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:
There’s a lot more big talk than reactor-building going on.

America’s technology companies need power, and lots of it.
Artificial intelligence combined with still-growing internet and smartphone use will likely require a game-changing investment in data centers — one that its already showing up in huge projected increases for electricity demand across the country. At the same time, many technology companies want to procure and invest in clean power, while many states have clean energy goals that may make it difficult to add new load to the grid without a corresponding investment in clean generation. All told, the Department of Energy estimates that some 700 to 900 gigawatts of new clean firm capacity — energy generation that doesn’t emit greenhouse gases and can run 24 hours a day — will be necessary to build a fully decarbonized grid. Even in the real world, technology companies are interested in acquiring whatever clean power they can.
This is where the nuclear industry would love to step in, specifically the segment of the industry making small modular reactors, otherwise known as SMRs. These reactors, which promise to be cheaper, smaller, and faster to build than the existing nuclear fleet, seem like an ideal match for what technology companies need. What could be better for data centers than on-site power (meaning no transmission costs) that runs all day (meaning no intermittency issues) with no carbon emissions (meaning no climate worries)? And if those nuclear power plants could be built quickly and cheaply out of pre-fabricated parts, all the better, right?
Whether SMRs actually can step in, well ... “If I had every agreement in principle SMRs have signed, I could walk from here to Europe without getting my feet wet,” Dan Yurman, the publisher of Neutron Bytes and a former project manager at the Idaho National Laboratory, told me.
The issue is that the most optimistic timeline for commercial deployment of SMRs starts in the late 2020s, with most observers putting actual deployment into sometimes in the 2030s. All the while, demand for data centers is growing now and is projected to accelerate sharply in the next few years.
As of today only a handful of small modular reactors are currently operational anywhere in the world, and none in the United States. The Nuclear Regulatory Commission, which governs all civilian nuclear construction in the country, has so far approved just one SMR design; NuScale, the company behind said design, recently laid off almost a third of its employees after its deal to build a power plant in Utah for a collection of local utilities fell through due to rising costs.
That approval process cost $500 million and took around five years, according to the Wall Street Journal — and, of course, NuScale has yet to get a functioning reactor out of it. The company is currently in the process of getting the go-ahead on a more powerful version of its existing design, which the company’s chief executive said could be approved “within 24 months.”
On paper, however, enthusiasm for co-locating SMRs with data centers and industrial sites abounds. Despite the collapse of the Utah project, during an earnings call this month, NuScale eagerly talked up a partnership with Standard Power to provide 2 gigawatts of electricity to data centers in Ohio and Pennsylvania. While its shares are down around 50% for the past 12 months, they are up about 35% (albeit to around $4.20) since the end of last year. In its presentation to investors, NuScale cited estimates that data center electricity consumption would triple by the beginning of the next decade.
“Management is quite enthusiastic around its opportunity with data center operators, noting that it's in discussions with large players as electricity demand accelerates via the AI buildout,” Ryan Pfingst and Chris Souther, two analysts for B. Riley Securities, wrote in a note to clients following the release of NuScale’s earnings report.
That enthusiasm notwithstanding, it’s not clear how far along the Standard Power project is. “A project of this size has a significant amount of detail that’s confirmed and structured before a project begins construction and those discussions are ongoing,” NuScale CEO John Hopkins told analysts on the company’s most recent earnings call. Standard Power did not return a request for comment asking for more details on the financing or construction timeline for its project. When asked for an update from NuScale, a spokesperson referred me to the earnings call.
Meanwhile, in Surry County, Virginia, work is advancing on a project adjacent to the existing Surry nuclear plant. The project would combine data centers, small modular reactors, and hydrogen fuel production; the data centers would come first, with SMRs following once costs come down, according to Michael Hewitt, the co-founder and chief executive officer of IP3, the project’s developer.
For Hewitt, the model for SMR deployment is to build them in factories and scale them directly for end users. “That’s the future of energy: If I want a gigawatt of data center, I build SMRs for the data center on day one,” he told me.
Which company will get there first? “If I had to guess right now, in terms of what will be factory-built first and available to consumers like us, it will more than likely be a light water reactor design — GE, NuScale, or perhaps Rolls-Royce,” Hewitt said. GE’s SMR design, the BWRX-300, is in the pre-application process with the NRC, and was picked by Ontario Power Generation for a nuclear development on its existing Darlington site. The Rolls-Royce SMR has been advancing through the British regulatory and procurement process, while the company currently designs light-water reactors for the Royal Navy.
“The first guy to get the factory built is the winner,” Hewitt said. But none will likely be ready for the Virginia project, at least not within the next eight to 10 years, though, he added. Nevertheless, urgent interest persists.
On Tuesday, Google, Microsoft, and the steel company Nucor announced that they were forming a group that would commit to purchasing clean firm technologies and included in its laundry list of potential power sources advanced nuclear. Another advanced nuclear developer, TerraPower, which is backed by Microsoft’s founder Bill Gates, announced Tuesday that it was applying for a construction permit for a plant in Wyoming and plans to start building non-nuclear portions of it in June. The company expects the full plant to come online in 2030.
There are dozens of other SMR designs at various stage of realization, but the absolute fastest a new design could get online, according to Adam Stein of the Breakthrough Institute, is around four years. “If a developer has not already submitted an application to the NRC to build a power plant — which none of them have for a specific site — then they mostly likely would not be able to operate a power plant before 2028,” Stein told me. “That is the soonest it could happen.”
That said, “If there’s more urgency from the market, a clearer and larger demand signal, then developers will move faster than they are right now,” Stein added.
What’s far more likely, according to Yurman, is that tech companies will sign power purchase agreements for existing nuclear power plants, as Amazon has with Talen Energy. “That’s immediate access to reliable power,” Yurman said.
And even if SMRs are actually built, they may not end up adjacent to data centers, but instead on the sites of existing nuclear and even coal plants (this is the plan for the TerraPower site) which have preexisting grid connections. “If I’m putting together this kind of deal,” Yurman told me, “I’m looking at an old coal power plant I can demolish and keep the grid connection.”
While American tech companies are eager to buy up new power, the real opportunity, should it ever come, may be overseas, where smaller countries without indigenous energy supplies could be especially interested in nuclear power.
“What we need to do is get to full rate production and start stamping out SMRs with low risk,” Hewitt said. “If we do that, we can take these things everywhere.”
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
Two new reports out this week create a seemingly contradictory portrait of the country’s energy transition progress.
Two clean energy reports out this week offer seemingly contradictory snapshots of domestic solar and battery manufacturing. One, released Wednesday by the Rhodium Group’s Clean Investment Monitor, shows a distinct decline in investment going into U.S. factories to make more of these technologies. The other, released today by the trade group American Clean Power Association, shows staggering recent growth in production capacity.
So which is it? Is U.S. clean energy manufacturing booming or busting?
Maybe both.
The U.S. is suddenly producing more solar and batteries than ever before — enough to meet current domestic demand — so it makes sense that investment in new factories is starting to slow. At the same time, there’s a lot of room for growth in producing the upstream components that go into these technologies, but the U.S. is no longer as attractive a place to set up shop as it was over the past four years.
The U.S. saw 30 new utility-scale solar factories and 30 new battery factories come online last year alone, according to ACP. The country now has the capacity to meet average domestic demand for storage systems through 2030, and can produce enough solar panels to satisfy demand two times over.
In both industries, nearly all of that capacity has been added since 2022, when the Inflation Reduction Act created new subsidies for domestic manufacturing. The advanced manufacturing production tax credit incentivized not just solar and battery factories, but also all the production of components that go into these technologies, including solar and battery cells, polysilicon, wafers, and anodes. On top of these direct subsidies, the IRA generated demand for U.S.-made products by granting bonus tax credits for utility-scale solar and battery projects built with domestically produced parts.
“The policy definitely laid the right foundation for a lot of this investment to take place,” John Hensley, ACP’s senior vice president of markets and policy analysis, told me.
Trump’s One Big Beautiful Bill Act has changed the environment, however. The utility-scale wind and solar tax credits were supposed to apply through at least 2033, but now projects have to start construction by July 4, 2026 — just over a month from now — in order to claim them. Any of those projects that got started this year will also have to adhere to complex new sourcing rules prohibiting Chinese-made materials.
Now, dollars flowing into new U.S. solar factories appears to be on the decline. Investment fell 22% between the fourth quarter of last year and the first of 2026. Battery manufacturing investment dropped by 16%.
The reason investment is declining is not entirely because of OBBBA — it’s partly a function of the fact that a lot of the projects announced immediately after the IRA passed are entering operations, Hannah Hess, director of climate and energy at the Rhodium Group, told me.
Rhodium’s Clean Investment Monitor tracks two metrics, announcements and investment. Announcements are when a company says it’s building a new factory or expanding an existing one, usually with some kind of projected cost. Investments are an estimate of the actual dollars spent during a given quarter on facility construction, calculated based on the total project budget and the expected amount of time it will take to complete after breaking ground.
According to Rhodium’s data, the peak period for new solar manufacturing project announcements was the second half of 2022 through the first quarter of 2025. During that time, announcements averaged more than $2 billion per quarter. New solar factories announced this past quarter, by contrast, fell to about $350 million.
Since it can take a while to get steel in the ground, the peak period for investment was slightly later, with $13.5 billion invested between the second quarter of 2023 and the third quarter of 2025.
“What we were seeing in that post-IRA period was huge, almost unconstrained growth in that sector, and that’s not happening anymore,” Hess said.
Most of this growth occurred all the way downstream, at the final product assembly level — i.e. factories making solar and battery modules that still had to import many of the components that went into them. This was the “lowest hanging fruit” to bring to the U.S., Hensley, of ACP, told me, as the final assembly is the least technologically challenging part of the supply chain.
“These supply chains have momentum as they get going,” he said, “so as you establish those far downstream component manufacturing, you start to recruit all of the upstream manufacturing.” In other words, a solar cell manufacturer is far more likely to build in the U.S. if there’s a robust local market of module factories to buy the cells.
There’s evidence that’s still happening in spite of changes to the tax credit structure. The ACP report says that three solar cell factories came online between 2024 and today — one per year. If all of the additional factories that have been announced are built by 2030, the U.S. will have nearly enough capacity to meet all of its own demand for solar with domestic cells. Battery cell capacity is growing even faster, with three factories as of the end of 2025 and seven more expected to be complete by the end of this year, which will produce more than enough units to meet average annual demand.
It’s the next step up on the supply chain that spells trouble. For solar, that’s ingots and wafers, followed by polysilicon. Today, the only producer of ingots and wafers in the U.S. is a company called Corning. It produces enough to meet about 25% of current domestic solar cell production, but cell production will more than quadruple by the end of this year compared to last year, according to ACP. Similarly, we produce enough polysilicon to meet Corning’s current needs, but not enough to meet anticipated cell demand. The announced projects in the pipeline will not add much on either front.
For batteries, it’s the anodes and cathodes. There’s currently one factory in California producing cathodes and at least one more under construction, but as there is nothing else in the pipeline, the ACP report expects cell manufacturers to rely on imported cathodes for the foreseeable future. Anodes are the one bright spot — there’s one factory producing what’s known as active anode material factory in the U.S., and four more anticipated by the end of this year. Together, they have the potential to meet demand by 2028, according to ACP.
The question now is whether that snowball effect kicked off by the IRA will continue. “A lot has changed about the outlook for future demand after the One Big Beautiful Bill Act passed,” Hess said. “We have seen some more project cancellations and pauses in construction recently.”
Most recently, a company called Maxeon Solar Technologies canceled a $1 billion cell and module factory in New Mexico. The company had been “fighting for its life” since 2024, according to Canary Media. It’s also majority owned by a Chinese state-owned company. The
OBBBA was likely the nail in the coffin, as it penalizes solar developers who source panels from companies with Chinese ownership.
OBBBA also shortened the timeline for the wind and solar tax credits, while the Trump administration’s hostility to wind and solar permitting has made it more difficult for projects to get built before the credits expire. Hensley said the Trump administration’s hostility toward clean energy has added a lot of risk into the system, complicating final investment decisions for manufacturers.
On the flip side, tariffs have the potential to help some domestic producers. Duties on imports from countries such as Cambodia, India, and Vietnam, all major manufacturers of solar panels, “have made their exports to the U.S. almost prohibitive,” Lara Hayim, the head of solar research at BloombergNEF, told me in an email. “This sort of policy framework could continue to provide some protection for domestic manufacturers,” she said, but there are still plenty of countries with low enough tariffs that they will continue to serve the U.S. and compete with domestic manufacturers.
Hensley said that the Trump administration’s tariffs were a double edged sword. They can help domestic manufacturers, but not if they make all of the inputs into the product more expensive.
“That’s a problem with these blanket type of tariffs that aren’t really fine-tuned to target the behavior that you’d like to see,” he told me. “I think we’re seeing a lot of that push and pull and tension in the system at the moment.”
Between Trump’s tariffs and the OBBBA, there’s no doubt that the manufacturing boom sparked by the IRA is slowing. But Hensley is optimistic that the progress will continue. “We haven’t attracted all of the supply chain yet. It’s still a work in progress, but so far the signs are quite good.”
This week’s conversation is with Duncan Campbell of DER Task Force and it’s about a big question: What makes a socially responsible data center? Campbell’s expansive background and recent focus on this issue made me take note when he recently asked that question on X. Instead of popping up in his replies, I asked him to join me here in The Fight. So shall we get started?
Oh, as always, the following conversation was lightly edited for clarity.
Alright let’s start with the big question: What is a socially responsible data center?
So first, there’s water, which I think is pretty solvable.
Part of me thinks water is not even the right thing to be focusing on necessarily, and it’s surprising that it became at least for a while the center of the controversy around data centers.
I think there’s energy, which is mostly a don’t-raise-people’s-bills kind of thing. Or in extreme cases, actually reducing people’s access to energy.”
I think air pollution is another key. This is one of the biggest own-goals our [climate] space is making, because people are installing behind-the-meter power and we can talk about why they’re doing that, the shifting reasons, but the real shame in it is you really shouldn’t have to run those 24/7. If you’re building your own power plant, it should enable you to get a grid connection, because you’re bringing your own capacity and they can provide you firm service, and you should only have to run that gas plant 1% of the year, so air pollution is a non-issue. If only the grid and its institutions could get their act together, this is a no-brainer. But instead people run them 24/7.
There’s noise, which has been very misunderstood and bungled on a handful of well-known projects. That’s just a do-good engineering and site layout type of problem.
And then there’s other. Beyond the very concrete impacts of a data center, what else can it do for the community it's siting itself in? That’s going to be specific for every community.
There’s going to be a perspective that data centers are takers. They get tax incentives. They’re this big new thing. If data centers were to bring something compelling when [they’re] siting in communities, and it is specific to whatever they’re dealing with, maybe they’d be considered socially responsible.
I don’t think I have the master answer here. Everyone’s trying to figure it out.”
What do you hear from other folks in decarb and climate spaces when you ask this question? Do you hear people come up with solutions, or do they knock down the entire premise of the question — that there isn’t such a thing as a socially responsible data center?
You get both. You definitely get both. It depends on who you're talking to.
I can understand both sides of the equation here. There’s definitely solutions, first of all. I do think there’s a group of people whether it is in the energy world or the data center world or tech who would have this incredulous disbelief that anyone could not want what they’re doing. And that then, after being poked and prodded enough, transforms into a very elitist, almost pejorative explanation of everybody’s just NIMBYs.
I think that’s really unproductive. It kind of just throws gas on the fire.
But there’s a lot of people working on solutions, too. The non-firm grid service thing is just a huge opportunity. To be able to connect these sites to the grid in such a manner they either get curtailed some small amount of hours per year or they show up with accredited capacity, absolving them from curtailing. I mean, we can do that. It’s very doable.
The second question becomes, what are the forms of accredited capacity that can be deployed quickly? I think that’s where there’s a lot of cool stuff around VPPs and such. Sure, build a gas power plant, run it once or twice a year. If anything that’s good for a community — back-up power at grid scale.
There’s also other solutions. A really cool effort right now, former Tesla people building a purely solar and battery DC microgrid in New Mexico.
And there’s also a lot of inertia. The folks making decisions about data centers have been doing stuff a certain way for 20 years and it’s hard to change. The inertia within the culture combined with the enormous pressure to deploy just makes it less dynamic than one would hope.
On my end, I’ve been grappling with the issue of tax revenue. We’re seeing a declining amount of money for social services, things that can really help people for both personal and academic reasons. There's quite a bit a lot of people could say on that topic. At the same time, this is another form of industrial development. People are upset at the amount of resources going to this specific thing.
So when it comes to the data center boom in general, where do you stand on social cost-versus-benefit analysis?
That’s a good question. I’m not an expert. I’m mostly just someone who designs energy projects. But I can say where I’m at personally.
Yeah, but isn’t everyone in the energy space talking about data centers? Shouldn’t we all be thinking about this?
Of course. I’m not in a place to proclaim what is right but I’ll tell you where I’m at right now.
With any large-scale industrial build out it is tough relative to other technological changes that were simpler at the infrastructure layer. Like, the smartphone. Massive technological change but pretty straightforward in a lot of ways. But industrial buildout stresses real physical resources, so people have much more of an opinion of whether it’s worth it or not.
I’m pretty optimistic about AI generally. It’s very hand-wave-y. It’s hard to cite data or anything, because we’re talking about something that hasn’t happened yet, but I’m very optimistic about increasing the amount of intelligence we have access to per person on Earth.
A similar thing I think about is when everyone stopped getting lead poisoning all the time, we all jumped five IQ points and killed each other less. Intelligence is good. A lot of our story as a species is about increasing intelligence and learnings-per-person so we can do more. The idea that we would be able to synthesize it, operate it as a machine outside of our own bodies. It feels pretty inevitable.
There’s questions about what that [AI] will do to the economy and jobs, which is what people are really concerned about and is the case with any major technological change.
Are data centers being deployed at a rate and in a way that is responsible? Like, does it need to be this fast? That’s a question people ask and that’s in a way the question being posed by the moratoriums. They’re not saying let’s ban this forever. They’re saying, let’s take a breather. And I do understand that.
There’s a lot of good solutions that could just be pursued and it’s hard for me to separate my feelings about the current path data centers are taking from what I think is objectively right. We could just be doing way better.
On the energy front, what do you make of the way our energy mix — carbon versus renewables, our resilience — is headed? And where do you think we’re heading in five years?
For the energy and climate world, this is the real question. Data centers are a complicated thing but at the end of the day, for us, they’re a source of electricity demand.
From an electricity perspective, there’s been no growth for 20 years. So the theory of addressing climate change was, as the old stuff breaks we’ll replace it with new clean stuff. That was what we were doing, while saying, a lot of the old stuff we’ll keep around. We’ll layer on the new clean stuff.
It was always the case though that we could enter a new phase of electricity growth. Actually, five years ago, when the phrase “electrify everything” was coined, it explicitly became our goal! We were going to massively and rapidly grow the electricity system in order to switch industry, heating, and transport off of fossil fuels. That’s the right prescription, the right way to do it.
My understanding of it is that while this feels really big, because we haven’t grown in so long, compared to the challenge we were all talking about doing is not big at all. It increases the challenge by 15% or 20%. That’s meaningful. But it just seems like we should be able to do this.
From a climate perspective, as someone who’s been trying to do everything I can on it for a while now, I can’t help but feel a little dismayed that today the growth we’re experiencing is some tiny, tiny percentage of what we actually set out to do. And it’s causing chaos. We’re institutionally falling apart from a single percent of what our goals should be.
This is the time for the electrification case. We can all demonstrate this is possible over the next few years. I think confidence in the electricity system as our energy path can remain high. Or this utterly fails, where it’s really hard to imagine governments and businesses making any sincere attempt at a high electrification pathway.
Plus the week’s biggest development fights.
1. LaPorte County, Indiana — If you’re wondering where data centers are still being embraced in the U.S., look no further than the northwest Indiana city of LaPorte.
2. Cumberland County, New Jersey — A broader splashback against AI infrastructure is building in South Jersey.
3. Washington County, Oregon — Hillsboro, a data center hub in Oregon, is turning to a moratorium.
4. Champaign County, Ohio — We’re still watching the slow downfall of solar in Ohio and there’s no sign of it getting any better.
5. Essex County, New York — Man oh man, what’s going on with battery storage in rural pockets of the Empire State?