on the software — you can improve the efficiency of the software, you can improve the algorithms that you use.
So even if Moore's law shrinkage of transistors stops, which it hasn’t fully stopped. But even if it did, there are a lot of other things we can do. And AI in particular is relatively new. Basically, people threw a whole bunch of money at existing processors because there was this rush to deploy technology. But now, everyone’s stepping back and saying, well, look at the cost of the energy cost and the infrastructure cost. Is there a way to do this better? And sure, there definitely is, and Nvidia proved it in their presentation that you referred to.
This episode of Shift Key is sponsored by…
KORE Power provides the commercial, industrial, and utility markets with functional solutions that advance the clean energy transition worldwide. KORE Power's technology and manufacturing capabilities provide direct access to next generation battery cells, energy storage systems that scale to grid+, EV power & infrastructure, and intuitive asset management to unlock energy strategies across a myriad of applications. Explore more at korepower.com.
Watershed's climate data engine helps companies measure and reduce their emissions, turning the data they already have into an audit-ready carbon footprint backed by the latest climate science. Get the sustainability data you need in weeks, not months. Learn more at watershed.com.
Music for Shift Key is by Adam Kromelow.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
It sure looks that way, at least. Democrats should start coming up with a plan.
For the first six months of President Trump’s term, the big question was about what would happen to the Inflation Reduction Act. We now have something like an answer.
President Trump’s memorably named One Big Beautiful Bill Act repealed many of the IRA’s most important clean energy tax credits, including incentives for wind, solar, and electric vehicles. And while it’s still unclear whether the Trump administration will let developers actually use the tax credits that remain on the books — especially the now-denuded credits for wind and solar — fewer “unknown unknowns” remain about what might come next.
So I’ve been trying to figure out where climate and energy policy might go from here. And one story that I keep coming back to is the flashing red lights around what could become a serious electricity affordability crisis.
It’s now widely understood that electricity demand is rising in the United States for the first time in a generation. The Energy Information Administration projects that electricity use will grow 1.7% in the next few years, after increasing by just 0.1% per year from 2005 to 2020. That growth is projected to come from new data centers, new factories, the (now) slow(er) but (still) steady adoption of electric vehicles, and population growth.
What is less well understood is how poorly the United States is prepared to match this rise in electricity demand with an equivalent increase in supply. To some degree, American electricity prices are already rising: So far this year, utilities have received or requested permission to increase customers’ bills by $29 billion, according to a July report from PowerLines, a think tank and advocacy group. That’s a large number in its own right, and it’s more than twice as much as had been approved at this time last year.
But when you look across the power system, virtually every trend is setting us up for electricity price spikes:
On top of all this, of course, the Trump administration has made it much more uncertain which new solar, wind, and battery projects will be able to secure tax credits — and with them, secure bank financing.
None of these trends alone would guarantee price increases or electricity supply constraints. But taken together, they reveal an electricity system that is coming under a variety of strains.
In the 2010s, cheap natural gas and technological advances in energy efficiency pacified much of the power system. We won’t have the same luxury this decade.
This is all going to be bad for the economy, bad for the climate, and bad for climate policy.
It’s a setback for the U.S. economy because, as President Trump somewhat alluded to in his second inaugural address, energy is a key input to virtually every other economic process, including manufacturing. But it’s especially bad for climate policy. The dominant plan to decarbonize much of the U.S. economy is to “electrify everything” — cars, appliances, home heating, and even many industrial processes. Americans will be far less eager to electrify everything if electricity is expensive.
If energy price hikes do arrive, Democrats are going to have a relatively straightforward time communicating about them in a narrow political sense. The story is just too simple: Democrats passed a law to encourage clean energy called the Inflation Reduction Act. Republicans repealed it. Energy prices inflated. QED.
That story alone might be too contrived, but the evidence we have suggests that OBBBA will raise energy bills. The REPEAT Project at Princeton University — led by Jesse Jenkins, my Shift Key podcast cohost — has a new report out projecting that the One Big Beautiful Bill Act will increase Americans’ electricity bills by $165 a year by the end of the decade. (If the law is allowed to stick around, and in the absence of intervening policies, it could raise bills by hundreds of dollars a year by the middle of next decade.)
OBBBA’s explosion of the federal deficit will make the situation worse: By expanding the deficit for such little public gain — that is, merely to memorialize earlier tax cuts, not even to make new ones — the Federal Reserve will have a more difficult time cutting interest rates in the future. That will in turn make it even more difficult for utilities and developers to finance new energy projects.
The political story will be so compelling here, I think, that Democrats will come under a lot of pressure to reinstate the wind and solar tax credits. And maybe they should do that — it could make sense as part of a larger energy or permitting deal. But stacking more solar and wind on the grid will not on its own lower people’s electricity bills.
Going into 2028, Democrats will need an actual plan to stabilize or cut electricity costs. They will need ideas about how (and whether) to speed up permitting, restructure wholesale power markets, and build new power plants in order to stabilize the power grid.
One thing that’s already clear is that in this inflationary environment, states like New York with publicly owned power authorities are able to intervene more forcefully in their own power markets than states that lack such capability. That’s because the state itself can act to build its own large-scale power plants. New York Governor Kathy Hochul recently directed the state’s power authority to build a new nuclear power plant upstate in order to grow the supply of zero-emissions electricity. Using their state own power authorities, governors in other states — or even the federal government, with an entity like the TVA— could take a similar step.
With all that said, I’ve been trying to come up with a scenario under which these price hikes will not materialize. In the late 2010s, for instance, America’s liquified natural gas exports surged essentially from zero, but domestic consumers didn’t see significant price hikes because drillers increased gas production to match the exports. Maybe that could happen again. And maybe utilities will — and this would, to be clear, be horrible for the climate — run their aging coal plants much more than they once anticipated doing.
Or maybe load growth won’t be as bad as we think. When Jesse and I spoke to Peter Freed, Meta’s former director of energy strategy, for Shift Key, he told us that the current data center boom is different from any previous buildout because of the presence of speculators. For the first time, he said, speculative data center developers are buying up prospective sites and requesting utility-scale hookups with the expectation that they will find a tenant for the data center in the future. In other words, the demand side of the electricity system is filled with an unusual amount of froth at the moment.
We also know that, more generally, the demand side of the power system is a mess. In the past few years, climate analysts have gotten used to talking about the power grid’s interconnection queue — that is, its supply side. But the demand-side queue — the process that lets new data centers, factories, and other new electricity users connect — is even more broken. In some jurisdictions, it’s little more than an Excel file that projects move up and down within as local politics requires.
We also know that one source of new demand — one planned factory or, more often, one data center — will sometimes apply to hook up to multiple states or utilities at the same time. It will get utilities to bid against each other, suss out the best construction sites and power rates, and only relatively late in the process make a final decision about where to build.
So if I were putting together a bear case for electricity demand, I would start here. Maybe aggressive data center speculators are bidding in multiple utilities, driving up projections across many states. That’s causing utilities to freak out about their supply, leading them to project the need for a lot of new investment — and, with it, a lot of electricity rate increases. But as data center speculators actually begin to build (or abandon) projects — and as some of the air inevitably comes out of the AI boom — some of this projected demand will start to evaporate. Perhaps the data centers that do get built will find ways to reduce their power usage, too.
Even this story won’t fully eliminate load growth on its own, though. Data centers make up the largest share of new electricity demand, but even then, they’re not the majority of it. The rest comes from, roughly, new factories, the slow electrification of the vehicle fleet, and new residential construction. But let’s say the One Big Beautiful Bill Act succeeds in hobbling the electric vehicle sector in the United States, many EV and battery factories get canceled, and fewer Americans buy EVs overall. Calculate in a mild recession, too, since all the AI and EV investment will be drying up.
In that world, most new sources of power demand really will be in abeyance. That’s how some of these power projections might not come true. But in most other scenarios, it’s time to hold on — and for blue-state leaders to think about how they can find cheap, zero-emissions electrons, as soon as possible.
The Department of Energy announced Wednesday that it was scrapping the loan guarantee.
The Department of Energy canceled a nearly $5 billion loan guarantee for the Grain Belt Express, a transmission project intended to connect wind power in Kansas with demand in Illinois that would eventually stretch all the way to Indiana.
“After a thorough review of the project’s financials, DOE found that the conditions necessary to issue the guarantee are unlikely to be met and it is not critical for the federal government to have a role in supporting this project. To ensure more responsible stewardship of taxpayer resources, DOE has terminated its conditional commitment,” the Department of Energy said in a statement Wednesday.
The $11 billion project had been in the works for more than a decade and had won bipartisan approval from state governments and regulators across the Midwest. The conditional loan guarantee announced in November 2024 would have secured up to $4.9 billion in financing to fund phase one of the project, which would run from Ford County in Kansas to Callaway County in Missouri.
In response to a request for comment, an Invenergy spokesperson said, “While we are disappointed about the LPO loan guarantee, a privately financed Grain Belt Express transmission superhighway will advance President Trump’s agenda of American energy and technology dominance while delivering billions of dollars in energy cost savings, strengthening grid reliability and resiliency, and creating thousands of American jobs.”
The project had long been the object of ire from Missouri Senator Josh Hawley, who recently stepped up his attacks in the hopes that a more friendly administration could help scrap the project. Two weeks ago, Hawley posted on X that he’d had “a great conversation today with @realDonaldTrump and Energy Secretary Chris Wright. Wright said he will be putting a stop to the Grain Belt Express green scam. It’s costing taxpayers BILLIONS! Thank you, President Trump.” The New York Times later reported that Trump had made a call to Wright on the issue with Hawley in the Oval Office.
Hawley celebrated the Grain Belt Express decision, writing on X, “It’s done. Thank you, President Trump,” and exulting in a separate post that “Department of Energy officially TERMINATES taxpayer funding for Green New Deal ‘grain belt express.’”
The senator had claimed that the plan would hurt Missouri farmers due to the use of eminent domain to acquire land for the project. In 2023, Hawley wrote a letter to Invenergy chief executive Michael Polsky claiming that “your company’s Grain Belt Express construction campaign has hurt Missouri’s farmers,” and that “they have lost the use of arable land, seen their property values decline, and been forced to operate under a cloud of uncertainty.”
Controversy over eminent domain and the use of agricultural land by transmission lines illustrates the difficulties in building the long-distance energy infrastructure necessary to decarbonize the grid.
Opposition to the project had been gestating for years but picked up steam in recent weeks. Earlier this month, Andrew Bailey, the Republican attorney general of Missouri, announced an investigation into the project. “This is a HUGE win for Missouri landowners and taxpayers who should not have to fund these green energy scams,” he wrote on X Wednesday following the DOE’s announcement.
As the project appeared to be more imminently imperiled, Invenergy scrambled to preserve its future, including making plans to connect gas to the transmission line. In a letter to Secretary of Energy Chris Wright written earlier this month, the Invenergy vice president overseeing the project wrote that the Grain Belt Express “has been the target of egregious politically motivated lawfare,” echoing language President Trump has used to describe his own travails.
If the author’s intent was to generate sympathy from the administration, it didn’t work. The end of the loan guarantee could be a death blow to the project, and will at the very least force Invenergy into a mad dash to try to match the lost capital.
Editor’s note: This story has been updated to include a comment from Invenergy.
The grant from Washington State will fund a facility where all kinds of fusion labs can run tests of their own.
Flash back to four summers ago, when aspiring fusion pioneers Robin Langtry and Brian Riordan were stuck designing rockets at Blue Origin, Amazon CEO Jeff Bezos’ aerospace and space tourism company. More specifically, they were ruminating on how their engine’s large size was preventing the team from iterating quickly.
“If your rocket engine is 12 feet tall, there’s like, three places in the country where you can get castings,” Langtry told me. One simple design change could mean another eight to nine months before the redesigned part came in. Smaller designs, they hypothesized, would lead to faster development cycles.
They decided to quit their jobs in June of 2021 and put their thesis to the test with what would become Avalanche Energy, a fusion startup aiming to commercialize tabletop-sized reactors via magneto-electrostatic fusion, a nascent technology that’s far less well-understood than even still-experimental large-scale fusion machines like tokamaks and stellarators. Today, though, Washington State is giving this emergent tech a big vote of confidence by announcing one of the largest government-led fusion investments to date: A $10 million grant for Avalanche to build out a commercial-scale test facility for fusion technologies.
This facility, called FusionWERX, is where Avalanche will test its own prototypes with the goal of achieving scientific breakeven — the point at which a fusion reaction produces more energy than the energy used to initiate the reaction. But as Langtry, the company’s CEO, explained to me, it will also be a hub where other fusion companies, universities, and national labs can come test their own proprietary technologies while keeping their intellectual property intact.
“It’s almost like a commercial wind tunnel test facility, but for fusion,” Langtry told me. For example, Avalanche’s early-stage reactors will produce neutrons that researchers can use to test novel materials and ensure they can withstand the extreme conditions found inside fusion reactors. Organizations can also test their own neutron capture methods, often referred to as "neutron blankets,” which are critical for producing the tritium fuel that’s needed for a sustained fusion reaction.
Thus, Avalanche will earn revenue from the groups using the FusionWERX facility well before it makes any money from commercial energy production. The startup also plans to bring in additional income by making and selling radioisotopes — atoms that emit radiation as they decay — for medical and energy applications such as diagnostic imaging, radiation therapy, and nuclear batteries that can generate electricity in space or remote areas like the deep ocean.
Langtry told me these additional opportunities make Avalanche attractive to a wider variety of investors than simply climate tech venture capitalists interested in fusion’s potential for utility-scale power generation. “There’s much bigger sources of capital if you can build a true business that commercializes this technology and generates revenue and scales it,” Langtry told me. “That’s really what we’re about.”
Prior to the $10 million grant, Avalanche had raised a total of $50 million from investors such as Lowercarbon Capital, Peter Thiel’s Founders Fund, and Toyota Ventures. And while the startup’s lineup of near-term use cases sets it apart, Avalanche too is ultimately aiming to produce commercially-relevant energy, with an eye towards replacing diesel generators for data center backup power or for use in remote communities or military outposts.
Avalanche’s chosen method, magneto-electrostatic fusion, uses ions that are injected into the reactor’s chamber and confined with extremely high voltage. This strong electric field accelerates the ions towards the center of the reactor, where they collide to produce a fusion reaction. Magnets surrounding the chamber also work to trap electrons alongside the ions, increasing the density of the plasma to achieve high fusion rates.
Avalanche announced today that it has successfully operated its machine at 300 kilovolts for multiple hours. When adjusted for size, this equates to 6 megavolts per meter, twice the voltage density of lightning. To reach breakeven, the company will need to operate its machine at about 700 kilovolts, which Langtry told me can be done by doubling the size of the reactor’s radius from 6 centimeters to 12 centimeters. Avalanche said in a follow up email that the company is waiting to gain operational experience at its current scale before raising the capital it will take to build a larger reactor.
The magneto-electrostatic method is well-suited to micro reactors as it doesn’t rely on giant magnets or lasers to create the fusion reaction. Ultimately, Avalanche plans to produce modular reactors from 5 kilowatts to 1 megawatt in size — enough to power just a couple homes at the least, and about 1,000 homes at the most.
But powering homes isn’t what Avalanche will actually do. Before energy dominance was even in vogue, the company was already focused on military applications for its tech. It received a contract from the Department of Defense’s Defense Innovation Unit in 2022 to develop technology for a nuclear-powered spacecraft by 2027. Avalanche did not elaborate on what its initial prototype might look like or be used for, only writing in a follow-up email that it’s “in active discussions about next steps for maturing this technology with DOD.”
“We were sort of contrarian, in that we always thought our path to commercial operations was through DOD and space, whereas most of the fusion companies were raising on climate and clean energy and building massive clean energy power plants,” Langtry told me. He cited support from Thiel, perhaps Silicon Valley’s most influential conservative voice, as helping influence the company’s direction.
At this moment, Langtry told me, there’s excitement around using Avalanche’s tech to make President Trump’s vision of a so-called “Golden Dome” missile defense system a reality. This would involve using satellites — theoretically powered by Avalanche — that could track and shoot down ballistic missiles fired at the U.S. “Right now, with solar, [satellites] could probably only take one shot during an engagement. But if you had 100 kilowatts or a megawatt, you could shoot continuously, and that system would be a lot more capable,” Langtry explained to me.
Depending on your feelings about nuclear war, this vision may bring more anxiety than comfort. It’s also a far cry from the more typical — and endlessly more idyllic — narrative of limitless clean energy and unprecedented prosperity that I’m used to hearing fusion enthusiasts promote. But such is the moment. And if the path to commercial fusion ends up running through a satellite-powered missile defense system, it probably won’t be the weirdest clean energy story of the Trump era.