What To Do About Nuclear

Why is innovation in nuclear energy so slow and how can we change the paradigm?

Construction of Vogtle Units 3&4. Source: Southern Company

I am going to start this article with a statement that is technically true but disagreeable to the majority of the energy industry: Nuclear power can be the lowest cost generation technology. Not the lowest cost “zero-carbon” or “dispatch-able” generation technology, and not at the expense of safety; the lowest cost generation PERIOD.

Now, that may seem an absurd statement given the vast majority of civilization’s experience with nuclear power:

• >$8,000/kW overnight capital cost
• $30/MWh long run marginal cost (cost of making power after the plant is paid for) — 80% fixed O&M, 20% fuel (which is capitalized so it is also more fixed than variable)
• >$90/MWh LCOE (when realistic weighted average costs of capital are used)
• 300+ staff
• 10+ years to build

To understand why a technology with those attributes can be the lowest cost energy generation technology requires some Elon Musk-style first principles thinking:

1. Nuclear generation technology can be orders of magnitude greater power density than wind or solar. Even today’s nuclear technology (e.g. Westinghouse AP1000) produces about 100 times more power than the solar insolation falling on the same area of the Earth’s surface. This means there is the potential for a lot of energy to come from very little land and capital infrastructure (overnight capital cost can be orders of magnitude lower than it is today)
2. Nuclear fuel is 10,000,000 times more energy dense than fossil fuels. This means there is the potential for a lot of energy to come from very little fuel. The only physically intrinsic marginal cost for a generation technology is the cost of fuel (O&M is a feature of the specific equipment design). Therefore, the marginal cost of nuclear can be orders of magnitude lower than it is today (on par with zero marginal cost wind and solar).
3. Unlike wind and solar, nuclear power can be dispatched. It has an inventory of potential energy on which it can draw rather than opportunistically harvesting energy wind or sun. This means that whatever upfront capital costs do exist, they can be more easily amortized over the asset’s productive hours, lowering the LCOE.
4. There is nothing inherent about nuclear fission or fusion that require humans present (staff can be reduced to zero with an autonomous design
5. There is nothing inherent about nuclear fission or fusion that takes a long time to build. Build time is a product of machine design, manufacturing approach, and supply chain management (hypothetically, building a nuclear generator could be as fast as manufacturing a solar panel, or faster)
6. Nuclear Power does not have any intrinsic externalities. It does not produce CO2 (like fossil fuels); it does not kill birds, make noise, or create a visual obstruction (like wind); and it does not alter massive land areas and ecosystems (like hydro and to a much smaller extent solar).
7. Nothing technically limits where nuclear power is sited (it does not even need oxygen and exhaust like diesel generators). It can generate energy anywhere.

Very few energy technologies have the potential for a 10X improvement in any respect. A quick look at nuclear power’s underlying physics shows it has the potential for multiple orders of magnitude improvement in almost every respect. So why are we no where near realizing this potential? Why does nuclear power seem to get more expensive over time when every other energy technology is getting cheaper?

There have been many studies that point at things like construction management, contract structuring, misaligned incentives, overly conservative regulations, lack of standardization, lack of R&D funding, etc.

However, none of these get at the root cause of the industry’s failure: nuclear power technology has lacked an iterative approach to development. Every technology that has shown meaningful advancement in features, performance, or cost has done so via an iterative feedback loop between customers and design engineers. A minimum viable product is built, sold, tested, and reviewed, then engineers build version 2, 3, 4, etc. Incremental learning combines with leaps of creativity to ultimately make a product that is far better than anyone originally thought possible with version 1. Smart phones, cars, airplanes, even rockets (thanks to SpaceX) have all gone through this cycle. Step 1 is to achieve product-market fit and step 2 is to scale the product via an efficient feedback loop of iterative improvements. This process is easiest to conduct for software, a bit less easy for personal electronics, and progressively harder for products with longer replacement schedules (cars, airplanes, buildings). This process is also made harder for more complex products that are highly regulated, but as SpaceX has shown, it still very much achievable. Nuclear power faces a combination of a long-lived product, high technical complexity, high regulations, and one additional challenge: a low profit-margin commodity market. The one thing SpaceX had going for it is a high margin business with little competition.

However, this does not mean that an iterative development cycle cannot be established for nuclear, it just requires a deeper look at the market and a few new strategies:

Understanding the competition

There are primarily three types of generation getting built in the U.S. today: wind, solar, and natural gas

Wind

• Overnight Capital Cost: $1,300/kW
• Long-run Marginal Cost: $3/MWh
• LCOE: $18-$35/MWh
• Staff: 1–5
• Construction Time: 2 years

Solar

• Overnight Capital Cost: $1,300/kW
• Long-run Marginal Cost: $1.73/MWh
• LCOE: $25-$35/MWh
• Staff: 1–5
• Construction Time: 1 year

Natural Gas

• Overnight Capital Cost: $900/kW
• Long-run Marginal Cost: $22.36/MWh (6% fixed cost, 94% variable, assuming $3/mmBTU gas)
• LCOE: $30-$40/MWh
• Staff: 30–40
• Construction Time: 2 years

(data from the EIA’s Annual Energy Outlook 2020 1, 2)

Takeaway: Nuclear needs to be cost competitive with wind, solar, and natural gas in all respects

Designing for the Grid’s Future Needs

As the world moves towards decarbonization, the growth of wind and solar generation capacity is inevitable. They have proven to be low cost, relatively easy to deploy, and clean. Their biggest shortcoming stems from integrating their intermittent generation with the rest of the grid. Energy storage, demand-side flexibility, and transmission build-out will help this integration but ultimately a dispatchable zero-carbon generator is needed to fill the majority of the gaps between supply and demand. Historically, natural gas has been filling this role because it has a high variable cost and low fixed cost. Even though nuclear plants can technically ramp power up and down today (albeit with some limitations), there is little financial incentive to do so.

Takeaway: Nuclear needs to be designed to replace the role of natural gas, which means it’s fixed costs need to drop substantially below its variable costs (fuel) or it needs to be able to divert its power to a production process that can be inventoried (see upstream net metering for more on this).

Developing a minimum viable product

Nuclear plants are traditionally large bespoke civil engineering projects akin to building an international airport. Several thousand specialized workers move to the project site for about a decade, many custom parts are fabricated on-site, and very few opportunities are presented for learning or improving anything once this construction has started. Complex construction schedules and critical paths must be strictly adhered to for 10 years or costs and delays will compound, which is almost always what happens. Before this process can begin, billions of dollars must be financed, contracts and liabilities distributed, and site studies finalized. Before even those items can begin, there has usually been 10–15 years of reactor design, regulatory review, and certification (usually costing as much as $1B itself). Following this current pace, a new 1 GW reactor concept started today wouldn’t produce a $1 of revenue or a single MWh for probably 25 years (at the earliest). Discounting that revenue for 25 years at a rate appropriate for the level of risk involved makes new nuclear a pretty unappealing investment under the current model. Even the new small modular reactor (SMR) concept, NuScale, took 15 years and $1B to design and will need another 4–5 years to build if everything goes right (and despite being small and modular, most of the capital cost is custom civil engineering site work).

For new commercial nuclear power to have any future, this 25 year, multi-billion dollar process needs to shrink dramatically. The first power producing reactor, EBR-I, took 2 years to build between 1949 and 1951. Nuclear innovators need to revisit this smaller approach to deploying their first product. Wind power technology did this before it started scaling to the 250+ m tall, 6+ MW turbines of today. Photovoltaic technology did this with rooftop and satellite deployments before low-cost utility-scale installations. Reactor designers need to spend more time thinking about the rooftop solar equivalent of a nuclear reactor, a nuclear distributed energy resource (DER). Rather than targeting the end-game wholesale energy market, nuclear entrepreneurs need to focus on niche off-grid markets for micro reactors (1–5 MW or less) such as industrial and university campuses, defense applications, remote mining operations, mobile applications, anyone willing to pay for a premium energy product rather than a mass-market commodity. It will be much easier to see the low-cost mass-market solution once the premium version works well. Oklo and other micro reactor developers are beginning to adopt this strategy.

Takeaway: entrepreneurs need to focus on micro reactors and premium markets to achieve revenue earlier and iterate towards a mass-market product.

Aurora reactor 1.5 MWe pilot from Oklo

Nuclear as a Product

In stark contrast to the bespoke civil construction format of nuclear, wind, solar, and natural gas plants are largely standardized, pre-fabricated equipment purchased “off-the-shelf”. A simple foundation is laid, and the equipment is shipped-in from a factory and bolted down. If more power is desired, more turbines or panels are installed. Rarely are any modifications made (certainly not for the lowest cost projects).

Nuclear power needs to migrate to a more “productized” form factor. Not only must the units be standardized, but a large majority of the overnight capital cost for a project must come from the factory-produced components rather than the one-off site work. To be clear, this was not the case for the AP1000 at Vogtle (despite being marketed as “modularized” construction) nor will it be the case for the upcoming NuScale design. The best way to achieve “learning-by-doing” and realize a cost curve that trends downward vs time (rather than upward) is to migrate as much fabrication into a controlled, centralized factory environment as possible.

One way to achieve this is by shrinking the size of the reactors (as is the strategy behind SMRs and micro reactors), but “productizing” nuclear does not have to mean shrinking it. Economies of scale are a very real thing and later iterations of a new reactor design may benefit from large reactor size as long as it can still be factory built. One way of achieving this is demonstrated by the shipyard-manufactured nuclear plants proposed by MIT and others. Building a nuclear plant in a shipyard enables much larger reactor sizes while preserving the same construction team, the same factory, and economies of mass-production.

Takeaway: Nuclear needs to be turned into an “off-the-shelf” product with all of the high-quality, complex, regulated work conducted in a centralized factory off-site.

Audacious Thinking

For 70 years, nuclear has been stuck in the same form factor of large light water reactors, requiring many workers, significant security, and essentially the same fuel technology. The most meaningful innovations in nuclear during that time were the “learning-by-doing” operators exhibited by maximizing capacity factors and minimizing anomalous maintenance activity. Politicians, financiers, and the public largely only think of nuclear in this somewhat antiquated way, causing many to think of nuclear as a technology of the past. In order for nuclear to achieve the technical potential outlined at the beginning of this article, innovators need to pursue some audacious goals. Key examples include:

Full Autonomy

Autonomous nuclear generators have been successfully deployed many times in space applications and have been extensively explored for military applications. Research reactors have operated for decades with minimal operational and security staff (in some cases even in urban environments). There is no technical reason nuclear plants could not be designed for full or nearly full autonomy. Inherent safety and control systems could minimize human intervention (in fact the biggest risk to newer designs like NuScale’s is human intervention itself) and new materials and sensors could minimize maintenance requirements. Even new security technology leveraging modern surveillance, AI, and robotics from companies like Anduril, Planet Labs, and Boston Dynamics could drastically reduce or eliminate the hundreds of armed security guards.

High Fuel Burn-up

Current nuclear fuel technology leaves over 95% of the potential energy in the uranium untapped before replacing it. Newer concepts such as liquid fuel molten salt reactors and other breeder reactor designs could substantially increase the fuel utilization and shrink the fuel costs. (Note: this is perhaps less important until fixed operating costs are shrunk first).

Direct Energy Conversion

Many nuclear plants designs rely on low temp Rankine cycles that only achieve ~33% efficiency. Higher temp reactors can use this into the low 40s and very high temp gas reactors can start to implement Brayton cycles to achieve 50–60%. However, all heat-engine power cycles suffer from cost, complexity, size, and maintenance (at least until supercritical CO2 achieves commercialization). A really audacious idea would be to try to pursue something like direct energy conversion from the fission products themselves. There are a few ideas on how to do this (e.g. U.S. patent 5,586,137) that would require significant R&D but could completely change the economic and business model behind nuclear power.

Takeaway: Radical new features are needed to help commercial nuclear power realize its potential.

Rethinking Investment

Historically, the DOE has been in the position of “picking winners” when it comes to nuclear technology: first with GE’s PRISM design, then with NuScale’s SMR, and potentially again as the next generation reactor’s like Terrapower’s MCFR and X-Energy’s Xe-100 gain momentum. This strategy has largely stemmed from the high capital costs and long development time nuclear technology has exhibited, but this funding approach also reinforces the bad habits that cause such massive upfront investments. Nuclear’s success cannot be dependent on the political whims of congress or the DOE.

While heavily criticized as being incompatible for “hard-tech” innovations, venture capital creates the incentive framework necessary to spawn lean, focused businesses. As long as the market-size and upside potential are realistically plausible, venture capital is prepared to take the necessary risk. That being said, there may be a few ways policy and DOE funding can help make venture capital and private equity a bit more targeted and patient:

1. A tax break for distributions made to limited partners from venture funds that targeted nuclear or perhaps clean tech generally (could be capped at a certain amount e.g. $50 million)
2. The DOE could provide non-dilutive matching funds for nuclear startups that raise from venture capital (again with certain limits on funding and the stage of the company)
3. For very aggressive intervention, the DOE could allocate a number of funds (e.g. 3 x $300 M) to general partners (GPs) at existing VC firms with the explicit criteria to invest in nuclear energy startups. GPs would be compensated with the typical 2% management fee and 20% carried interest but would also need to match at least 1% of any investment with their own money (to bear some risk and ensure high quality investments). Such a fund could have much longer time horizons for returns than the typical 7–12 year life of traditional closed-end venture funds.
4. The DOE could also match funding (e.g. 50%) for any venture fund raised from private capital sources that have the criteria to invest in nuclear energy tech (turning a $100M fund into a $200 M fund). This would catalyze private capital and put more onus on the GPs to perform.

These ideas would need further refinement but allude to the types of strategies the DOE could try to incentivize faster and more impactful nuclear innovation.

Takeaway: DOE funding resources for nuclear innovation need to find ways of being complimentary to the existing venture and private equity models that develop breakthrough innovations.

Jake Jurewicz offers professional consulting at 1st Principles Consulting, LLC.