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:
Rob talks New Jersey past, present, and future with Employ America’s Skanda Amarnath.
Though high costs have become central to the upcoming election, they’re mostly out of the state’s control.
In a press conference about the newly recast program’s first loan guarantee, Energy Secretary Chris Wright teased his project finance philosophy.
The administration seems to be pursuing a “some of the above” strategy with little to no internal logic.
The country’s underwhelming new climate pledge is more than just bad news for the world — it reveals a serious governing mistake.
You might even call the Energy Secretary ... Chris Wrong.
The company, Nuclearn, aims to speed development and licensing processes with the help of a specially trained large language model.
You’d be hard-pressed to dream up a buzzier clean tech concept than an AI platform custom-designed for the nuclear industry. Yet Phoenix-based startup Nuclearn has been betting on the role of artificial intelligence in the booming nuclear sector since 2021 — predating the wide launch of ChatGPT and the Trump administration’s recent embrace of nuclear energy.
Now the funds are rolling in. The company announced today that it raised a $10.5 million Series A round led by the climate tech venture fund Blue Bear Capital. With this cash, Nuclearn plans to expand its repertoire of AI offerings, which spans everything from identifying and documenting faults in a reactor to project scheduling, engineering evaluations, and licensing and permitting for new or modified reactors.
To expedite these processes, the company has developed its own, nuclear-specific language model, built atop existing open source models and trained on public data from the Nuclear Regulatory Commission and other government agencies, Nuclearn’s cofounder and CFO, Jerrold Vincent, told me. This allows the model to pick up on “a lot of nuclear specifics, whether it’s the acronyms, vernacular, specific processes, even just sometimes the way [the nuclear industry] thinks about certain types of issues and the level of scrutiny they put on one thing versus another,” he explained.
By way of example, Vincent told me that one of the startup’s current customers is working on a licensing application and wanted to conduct some background research to identify potential gaps or areas where the NRC might raise additional questions. Every other time the company has pre-checked an application like this, Vincent said, it was a 400-hour process. Nuclearn helped reduce that timeline to less than a day.
It’s a deeply resonant win for Vincent and his cofounder, Bradley Fox, who are all too familiar with the inefficiencies of the industry themselves. Prior to founding Nuclearn, both worked in data science at the Palo Verde Nuclear Generating Station in Arizona, where employees spent thousands of hours every year on “a lot of documentation, a lot of paperwork, a lot of manual work,” Vincent told me.
Natural language processing had some very obvious applications for the nuclear industry. “Everything in nuclear is text. Everything’s written down,” Vincent said. So when some of the seminal research on novel deep learning models started coming out in 2017 and 2018, Vincent and Fox took note, exploring ways they could apply this to their own work. “Those were trends we jumped on very, very early, not because they were particularly fashionable at the time or because there was a lot of hype around it, but because that was the type of techniques we needed to be able to solve these problems,” Vincent told me. “That’s why we got into the language model space half a decade before ChatGPT.”
For the majority of jobs, such as working on permitting or license renewals, Nuclearn uses a software layer on top of its language model to coordinate various AI agents working on tasks linked to different data sets, such as analyzing design functions, safety protocols, or systems degradation over time. The software then integrates these various outputs to generate reports or summary analyses. On the operational side, the company has its own benchmarks to evaluate how its AI tools are performing on nuclear-specific tasks.
There is, of course, a certain poetic irony to the fact that AI is being used to license and manage operations for the very reactors that are now in such high demand for their ability to consistently and cleanly power AI data centers. The better AI gets, the more we need nuclear; the more we need nuclear, the more useful AI-powered tools like Nuclearn become.
To date, the company has integrated its AI platform into the operations of more than 65 reactors both domestically and abroad, which Vincent told me represents a mix of standard commercial reactors and small modular reactors. As the market heats up, demand may well follow. With the Trump administration pushing to accelerate nuclear development, electricity demand rising, and tech giants prioritizing clean, firm power, it’s boom times for companies looking to build everything from conventional nuclear plants to small modular reactors, microreactors, and the long-elusive fusion reactor, each and every one of which will have to be licensed and permitted.
All this activity also means that the nuclear workforce is under strain, especially given that 25% set to retire in the coming decade. “We’ve had knowledge and workforce challenges for several years now, and now it’s getting exacerbated quite substantially from all the macro trends going on,” Vincent told me. Given this situation, he doesn’t anticipate that the adoption of AI tools will necessarily lead to layoffs. These days, he said, the industry is just wondering “how do we do the things we need to do to operate a nuclear power plant safely and efficiently with less people?”
With this new capital, the startup plans to scale its operations to encompass even more aspects of nuclear reactor management. One future use case Vincent anticipates is helping to automate the sourcing of unique, industry-specific parts. There are plants operating today, he told me, that rely on equipment from vendors that may be long out of business. Figuring out how and where to source equivalent components is the type of niche challenge he’s excited to take on.
“It just tends to be very manual, labor intensive, and very documentation heavy,” Vincent told me of the industry as a whole. Luckily, “those are all things that AI is very good at solving these days.”
Editor’s note: This story has been updated to note some poetic irony.
Jesse gives Rob a lesson in marginal generation, inframarginal rent, and electricity supply curves.
Most electricity used in America today is sold on a wholesale power market. These markets are one of the most important institutions structuring the modern U.S. energy economy, but they’re also not very well understood, even in climate nerd circles. And after all: How would you even run a market for something that’s used at the second it’s created — and moves at the speed of light?
On this week’s episode of Shift Key Summer School, Rob and Jesse talk about how electricity finds a price and how modern power markets work. Why run a power market in the first place? Who makes the most money in power markets? How do you encourage new power plants to get built? And what do power markets mean for renewables?
Shift Key is hosted by Jesse Jenkins, a professor of energy systems engineering at Princeton University, and Robinson Meyer, Heatmap’s executive editor.
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:
Jesse Jenkins: If I’m just a utility operating on my own, I want to basically run my fleet on what we call economic dispatch, which is rank ordering them from cheapest to most expensive on a fuel or variable cost basis, and trying to maximize my use of the less expensive generators and only turn on the more expensive generators when I need them.
That introduces this idea of a marginal generator, where the marginal generator is the last one I turned on that has some slack to move up or down as demand changes. And what that means is that if I have one more megawatt-hour of demand in that hour — or over a five-minute period, or whatever — or 1 megawatt-hour less, then I’m going to crank that one generator up or down. And so the marginal cost of that megawatt-hour of demand is the variable cost of that marginal generator. So if it’s a gas plant that can turn up or down, say it’s $40 a megawatt-hour to pay for its fuel, the cost on the margin of me turning on my lights and consuming a little bit more is that that one power plant is going to ramp its power up a little bit, or down if I turn something off.
And so the way we identify what the marginal value of supplying a little bit more electricity or consuming a little bit more electricity is the variable cost of that last generator, not the average cost of all the generators that are operating, because that’s the one that would change if I were to increase or decrease my output.
Does that make any sense?
Robinson Meyer: It does. In other words, the marginal cost for the whole system is a property of the power plant on the margin, which I realize is tautological. But basically, the marginal cost for increasing output for the entire system by 1 megawatt-hour is actually a property of the one plant that you would turn on to produce that megawatt-hour.
Jesse Jenkins: That’s right, exactly. And that can change over the course of the day. So if demand’s really high, that might be my gas plant that’s on the margin. But if demand is low, or in the middle of the day, that gas plant might be off, and the marginal generator during those periods might be the coal plant or even the nuclear plant at the bottom of the supply curve.
Mentioned:
Jesse’s slides on electricity pricing in the short run
Jesse’s lecture slides on electricity pricing in the long run
Shift Key Summer School episodes 1, 2, and 3
This episode of Shift Key is sponsored by …
Accelerate your clean energy career with Yale’s online certificate programs. Gain real-world skills, build strong networks, and keep working while you learn. Explore the year-long Financing and Deploying Clean Energy program or the 5-month Clean and Equitable Energy Development program. Learn more here.
Join clean energy leaders at RE+ 25, September 8–11 in Las Vegas. Explore opportunities to meet rising energy demand with the latest in solar, storage, EVs, and more at North America’s largest energy event. Save 20% with code HEATMAP20 at re-plus.com.
Music for Shift Key is by Adam Kromelow.
