Microreactors vs Renewable Grid: Cost Challenges and Outlook

The discussion begins by revisiting a previous calculation on the cost of a fully renewable energy grid (solar and wind) with storage and continental transmission, estimated at $131 per megawatt-hour. This figure, derived from 43 years of weather data to ensure 99.9% grid reliability, serves as an upper bound or "litmus test" for evaluating other carbon-free energy technologies. For comparison, the current wholesale price of electricity in Boston is around $50 per megawatt-hour, making the all-renewable scenario about 2.5 times more expensive.

The conversation then shifts to nuclear energy, specifically the capital cost of building a reactor. The current Energy Information Administration (EIA) prediction for an AP1000 reactor is $7,700 per kilowatt of capacity. This sets a benchmark for nuclear energy to beat.

Microreactors: A Closer Look

Microreactors are a significant focus in the fission industry, with many startups and research efforts dedicated to their development. The core question is what makes these microreactors truly new or better, as simply being "different" doesn't guarantee improvement.

Historically, smaller reactors have existed. The term "Small Modular Reactor" (SMR) typically refers to reactors below one-third the size of a full-scale plant, while "microreactors" are even smaller, often below 150 MW thermal. Major nuclear vendors focus on large reactors, but startups are pursuing smaller designs, revisiting a path previously abandoned.

The reason for abandoning small reactors historically was their lack of capital efficiency. A regression analysis of 342 reactors showed that for every doubling of reactor size, the cost decreased by 20%. This scaling relationship suggests that smaller reactors are inherently more expensive per kilowatt of capacity. For instance, a single module of a NuScale reactor was expected to cost 2.35 times more than an AP1000 on a per-kilowatt basis. While this scaling was derived from Light Water Reactors (LWRs), the fundamental geometric volume-to-surface area relationship that drives it is expected to hold for other reactor types, even if the exact formula changes.

Modular design, while often touted as an advantage, isn't new. The AP1000 itself is modular, with components built offsite. However, this didn't significantly reduce costs in practice, sometimes even leading to rework due to fit issues. Therefore, the true advantages of microreactors must lie elsewhere: either in fundamentally different technologies or in a new approach to fabrication, such as factory manufacturing.

Proposed Microreactor Concepts and Their Challenges

NuScale

NuScale's design, a mature concept with 20 years of R&D, involves six 77-megawatt modules submerged in a giant water-filled bathtub. This design aims for enhanced safety by ensuring abundant coolant. Initially designed at 50 MW thermal, the modules were upsized to 77 MW for economic viability, demonstrating the persistent influence of cost scaling. To achieve economic feasibility, NuScale's approach involves housing six modules in a single building, sharing a containment, swimming pool, steam turbine, site, and licensing. This effectively makes it a full-sized reactor with its nuclear core broken into smaller pieces.

Holtec

Holtec proposes a traditional PWR scaled down to one-third to one-sixth size, without significant design changes.

Last Energy

Last Energy offers a NuScale-like concept but pushes the size down to 20 MW. This smaller size is expected to increase costs further (estimated 3.6 times more expensive than AP1000). They replace the expensive concrete containment with a 12-inch-thick solid steel vault and use air cooling, burying the reactor underground. This design prioritizes safety and reduced water issues over cost, targeting customers who value safety above all else. However, the existence of such a market, especially given that many data center contracts with these startups involve selling electricity at regular grid prices, remains questionable.

Deep Fission

Deep Fission proposes even smaller 15 MW reactors buried one mile underground. This concept aims to eliminate containment by leveraging the earth as a shield. However, drilling deep, perfectly round holes is challenging, and servicing such buried reactors would be extremely difficult. This approach, while appealing to the "better safety" crowd, is unlikely to be cost-effective.

High-Temperature Gas-Cooled Reactors (HTGRs)

These reactors, like those proposed by the now-bankrupt UltraSafe (and its successor Valor) and X-Energy, use TRISO fuel (sand-like grains in a graphite matrix) cooled by non-reactive helium gas. Advantages include simpler design, no phase change, reduced complexity, and potentially no need for containment due to the fuel's high temperature resistance. The downside is low power density, meaning a large reactor for little power output, which negatively impacts economics. However, low power density can be spun as an advantage for passive radiative cooling in accident scenarios.

X-Energy uses a pebble bed concept where fuel pebbles are continuously cycled, while others use prismatic blocks. Historically, pebble bed reactors have faced technical challenges like abrasion and complex fuel sorting.

HTGRs are not new; several have been built and operated historically. - Fort Saint Vrain: The largest HTGR, it cost $12,000 per kilowatt (inflation-adjusted) due to a 15% lifetime capacity factor (85% of the time spent on repairs). - AVR: A pebble bed reactor with a 62% capacity factor, it is now the most strontium-contaminated facility globally due to temperature excursions and fission product diffusion. - HT-10 (China): A prototype with very high costs and a current capacity factor of 16% for its successor, HTRPM, indicating similar operational issues.

The historical performance of HTGRs suggests that claims of their cost-effectiveness should be viewed with skepticism, as smart people abandoned them for PWRs in the 1960s for good reasons.

Molten Salt Reactors (MSRs)

MSRs aim to improve power density by replacing helium coolant with molten salt, typically FLiBe (fluorine, lithium, beryllium). Challenges include: - Tritium production: Lithium's neutron absorption creates tritium, requiring expensive isotopic modification of lithium. - Corrosion: Chemical contaminants make FLiBe highly corrosive, a major unsolved issue. - Material lifespan: Vessels made of 316H stainless steel, as planned for Kairos reactors, would only last about eight years at operating temperatures, far short of the desired 40-60 year lifespan.

While MSRs offer four times the power density of HTGRs, potentially reducing hardware costs, this is offset by frequent component replacement and unsolved corrosion issues. Water reactors, by comparison, offer 5-10 times better power density, use cheap water, and have well-understood corrosion management. MSRs, despite being discussed for 23 years, have never been built and face many unsolved problems.

Liquid Fuel Molten Salt Reactors

These reactors dissolve uranium directly into the molten salt, making the fuel itself liquid. While many companies are pursuing this (Terrapower, Terrestrial, Copenhagen, Naarea, Elysium, Natura), the idea is less developed and faces significant challenges: - Fission product management: All fission products are within the liquid fuel, making the entire reactor radioactive. Noble metals plate out on surfaces, making maintenance extremely difficult and requiring full radiation protection. - Gaseous fission products: Iodine and radio-krypton bubble out, requiring complex systems to process, cool, and safely store these radioactive gases. - Historical performance: The Molten Salt Reactor Experiment (MSRE), a 7 MW reactor, operated at 30% capacity and had an estimated cost of $72,000 per kilowatt (capacity-corrected), far from competitive.

The general conclusion for Generation IV reactor concepts is that they move away from the economics of PWRs and, despite decades of discussion, have not been widely built due to reliability and cost-efficiency challenges.

Nuclear Batteries (eVinci Reactor)

Westinghouse's eVinci reactor is a TRISO-fueled, graphite-moderated reactor that uses heat pipes for cooling instead of flowing helium. This represents a fundamentally different approach to reactor design.

Cost Predictions for Nuclear Batteries

Cost predictions for these abstract, battery-style reactors vary widely. - First-of-a-kind (FOAK) costs: These are predictions for building a single reactor. - N-th-of-a-kind (NOAK) costs: These predict costs after many units have been built, assuming learning effects.

All predictions assume 95-100% capacity factors, meaning the reactors are not load-following. If they were required to load-follow, costs would increase by approximately 1.4 times.

Specific examples of cost predictions: - CANDU (CAREM): A 25 MW traditional PWR, not factory-made, currently being built in Brazil. Based on actual costs, electricity is estimated at $370 per megawatt-hour. - Jacopo Buongiorno (MIT): Most optimistic estimate, only providing NOAK costs. By comparing reactor internals to a GE 90 jet engine, he estimates NOAK costs between $85 and $355 per megawatt-hour. With load-following correction, this becomes $120-$500 per megawatt-hour. - NEI (Nuclear Energy Institute): Published estimates of $270 and $148 per megawatt-hour, but without any clear basis for their numbers. - Idaho National Lab (INL): Their "Design A" has a FOAK cost of $2,174 per megawatt-hour. Their NOAK cost was derived by asking colleagues for their opinions. - Department of Defense (Project Pele): A single reactor concept (1-5 MW) with high uncertainty and little public detail.

A key takeaway is that no one is claiming these battery-sized microreactors are cheaper than conventional nuclear power. Their market is likely outside the grid, in heat markets or remote locations, rather than solving the general climate change problem through cheap electricity.

Realism of Cost Estimates and Learning Rates

Reactor designs tend to become more expensive as they mature. NuScale's cost estimate, for example, has doubled multiple times since 2007 as the design was filled out. X-Energy and GE-Hitachi have seen similar cost escalations (factors of 4 and 4.25, respectively).

A rule of thumb based on the AP1000 experience: - Early design to start of construction: Price triples. - Construction-related issues: Price doubles. - The Congressional Budget Office found that, on average, 75 US nuclear plants cost 207% more than originally estimated.

This suggests that initial cost estimates for immature designs should be multiplied by a factor of 6-9 to reflect final costs. This phenomenon is not unique to nuclear but is common in mega-projects, though some argue it's exacerbated by venture capital culture encouraging overly optimistic projections.

Factory Fabrication and Learning Rates

The hope for microreactors lies in factory fabrication and mass production. - Learning rate based on cumulative units: Historical data for non-factory-fabricated reactors shows a 1% cost reduction for every doubling of units. A Soviet naval reactor factory manager claimed a 5% cost reduction for every doubling of factory-produced reactors. While better, this is still much slower than solar cells (which have seen dramatic cost reductions due to high production volumes). Even with a 5% learning rate, achieving cost competitiveness would require building hundreds of thousands, if not a trillion, reactors.

• Learning rate based on production rate: The rate of production is crucial. A study by NREL on stationary fuel cells found a 30% cost reduction for every tenfold increase in production rate. If this applies to nuclear, producing 50 units per year (to replace current natural gas additions in the US) could bring costs down to $20,000 per kilowatt, competitive with traditional nuclear. Producing 500 units per year could further reduce costs to $14,000 per kilowatt, but market size quickly becomes a limiting factor.

The automotive industry provides data on minimum efficient scales: 200,000-300,000 units for a car plant, 400,000 for engines, and 500,000 for transmissions. Similar data is needed for nuclear to understand optimal production volumes.

Timeline for Microreactor Deployment

Reactor development is a lengthy process: - NuScale: 25 years from conceptual inception to first unit in service (project now halted). - AP1000: 20 years from design conception to final license approval, plus 5-10 years for construction, totaling a quarter-century.

This long timeline means that microreactors coming online in 20-25 years will enter a world transformed by massive wind and solar installations. They will need to be load-following to compete or focus on non-electrical grid applications like industrial process heat.

Potential Markets for Microreactors

Proposed markets for microreactors include: - Disaster relief (low volume) - Arctic and remote communities (e.g., 40-60 reactors for Alaska and Canada, insufficient for mass production) - Mining operations (a few dozen worldwide, insufficient for mass production) - Island communities (limited market) - Remote military installations (hundreds, but concerns about deploying nuclear in war zones due to contamination risk from kinetic explosions).

Developing World Electrification

A significant potential market is electrifying the developing world. Satellite data on nighttime lights, combined with GIS data, suggests that 630 million people in areas with low electrification could theoretically be served by small nuclear reactors. However, this is contingent on: - Governance and regulation: Excluding countries without strong governance (e.g., below Ukraine's standard) reduces the population to 98 million. - War zones: Further exclusions for conflict areas.

Even with these exclusions, there could be a market for 300,000 reactors, a number that could potentially drive economies of manufacturing. African electricity prices ($80-$600 per megawatt-hour) might seem to make nuclear affordable, but this ignores the reality that most African electricity sectors operate at a loss, subsidized by governments. Scaling this to 100 million additional people is financially unfeasible.

Finally, a major challenge is the human capital required. If each reactor needs two engineers, 400,000 reactors would require 4 million nuclear engineers, necessitating a massive expansion in training or fully automated, computer-controlled operations.

Conclusion

For microreactors to succeed, the focus should be on: - Long-term vision: Thinking 25 years into the future. - Factory fabrication: High production rates and large numbers. - Standardized design: A single, widely adopted design rather than many competing ones. - Non-electrical applications: Prioritizing process heat or other industrial uses over grid electricity, unless they can effectively load-follow.