Nuclear's AI Opportunity
Nuclear's Big Tech Moment, the History of Nuclear, the Challenges Ahead, and Why I'm Excited
We’re seeing an interesting convergence of trends opening the door for a nuclear revival: AI’s energy-intensive nature, demand for clean baseload power to meet those needs, and big tech’s willingness to finance it.
Industry veterans, utilities, and the people building nuclear companies know the challenges ahead; nuclear power is a difficult problem to solve. However, if nuclear power can be built affordably, the payoffs will be clear.
This has not been an under-covered story over the last month; we’re seeing articles every day on the potential for nuclear to address our electricity needs. Buoyed by the hyperscaler’s recent investments in nuclear, see recent announcements here:
Amazon invested in a $500M round into nuclear startup X-energy while supporting the construction of 320MW worth of nuclear generation.
Google signed a deal to support the development of over 500MW of power from nuclear startup Kairos.
Amazon bought a nuclear-powered data center for $650M, intending to expand energy consumption from 300MW to 960MW over the coming years.
Microsoft signed an agreement to restart the reactor at Three Mile Island. It is rumored that they have signed a power purchase agreement (PPA) for approximately $800M/year for 20 years.
While that last point may seem expensive, the hyperscalers are more concerned about electricity availability than affordability right now.
On the surface, nuclear provides a clear narrative, ”Consistent, clean power. It is the most efficient form of energy creation known to mankind. The incidents decades ago soured public opinion, and we haven’t recovered since.”
One layer deeper, the story gets more interesting. Nuclear faced real challenges before those incidents. It’s one of, if not the only, forms of energy that’s gotten more expensive over time. When I talked to nuclear experts, the problem with the adoption of nuclear is not a technology problem. It’s an engineering problem. It’s a construction problem. It’s an economic problem.
Companies across the nuclear spectrum are trying to solve this problem.
This will be a deep dive into the nuclear industry, its technology, its history, the market today, and what the future might hold.
For readers who are new to the space, I recommend reading recent AI data center and electric grid pieces before reading this. They will provide context on the energy sector’s tailwinds and bottlenecks.
1. A Brief Overview of Nuclear Technology
Nuclear reactions come in two forms: fusion and fission. Fusion, i.e. the sun, joins two atoms together and forms energy. It’s essentially the final form of energy for humanity, providing abundant clean energy if commercially viable. It’s unclear how far out we are from developing fusion technology at scale, but that isn’t stopping companies from trying.
This article will mostly focus on fission, the splitting of atoms to release energy. For every fission reaction, the goal is to achieve a chain reaction splitting heavy atoms like Uranium-235. Each reaction releases more neutrons, splitting more atoms, and so on until fuel is depleted.
Nuclear Reactors Today
Most nuclear reactors have the same primary components components:
A reactor with fuel (uranium, for example), control elements (rods/blades that control the rate of chain reaction), and moderators (elements to slow neutrons, maintaining the chain reaction).
A coolant cycling through the reactor to collect heat.
A power generation mechanism to turn that heat into electricity.
For example, the most common reactor in the US is a pressurized water reactor (PWR):
The reactor generates heat while water is pumped through the reactor. That water is heated, a second water loop collects that heat, turning it into steam, and that steam turns a turbine, creating electricity.
In comparison, a boiling water reactor only has one loop:
Water is heated by the reactor, turned into steam, and converted to heat. The downside of the BWR is that radioactive water is not contained within a separate loop like the PWR. The single loop of the BWR is simpler but can complicate the waste management process.
For context, the average large nuclear reactor will generate between 1000-1400MW. As will be discussed later, building one of these reactors is costly and timely. Ergo, the rise of the small modular reactor or SMR.
Small Modular Reactors & Microreactors
Note that the following phrases are somewhat arbitrary. I refer to specific reactors based on the megawatts of electric energy (MWe) they create. The difference between a 300MWe and a 10MWe reactor is night and day, but they’re often bucketed under the “SMR” term.
We can see the various sizes of small and microreactors mapped out here:
Generally, any reactor smaller than 50MW is considered a microreactor. Although I’ve heard some consider a microreactor to be less than 20MW. And there are portable microreactors, which tend to be ~1-5MW!
Packy McCormick made a great chart breaking down the various approaches in the small reactor landscape, which vary based on size, fuel, and type of coolant used (water, gas, metal).
Before diving deeper, we should take a step back and look at the industry's history. The problems the industry faced when it stalled in the 70s are many of the same ones companies are trying to solve today.
2. A History of the Nuclear Industry
I want to start with a quote from Nick Touran via Packy McCormick’s deep dive on nuclear startup Radiant. Much of this section comes from this deep dive and associated Age of Miracles podcasts, I highly recommend the read/listen. Another good resource is Fabricated Knowledge’s recent post on nuclear.
The Atomic Age
There was a time back in the 40s, 50s, 60s, where there were a hundred thousand people around the world, the smartest people in the world, all focused on nuclear reactor technology. It was like The Thing. And so there's so much interesting information and history and things that people did back then. It just blows my mind. Every time I go, look, I find something new. We had lots of small reactors, dozens of small reactors, some of which pretty much check all the boxes of the things that we're excited about now. We actually built them.
Sounds a lot like AI today! Post WW2, the US was excited to harness the power of the atom for humanity’s benefit. Ergo, the famous “Atoms for Peace” Speech.
The Atomic Energy Commission (AEC) was responsible for supporting bringing this to life. In the 50s, we saw most of the technology emerge used today in nuclear reactors.
In 1953, the first PWR reached criticality, and Westinghouse then created the first commercial reactor in 1957.
In the 60s, we started to build. Nuclear submarines and power plants, for example. We even investigated nuclear-powered airplanes, which could fly in theory for years (although these plans would later be canceled).
However, as we started to build, there was an early crack in these plans: the assumption of decreasing cost curves on nuclear power plants.
Companies like Westinghouse sold early reactors, likely at a loss, with fixed-price contracts. They wanted to be early and gain the benefits of customer trust and the learnings from being the first to build reactors. It worked!
Westinghouse built trust with customers, and customers then started buying reactors on variable cost contracts, where customers would pay based on the construction costs of the reactor. Everyone assumed the price of nuclear builds would go down. Well, the cost curves did go down, and then they didn’t.
It’s debated why this happened, but a few factors are rising costs from regulation and labor costs, which make up most of a reactor’s costs:
As the industry continued to see increasing costs, nuclear plants saw increased hesitancy from buyers.
Nuclear Stalls
In addition to the economic concerns, a few things happened in the early 70s that complicated the nuclear landscape. First, there were various groups opposed to nuclear energy, both environmental groups and lobbying organizations. Second, policies were put in place, such as “ALARA” or “As Low as Reasonably Achievable,” which refer to radiation minimization of nuclear reactors. While this principle sounds logical, the ambiguity made it difficult for companies to comply.
Additionally, in 1974, the AEC was split into two organizations: The Energy Research and Development Administration (ERDA) and The Nuclear Regulatory Commission (NRC). This delegated R&D to the first org and regulation to the second org. This put the NRC in a difficult situation, with its primary goal being the regulation of civilian nuclear power.
Since that decision, no advanced nuclear reactor design has been approved, and no construction of any new reactor design has been approved.
All of these things happened before Three Mile Island, Chernobyl, and the cultural fears around nuclear at the time. The initial downfall of nuclear was economic, and it remains the challenge today.
You might ask, well how many nuclear plants have been built since those events? Well…not many.
https://www.eia.gov/energyexplained/nuclear/us-nuclear-industry.php
3. Nuclear Markets Today
Despite these challenges, nuclear is still a key part of the energy market. About 20% of US energy comes from nuclear. Most of these nuclear plants are operated by utilities like Constellation and Duke Energy. The majority of these plants are 30-50 years old.
Nuclear is the most reliable form of baseload power we have. After plants are built, the economics of nuclear power are quite good.
The problem comes with the massive upfront costs of building nuclear. For example, the Vogtle plant in Georgia, the most recent reactor built in the US, cost $35B ($21B over budget) and 7 years over schedule. The high financing costs and the construction risks make this quite an unattractive risk/reward for utilities.
The liftoff report argues that economics will get better as more reactors are built, and this is likely true. Vogtle had some unique missteps that led to cost overruns as well.
There’s another path forward: what if the utilities aren’t the financier?
Enter big tech with their deep pockets, political influence, and, most importantly, a deep crisis around where to get clean, reliable power.
But wait, there’s more! What if we can reduce the need for these massive upfront costs and gradually increase capacity as necessary? Enter SMRs and microreactors.
SMRs
SMRs are meant to gain economies of scale from manufacturing. The smaller size and simpler designs, in theory, should lead to an easier construction process.
First, a disclaimer that, to my understanding, no SMRs have been built in Western countries. There are billions of dollars worth of agreements but no revenue yet. In light of that fact, this section may be premature. What matters right now is not contracts but execution.
The companies that can meet deadlines, both from an engineering and regulatory perspective, fastest will be most successful in this market. I’m of the opinion that the approaches that put companies in the best position to do that are the most likely to be successful.
The DoE Pathways to Commerical Liftoff report describes SMRs well:
For Small Modular Reactors (SMRs), “small” is generally considered to be under approximately 350 MW, while “modular” typically refers to standardized factory production.
Because civil works construction drives nuclear capital costs, the value proposition for SMRs centers on maximizing design standardization and factory production. To realize this potential, SMRs must move a substantial portion—more than 50%—of overall spending into the factory setting. Without this shift, an SMR risks becoming merely a civil works construction project, lacking the benefits of economies of scale.
SMR construction will require dedicated modular assembly capabilities, with requirements differing by design. Unique capacities will be needed for each design, making design down-selection critical for standardization and reducing total industry costs.
So, SMRs might be more expensive in $/kW, but the value of faster builds and more digestible upfront costs make them attractive:
As called out earlier, Amazon, Google, and Oracle have publicly said they’re working with SMRs. There are only 2 SMRs currently operational worldwide, so this is a bet on the future.
So, in theory, smaller modular reactors make construction costs more digestible. What if we can make them even smaller?
Enter microreactors.
Microreactors
A particularly interesting approach is building portable microreactors anywhere from 300kW to 10MW of energy.
We can use Radiant Nuclear as an example here. Radiant was founded by Doug Bernauer, a SpaceX veteran, with the goal of bringing power to Mars. To do that, he needed to develop nuclear reactors that worked on Earth and could make money.
They eventually settled on a 1MWe microreactor that could fit in a shipping container:
The technology is similar to the dual loop reactors described above with PWRs. There’s the reactor with a graphite core and HALEU fuel pellets. Helium, the coolant, is pumped through the reactor to collect the heat. Via a heat exchanger, that heat is passed on to a supercritical CO2 loop to generate electricity:
This design is meant to optimize for safety via the fuel (meltdown-proof pellets), coolant (helium is not radioactive), and passive cooling (the reactor shuts down safely on its own).
On to the business model of microreactors.
Microreactors aren’t meant to solve grid-scale energy problems. Their value comes from providing a portable, reliable, clean source of energy. Today, the best comparison is diesel generators. The problem with diesel generators is that they require constant fuel, which means constant fuel shipments. In remote areas, military bases, and disaster relief, this may not be viable.
This isn’t to say data centers won’t use microreactors; keep in mind they use diesel generators themselves. They just won’t solve the GW-scale power needs of data centers.
If an approach like Radiant’s works, they can mass-produce microreactors that can provide a clean, reliable, portable, scalable energy source.
This brings us to the current state of nuclear markets today:
A new wave of buyers (and financiers) for nuclear power.
A new wave of companies aiming to make the cost of nuclear more reasonable, mostly through better engineering and better manufacturing.
4. Some Thoughts on the Space
I think it’s easy to see the array of SMR and microreactor companies and think about the competition between them.
I think the right approach is to view this as an industry-wide movement with the goal of replacing fossil fuels OR addressing net new energy demands. For example, if a microreactor company gets its design approved, every other company then has a roadmap to follow on design approval. That first company will start getting customers and building trust in nuclear power.
That tide rises all boats. The supply chain starts to develop, customer trust rises, and some economies of scale may start to be gained. The industry, from 1MW reactors to 300MW reactors, benefits from this.
My final comment is to acknowledge my clear optimism on the space. The problems the industry faces are daunting; they’ll be hard to solve and will take years to do so successfully. The recent hype around nuclear shouldn’t cloud our judgment, and the people building companies in this space are well aware of those challenges.
However, the best thing about innovation is acknowledging those challenges and moving forward anyway. This movement towards nuclear embodies that. These companies, led by ambitious people, are taking on seemingly impossible challenges with the hope of moving us forward. That’s something we all can be proud of.
As always, thanks for reading!
Disclaimer: The information contained in this article is not investment advice and should not be used as such. Investors should do their own due diligence before investing in any securities discussed in this article. While I strive for accuracy, I can’t guarantee the accuracy or reliability of this information. This article is based on my opinions and should be considered as such, not a point of fact. Views expressed in posts and other content linked on this website or posted to social media and other platforms are my own and are not the views of Felicis Ventures Management Company, LLC.