A nuclear startup will probably not be the next SpaceX

This article represents my personal views of opinions and not those of my employer.

For the last decade in nuclear, it’s been in vogue to make comparisons to SpaceX. It’s understandable why: In 2008, launch services were a stilted industry, funded almost entirely by governments and dominated by massive companies operating under cost-plus contracts. In short order, they were disrupted by an upstart, leading to radical price cuts and performance improvements. Who wouldn’t want to replicate that in the nuclear industry?

By all means, there are improvements to be made in the nuclear industry, and I’m all for challenging the status quo. But for all the SpaceX/Nuclear fanfiction out there, I haven’t seen a thoughtful explanation of some of the fundamental differences between the nuclear industry today and the space launch industry as it was in 2008. As someone who was closely involved with several nuclear startups, has a background in aerospace engineering, worked with NASA on nuclear propulsion, and now works at a company building and maintaining reactors worldwide, I was inspired to to use my free time helping people understand something.(Shoutout to Nick Touran for inspiration).

Unquestionably, radiation hazard (and public fear of radiation hazards) is a key difference between reactors and rockets; that said, I’m not going to address that below since if this topic is broached at all, that’s usually the only thing covered. From a “pure” technology development perspective, there are three other significant differences between developing a launch vehicle and a nuclear power plant:

The timeline to evaluate success for nuclear reactors is decades not minutes. Electricity is already a commodity market with robust competition, launch services was not. Nuclear deployment cannot vertically integrate at useful scale, while SpaceX could.

Timeline to innovate

The fourth flight of Falcon 1 launched at 23:15 UTC on September 28, 2008. About 9 minutes later its payload was in orbit and the mission could be legitimately claimed a success. Each implementation (especially the ones that never made it to the launch pad) could be fully evaluated over its entire lifecycle within a number of months. “Did it work,” and “how much did it cost” were answerable questions.

In contrast, nuclear reactors are investments with high upfront capital cost and (ideally) low operating costs. To result in a positive return on investment, they don’t just have to work, they have to work, and work reliably, for decades. The first commercial Pressurized Water Reactors were built with Alloy 600 steam generators. The alloy was used for a range of plants built starting in the late 1950’s through the early 1970’s. In 1971, a German nuclear plant first identified a phenomena known as stress corrosion cracking (SCC) which could lead to undetectable cracks in hard to access areas. Over the next two decades, a range of chemistry, alloy, and material control processes were developed to alleviate SCC, but many plants had to be retrofitted, costing billions of dollars and resulting in lengthy outages. San Onofre Nuclear Generating Station had a botched steam generator replacement which eventually contributed to the shutdown of the multi-billion dollar reactor after the billion dollar retrofit to fix SCC issues.

When each iteration of a nuclear plant can take years and hundreds of millions or billions of dollars, and each unit is expected to operate for decades, you simply can’t get to the same level of confidence with rapid testing. How much do you want to invest before you know what the long-term operating performance is? While non-nuclear testing has value, the non-nuclear parts of water reactors were figured out on coal plants by the 1950’s (seriously, they haven’t changed that much). It still took decades before nuclear costs truly came down to the level to compete with other electricity sources as maintenance and operating issues were resolved.

Said another way, given the long lifecycle of reactors, the information you get from a multiple-month reactor test is similar to building a rocket, putting it on the launch pad, taking it down after the photo op, and calling that a success.

Electricity markets don’t have juicy margins

A second thing I see overlooked is what competitive market forces exist within the nuclear industry, compared to those in launch services.

In the early 2000’s there were essentially four launch vehicle providers, (and really only one in the US after Boeing and Lockheed formed ULA). Similarly, there were an equally small number of customers, mostly governments. If you wanted to get a satellite to space, you had to buy a launch vehicle, and you had to buy it from those providers. Because so few commercial customers could afford the exorbitant launch costs, there wasn’t that much market pressure to innovate, and government cost-plus contracts provided little incentive to improve the underlying technology. This situation left massive margins if someone could even modestly reduce costs. As launch costs came down, newly profitable markets opened, significantly expanding the addressable market that SpaceX was the cheapest provider for.

Compare that to electricity: almost everyone buys electricity, and they do so from a technologically, geographically, and organizationally diverse group of producers. While AI and crypto-currency drive more aggressive load growth projections, few expect the electricity market to grow by an order of magnitude even in the most optimistic situations. Some customers might pay a premium for nuclear (northern Alaskan towns and military bases for the reliability or a tech company for low carbon). But for pretty much everyone else, electricity is like gasoline: as long as it works, you don’t care where it’s from, and you just want it to be cheap. So while nuclear innovators can try to disrupt the nuclear industry, they are competing within a mature electricity industry.

When SpaceX made even a little better launch vehicle, they could beat everyone on cost and rapidly gain market share while simultaneously expanding the market. Making a little better nuclear plant won’t win the electricity market; it will just put you in the same fierce competition as everyone else, where margins are already razor thin and a diverse marketplace of technologies has been cutting costs for decades.

Vertical Integration of infrastructure is different from equipment

A final point concerns the ability to vertically integrate nuclear power plants. Many well-funded startups aim to replicate SpaceX’s vertical integration strategy and reap the cost and performance benefits. SpaceX could control cost and performance throughout the supply chain by bringing it in house, wresting design and price control from suppliers and contractors. In addition to having the capital and willingness to invest, one of the logistical reasons SpaceX could do this at all was they could literally bring rocket components into the house (or at least an assembly building). Even the largest components of each Falcon 9 can be shipped by truck, allowing manufacturing, transport, and delivery to happen in controlled, centralized locations from a finite number of suppliers. The fully integrated rocket is launched from where it is built or can be transported, allowing final integration to occur in a similarly controlled, consistent location.

Building a nuclear plant is a much bigger endeavor by a few orders magnitude. Each reactor must be integrated and operated in the same region it is used. While reactor vessels and other equipment for even GW-scale reactors can be readily shipped by truck or rail, most of the cost of building a nuclear plant is infrastructure-scale construction. Even the 2 MW reactors built by the Army in the 1960’s had sizeable earthworks, concrete shielding, control buildings, switchyards, cooling systems etc. Yes, if you can really shrink reactors down to where they fit on a truck (and concurrently figure out how to dramatically reduce the cost of fuel, site prep, construction, licensing, disposal, security, inspection, and maintenance) then building a thousand 1 MW reactors might be cheaper than building one 1 GW one. In that case, you can vertically integrate in the SpaceX model. But at least in the 1960’s when we tried, that didn’t pan out.

For infrastructure-scale reactors, it is nearly impossible to vertically integrate the majority of your cost centers, and few utilities want to sign over their billion dollar project to an Architect-Engineer and EPC firms they don’t know, especially when that AE and EPC also has no experience with local construction contractors, providers, laws, or logistics. This situation can be improved relative to prior approaches, but the reality remains that few companies have the capital to fully integrate.

Shin Kori 4 makes a better poster than Falcon 9

There is a lot the nuclear industry can learn from other highly technical, integrated systems design challenges, including launch vehicles, but also medical devices, automotive, and defense. We need to be rigorous about questioning whether “the nuclear way” is the only way. Along those lines, while testing and vertical integration cannot address all problems, that also doesn’t mean it is a waste of time; the nuclear industry in the US has not seen any new integrated system development and deployment since the 1970s; designing new systems is fundamentally different from maintaining or modifying existing ones and its a skill we need to foster. Testing provides real world data that can cut off endless design-analysis loops.

Still, I’d prefer the nuclear industry spend at least as much time looking at the successful nuclear buildouts in France, South Korea, and Japan as they do successful design in different industries. Those countries applied standard designs with the same utility, AE, EPC and regulator, to great effect. If we want to hang a poster on the wall, perhaps it should be of South Korea’s Shin Kori Unit 4, built in 2015 for approximately $2,900/kW in 2025 dollars (though certainly caveats are needed on the cost).

If we could build a nuclear plant in the US today at that cost, it would be cheaper to write off a brand new combined cycle natural gas plant, scrap it, and build that nuclear plant.

I think that’s cooler than landing a first stage.

This article first appeared here and was hosted here with permission from the author for ease of access for non-LinkedIn users.