Nuclear Reactor Types: A Practical Reference

The PWR is the most common reactor on earth, making up ~70% of the world's ~440 power reactors. Ordinary water at 155 bar serves as both coolant and moderator. The pressure keeps it liquid at ~325 °C, and it hands its heat to a separate secondary loop that raises steam, so the two-loop separation keeps radioactivity out of the turbine. The design descends directly from Rickover's naval program. It became dominant because it was first, it was funded, and it has been iterated on endlessly.

The PWR has an operating record nothing else comes close to (tens of thousands of reactor-years), a global supply chain, mature regulation, and 90%+ fleet capacity factors in the US. Weaknesses: 155-bar pressure demands massive forgings and containments; ~33% thermal efficiency and ~300 °C output rule out industrial-heat markets; large cores need active safety systems layered in depth; and Western new-builds routinely run far over budget.

Default to the PWR for gigawatt-scale baseload on a large grid whenever you have a serial builder (Korea, China, or a committed national program) and state-grade balance sheets. It's the lowest-physics-risk, lowest-regulatory-risk path to nuclear at scale, and it's the right anchor technology for a first-time nuclear nation that wants a proven design with a functioning global fuel market. Look elsewhere if your grid is under ~10 GW, if you need industrial heat above 300 °C, or if the buyer can't carry a $10B+ decade-long project. In those cases light-water SMRs give you the same physics at a financeable size, and HTGRs are the ones that serve the heat market.

Electric output 900–1,600 MWe per unit from 2,700–4,600 MWt · light water as both coolant and moderator at 155 bar · outlet around 325 °C for roughly 33% thermal efficiency · UO2 fuel at 3–5% enrichment, 45–55 GWd/t discharge burnup · refueling outage every 12–24 months · overnight capex $2.5–4k/kW in China and Korea, $12–15k/kW in recent Western first-of-a-kind builds.

Vogtle 3&4 — first US new builds in 30 years, ~$35B for 2.2 GW; Barakah (UAE, four APR-1400s delivered near schedule); China connecting several Hualong Ones yearly; France's 56-reactor fleet.

The same design costs $2.5–4k/kW in China and Korea and $12–15k/kW in recent US/EU FOAK builds. The gap comes from serial execution rather than physics. Fuel comes from a functioning global LEU market (~$5–10/MWh contribution). So the PWR is both the safest bet in nuclear and the reason the SMR and advanced-reactor industry exists at all, since that industry grew up around Western PWR cost problems.

6 minPressurized Water Reactormckuyver · 100k+ views

6 minsaVRee Snacks #13 -How Pressurized Water Reactor (PWR) Power Plants Work!saVRee · 10k+ views

4 minPressurized Water Reactor: General ProcessAdam Sedgwick · 50k+ views

Nuclear Power Reactors (World Nuclear Association) · The Future of Nuclear Energy in a Carbon-Constrained World (MIT Energy Initiative)

The BWR is a simpler version of the PWR and makes up ~14% of the world fleet. Water boils directly in the core at ~75 bar, and that steam drives the turbine with no secondary loop. You get fewer components, half the pressure, and a bigger vessel. In exchange, the turbine hall becomes mildly radioactive during operation. The design also carries the reputation of Fukushima Daiichi, a 1960s BWR whose station-blackout vulnerability defined the accident.

The plant is simpler, with no steam generators (historically the PWR's most troublesome component), lower pressure, and good load-following via recirculation flow. Weaknesses: the radioactive steam path complicates maintenance; the Mark I containment's Fukushima performance drove costly retrofits fleet-wide; and new BWR construction stalled globally for a decade. The design's future now rides almost entirely on its SMR descendant, the BWRX-300.

Choose a BWR today mainly where the lineage already exists: Japanese restarts, life extension of the operating fleet, and jurisdictions with BWR-trained operators and regulators. For new capacity the decision has moved down-market. The BWRX-300 is the live way to buy the design's simplicity, so put it on the shortlist for 300 MW-class grids and coal-site replacements. Don't launch a new full-scale BWR program from scratch; for new large builds a Gen III+ PWR is the default.

Electric output 600–1,400 MWe per unit · light water as coolant and moderator, boiling in the core at about 75 bar · saturated steam near 290 °C for roughly 33% thermal efficiency · UO2 fuel at 3–5% enrichment, 45–50 GWd/t discharge burnup · refueling outage every 18–24 months · the BWRX-300 derivative rated at 300 MWe.

~60 operating units across the US, Japan, Sweden, Taiwan (retiring); Japan's post-Fukushima BWR restarts (Onagawa, Shimane); Hitachi-GE and GE Vernova as the surviving vendor line.

Economics are comparable to a PWR, with a somewhat cheaper plant and costlier maintenance logistics. The more interesting part is the lineage. GE Hitachi took the BWR's simplicity argument down to 300 MW as the BWRX-300, currently the West's most-ordered SMR, which puts the old design's economics at a financeable size.

19 minHow Boiling Water Reactors Work (BWR Working Principle)saVRee · 50k+ views

8 minThe Most Used Nuclear Reactors: PWR and BWREnergy Encyclopedia · 50k+ views

4 minBWR reactor (Fukushima) - How it worksEnergy Encyclopedia · 10k+ views

Fukushima Daiichi Accident (World Nuclear Association) · BWRX-300 Small Modular Reactor (GE Vernova)

The VVER is Russia's PWR lineage: horizontal steam generators, hexagonal fuel assemblies, and an unbroken design evolution from Soviet VVER-440s to the modern VVER-1200 with passive safety systems. The physics is ordinary PWR physics. The business model is what sets it apart. Rosatom sells the reactor, the financing, the fuel, the operators, and the waste take-back as one state-backed package, which has made the VVER the default reactor of the developing world.

The VVER has a long operating record across 30+ countries' worth of units, genuinely competitive delivered cost, and it's the only product offering full-stack financing-to-fuel service. Weaknesses: buying one creates decades of dependency on a sanctioned state; Western fuel alternatives (Westinghouse's VVER fuel) exist but are recent; and geopolitics now hangs over every project. Finland canceled Hanhikivi, while Hungary's Paks II proceeds.

The VVER is the rational choice if you're a first-time nuclear nation without investment-grade credit and you want a single counterparty for financing, fuel, operations, and waste. That's the buyer profile of Egypt, Bangladesh, and Turkey. Treat the decision as geopolitical before technical: only accept it if six decades of dependency on Moscow is tolerable and sanctions exposure is manageable. If you're a NATO/EU-aligned buyer, or you're otherwise hedging Russia risk, price the APR-1400 as the closest Western analogue on delivered cost. If you already run a VVER fleet, qualifying Westinghouse fuel is the way to reduce the dependency.

Electric output 440–1,200 MWe per unit · light water as coolant and moderator at roughly 160 bar · outlet around 320 °C · hexagonal UO2 assemblies at 3–5% enrichment, roughly 50 GWd/t discharge burnup · refueling outage every 12–18 months · export price around $5–6k/kW with Rosatom state financing.

VVER-1200s operating at Novovoronezh and Leningrad; export builds at Akkuyu (Turkey, four units), El Dabaa (Egypt), Rooppur (Bangladesh), Kudankulam (India), Tianwan (China); ~20 units in Rosatom's export book.

Quoted export prices run ~$5–6k/kW with state financing at concessional rates, a package Western vendors structurally cannot match. Rosatom's real product is geopolitical lock-in: 60-year relationships covering fuel, service, and spent-fuel return. For market analysis, treat the VVER as the pricing floor and the diplomatic competitor that every Western SMR export campaign is really up against.

11 minVVER1200 - The Russian VVER-1200 (NPP-2006 or AES-91)Micro-Simulation Technology (Micro Simulation) · 10k+ views

5 minVVER 1200 ConstructionAtomenegromash · 100k+ views

11 minRussian Nuclear Power Plant Reactor VVER 1200 (animation)Marsh Swog · 1k+ views

Nuclear Power in Russia (World Nuclear Association) · VVER Working Group Design Comparison Report (OECD-NEA MDEP)

Light-water SMRs are a bet that nuclear's cost problem comes from construction rather than physics. You shrink a proven light-water reactor to 50–470 MWe, build most of it in a factory, integrate components into one vessel (steam generators inside, in NuScale's case), and use passive safety to shrink the emergency planning zone. The physics is deliberately boring (LEU fuel, water coolant, known regulation) so that the whole innovation budget goes to manufacturability and financing size.

Unit size is financeable at ~$2–5B rather than $15–35B, decay-heat removal is passive, siting is flexible (retiring coal plants included), and the regulatory path reuses six decades of LWR precedent. Weaknesses: small reactors give up economies of scale, so per-kW costs are structurally higher until factory volume compensates, and that requires an order book nobody has yet. NuScale's Utah project died at ~$20k/kW, and the "modular" savings are still a projection rather than measured data.

Choose a light-water SMR when your binding constraint is financing size or grid size rather than $/MWh. That covers utilities that can't carry a $15–35B single-asset risk, 300 MW-class grids, coal-site conversions with transmission already in place, and data-center offtakers paying a premium for firm power on a date. Prefer designs with concentrated order books (BWRX-300 today), because the whole thesis rests on the manufacturing learning curve, and an orphan design gives you small-reactor costs without the volume that pays for them. If you can finance a full-size PWR from a serial builder, do that instead; it beats any FOAK SMR on cost per kW.

Electric output 50–470 MWe per module · light water as coolant and moderator · outlet around 300 °C · UO2 fuel below 5% enrichment · refueling every 18–24 months · $2–5B per unit instead of $15–35B for a full-size plant · overnight capex $4–6k/kW claimed at NOAK, $8–15k/kW realistic for first-of-a-kind.

Darlington's four-unit BWRX-300 program (first concrete 2025), Linglong One nearing completion at Changjiang, TVA Clinch River, Amazon/X-energy and Google/Kairos-style hyperscaler offtakes reshaping demand.

Vendors pitch $4–6k/kW at NOAK; realistic FOAK expectations are $8–15k. The math only works if one design wins enough repeat orders to move down a manufacturing learning curve, so order-book concentration (Ontario + TVA + Poland on BWRX-300) matters more than design elegance. Data-center power purchase agreements at premium prices are the demand shock that may finally fund the first fleet.

8 minSMRs Explained | Why Small Modular Reactors Are Changing Nuclear DeploymentWorld Nuclear Association · 100k+ views

2 minNormal System OperationNuScale Power · 100k+ views

2 minEnergy 101: Small Modular ReactorsU.S. Department of Energy · 10k+ views

The NEA Small Modular Reactor Dashboard, Third Edition (OECD-NEA) · Small Nuclear Power Reactors (World Nuclear Association)

The PHWR was Canada's answer to enrichment scarcity. It moderates with heavy water (D2O), which absorbs so few neutrons that natural, unenriched uranium sustains the chain reaction. The fuel sits in hundreds of horizontal pressure tubes rather than one massive vessel, which allows on-power refueling, so CANDUs never stop to swap fuel. India imported the design in the 1960s and industrialized the same lineage into its indigenous PHWR fleet.

There's no enrichment dependency (energy sovereignty was the point), on-power refueling keeps availability high, pressure-tube construction avoids giant forgings, and the neutron economy is good enough to burn thorium blends or recycled uranium. Weaknesses: heavy water costs hundreds of millions per plant; pressure tubes embrittle and require mid-life retubing (~$10B-class refurbishments); tritium production has to be managed; and that same good neutron economy is proliferation-relevant, since India's 1974 test traced back to a CANDU-lineage research reactor.

Choose a PHWR if enrichment independence is a strategic requirement, since the design runs on natural uranium from any supplier. It also makes sense wherever the CANDU industrial base already exists (Canada, India, Romania, Argentina), where refurbishments and India's 700 MWe builds are proven, competitively priced work. It's the practical route to burning recycled uranium or thorium blends without new reactor physics. If you don't have a sovereignty argument or a heavy-water supply chain, default to the PWR and its larger vendor ecosystem. And whatever you do, make any PHWR business case carry mid-life retubing as a second capital project.

Electric output 220–900 MWe per unit · heavy water as both coolant and moderator, roughly 100 bar inside the pressure tubes · outlet around 310 °C · natural uranium UO2 bundles at 0.71%, discharge burnup roughly 7 GWd/t · on-power refueling, so no refueling outage · overnight capex $2–3k/kW on India's 700 MWe builds, plus mid-life retubing as a second capital project.

Ontario's Bruce, Darlington, and Pickering fleets (mid-refurbishment through the 2030s); India's 22-unit PHWR fleet and its 700 MWe standardized builds; Qinshan (China), Cernavodă (Romania, units 3–4 revival), Embalse (Argentina).

India builds 700 MWe PHWRs at roughly $2–3k/kW, which is among the cheapest credible nuclear anywhere. Canadian refurbishments have lately run on time and on budget, which is rare for Western nuclear execution. New CANDU sales stalled for decades, but AtkinsRéalis's MONARK 1,000 MWe pitch and Ontario's momentum give the lineage its first real revival window since the 1990s.

2 minHow CANDU reactors workBruce Power · 10k+ views

7 minCandu Reactor RefuelingSteve Packard · 100k+ views

14 minNuclear Physicist Explains - What are CANDU Reactors?Elina Charatsidou · 100k+ views

The Essential CANDU Textbook (UNENE) · Nuclear Power in Canada (World Nuclear Association)

The RBMK is the Soviet graphite-moderated, water-cooled channel reactor, and it's Chernobyl's design. It was built for cheap plutonium-compatible scale using pressure tubes and no containment. Two design choices caused the 1986 accident: a positive void coefficient (boiling coolant raised reactivity instead of suppressing it) and control rods whose graphite tips briefly added reactivity on insertion. Together they turned a test into the worst reactor accident in history. Post-Chernobyl modifications removed the worst characteristics.

The RBMK offered on-power refueling, no enrichment-heavy fuel demands, and no giant pressure vessel, so it was cheap and scalable, which is why 17 were built. Weaknesses: the void coefficient and scram behavior made it unusually unforgiving; graphite fires spread contamination in a way water reactors can't; there's no containment; and the design can't be licensed anywhere outside its legacy fleet. It's the textbook case of physics choices setting up an institutional catastrophe.

There's no deployment case. The design can't be licensed outside its legacy Russian fleet, and it exits with the 2030s retirements. Today it's useful mainly as a reference point: it's the benchmark that every "inherent safety" claim gets priced against, and the case study in how a positive void coefficient compounds with institutional secrecy. If you actually want a channel-type reactor with on-power refueling, look at the CANDU lineage instead.

Electric output 1,000 MWe per unit from 3,200 MWt, and 1,500 MWe at Ignalina · light water coolant boiling in pressure tubes at about 70 bar, graphite moderator · outlet around 284 °C for roughly 31% thermal efficiency · UO2 fuel at 2–2.8% enrichment · on-power refueling, and no containment building · 7 units still operating, all retiring through the 2030s.

Chernobyl (all units closed, 1986–2000); Leningrad, Kursk, and Smolensk units in Russia — seven still operating with lifetime extensions, retiring through the 2030s as VVER-1200s replace them.

The RBMK has no commercial future, so its relevance is analytical. It's the cautionary entry in any reactor comparison: a design that was cheap because it gave up safety margin, run by an institution that couldn't admit the flaw. Every modern "inherent safety" pitch (passive decay heat, negative coefficients everywhere) is implicitly priced against this history.

13 minRBMK Reactors - Design Overview | The Nuclear Reactor that was used in ChernobylAntares · 100k+ views

4 minRBMK reactor (Chernobyl) - How it worksEnergy Encyclopedia · 100k+ views

13 minRBMK: The Soviet Reactor That Was Doomed from the Start | Chornobyl Uncharted Ep 04Chornobyl Family · 500k+ views

RBMK Reactors (World Nuclear Association) · Chernobyl Accident (World Nuclear Association)

Britain took a different path from the rest of the world: CO2-cooled, graphite-moderated reactors. Magnox (1956's Calder Hall, the world's first commercial nuclear power) ran natural uranium in magnesium-alloy cladding. The Advanced Gas-cooled Reactor (AGR) followed with enriched oxide fuel at 640 °C gas temperature. AGRs are still the highest-temperature power reactors ever fleet-operated, and they reach ~41% thermal efficiency, which no LWR matches. The lineage died because nothing was standardized: every AGR station was practically a prototype.

The lineage delivered high efficiency, on-load refueling (Magnox), benign coolant chemistry, and real engineering achievement in graphite-core longevity. Weaknesses: cores and civil works were enormous per MW; each of the seven AGR stations differed enough to give up any series learning; graphite ages unpredictably, which ended up setting fleet lifetimes; and Magnox's dual civil-military plutonium role tangled the program politically. The UK eventually gave up and bought a PWR (Sizewell B).

There's no new-build case. What's left is practical experience in decommissioning program management and the graphite-aging dataset the NDA effort generates at scale. If you're doing HTGR diligence, treat the AGR as the null hypothesis: 640 °C gas-cooled reactors were fleet-operated for decades and still lost on economics, so any modern high-temperature-gas pitch has to explain what breaks that precedent beyond TRISO fuel and factory build. If you want high-temperature gas output today, evaluate pebble-bed HTGRs instead.

Magnox 50–490 MWe per unit, AGR roughly 600 MWe per reactor · CO2 coolant with a graphite moderator · outlet about 400 °C on Magnox and 640 °C on AGR · AGR thermal efficiency around 41%, against ~33% for a light-water reactor · Magnox burned natural uranium metal in magnesium-alloy cladding, AGR enriched UO2 at 2.5–3.5% · 26 Magnox units and 7 AGR stations built, all retired or retiring by about 2030.

Calder Hall (1956–2003), the 26-unit Magnox fleet (all retired; Wylfa last, 2015); seven AGR stations (Hinkley B, Hunterston, Torness…) retiring through ~2028–30; decommissioning managed by the NDA at multi-decade, multi-£10B scale.

The program shows how not to build a fleet: one country, one technology, zero standardization. The AGR's 640 °C output is exactly what today's HTGR ventures promise, which is a reminder that high-temperature gas reactors aren't new physics and that their historic failure was economic. The UK's decommissioning bill is the long tail every national program should price in.

8 minMAGNOX, worlds first commercial scale nuclear reactorOsama Baig · 10k+ views

3 minUnderstanding graphite cores in nuclear reactorsEDF 🇬🇧 · 50k+ views

5 minGas cooled Nuclear Reactors | Skill-LyncSkill Lync · 10k+ views

Nuclear Development in the United Kingdom (World Nuclear Association) · Nuclear Decommissioning Authority (GOV.UK)

The core is made of about 420,000 billiard-ball fuel pebbles, each containing thousands of TRISO particles (poppy-seed-sized uranium kernels wrapped in ceramic layers that retain fission products to 1,600 °C+). Helium coolant exits at 750 °C+. The physics is meltdown-proof in a literal sense: power density is low enough and temperature margins high enough that a core losing all cooling just cools down, which test reactors have demonstrated on camera. Pebbles circulate continuously, so the reactor refuels on-line.

This is the strongest inherent-safety story in fission, it produces high-grade heat for industry and hydrogen, and the TRISO fuel acts as its own containment. Weaknesses: energy density is very low, so cores and vessels are huge per MW; TRISO fabrication is expensive and only now scaling in the West; helium leaks through everything; pebble flow and dust caused real trouble in Germany's AVR/THTR (shut down after mechanical and political failures); and HALEU dependence gates every Western project.

Choose a pebble-bed HTGR when your product is heat rather than electricity: 550–750 °C heat for chemicals, refining, synfuels, or hydrogen, sold to a creditworthy industrial or hyperscaler offtaker who values the walk-away safety case for close-in siting. If emergency-planning-zone size is what decides whether you can get a site, this is the right pick. It's the wrong tool for merchant bulk power, where the energy-density penalty makes LWRs cheaper per kW. Any Western schedule depends on HALEU and TRISO supply, so secure the fuel line before you commit to offtake dates.

Thermal output 200–250 MWt per module for 80–105 MWe · helium coolant with a graphite moderator · outlet 750 °C, raising steam at 550–566 °C for roughly 40% thermal efficiency · TRISO fuel in 60 mm pebbles at 8.5–15.5% enrichment, fission products retained above 1,600 °C · discharge burnup roughly 90–160 GWd/t · around 420,000 pebbles in a 250 MWt core, circulated and refueled on-line.

China's HTR-PM at Shidaowan — twin 250 MWt pebble beds, grid-connected 2021, full power 2022: the world's first operating Gen-IV plant, with follow-on HTR-PM600 planned. X-energy's Xe-100 (80 MWe, quad-packs) backed by Amazon and Dow for the Seadrift chemical plant; Germany's AVR/THTR as the instructive ancestors.

HTR-PM's FOAK cost ran well above Chinese PWRs, so the safety case doesn't come cheap yet. The Western thesis (X-energy) sells what PWRs can't: 565 °C steam and process heat up to 750 °C for chemicals, refining, and data-center campuses, sold via corporate offtakes rather than merchant power. TRISO fuel lines (X-energy's TF3, BWXT) are as much of a moat as the reactor itself.

7 minHTR PM Technology (Pebble Bed Reactor)Nuclear Reactor Physics · 10k+ views

10 minChina’s MELTDOWN-PROOF Nuclear ReactorKyle Hill · 500k+ views

3 minTRISO FuelEPRI · 10k+ views

TRISO Particles: The Most Robust Nuclear Fuel on Earth (US Department of Energy) · Xe-100 Reactor Design (X-energy)

Prismatic HTGRs use the same TRISO fuel and helium coolant as the pebble bed, but the particles are pressed into compacts and stacked in hexagonal graphite blocks, so the core is fixed and machined rather than flowing. You give up the pebble bed's on-line refueling and get precise core geometry, higher achievable outlet temperatures (Japan's HTTR has run 950 °C, hot enough for thermochemical hydrogen), and none of the pebble-flow and dust issues.

These are the highest demonstrated outlet temperatures in fission, the core mechanics are well characterized, and you get the same walk-away TRISO safety case. Weaknesses: batch refueling requires complex block shuffling; graphite blocks cost money to machine and replace; power density and big-vessel economics are as bad as in any HTGR; and the commercial bench is thinner than the pebble-bed camp's. The design's fifty-year history (Peach Bottom, Fort St. Vrain's troubled run, GA's canceled GT-MHR) is full of prototypes and short on fleets.

Reach for a prismatic HTGR only when your requirement genuinely exceeds the pebble bed's range: 900–950 °C for sulfur-iodine hydrogen or the hottest process heat. HTTR is the only place in fission where that's been demonstrated. If you're investing, two things have to be true: there's a real market at 950 °C that 750 °C can't serve, and the balance sheet survives years of fuel qualification before revenue (that's the failure mode USNC ran into). For anything at 750 °C and below, the pebble-bed camp has the operating plant, the order book, and the momentum.

Thermal output 15–600 MWt across built and proposed units, 5–330 MWe · helium coolant with a graphite moderator · outlet 750–950 °C, the highest demonstrated in fission (HTTR held 950 °C for 50 days in 2010) · TRISO compacts stacked in hexagonal graphite blocks, HALEU enrichment · discharge burnup roughly 100 GWd/t · batch refueling on a roughly 18-month cycle, with block shuffling.

Japan's HTTR (operating, 950 °C demonstrations, coupled hydrogen-production testing); Fort St. Vrain (US, 1979–89, plagued by water ingress); Ultra Safe Nuclear's MMR (Canadian licensing at Chalk River before the company's 2024 bankruptcy and asset sale); General Atomics' prismatic lineage.

Commercially it trails the pebble bed: same heat markets, less momentum. Its one distinctive asset is the 950 °C capability, the only demonstrated fission route to sulfur-iodine hydrogen and the hottest industrial heat. USNC's bankruptcy is the warning for the sector, since TRISO-based microreactor ventures burn capital on fuel qualification long before they see revenue.

15 minHTGR The High Temperature, Gas-Cooled Power Reactor SystemWhat is Nuclear? · 5k+ views

2 minHow Do High-Temperature Gas Reactors Work?U.S. Department of Energy · 5k+ views

3 minLoss of Forced Coolant test using the High-Temperature Engineering Test Reactor (HTTR)OECD Nuclear Energy Agency · under 1k views

HTGR Research and Development — HTTR (JAEA) · Very-High-Temperature Reactor (Generation IV International Forum)

The SFR is the most-built advanced reactor in history. It uses liquid sodium coolant, no moderator, and a fast neutron spectrum that can breed more fuel than it burns or destroy the actinides that make waste long-lived. Sodium transfers heat very well, so cores can be dense, run at atmospheric pressure, and put out 550 °C. Twenty-odd SFRs have operated since 1951, so feasibility was never the problem. What held the technology back was sodium's chemistry, plus breeder economics that cheap uranium made pointless.

There's a lot of operating experience here (EBR-II's well-known passive-safety tests, Russia's BN-600/800 running commercially for decades), plus atmospheric pressure, waste transmutation potential, and metal-fuel passive shutdown physics. Weaknesses: sodium burns on contact with air or water, and leaks and fires dogged Monju and Superphénix; the coolant is opaque, which complicates inspection; an intermediate loop adds cost; and breeding start-up cores need plutonium or HALEU in quantities that strain supply.

If you're a utility, the near-term case is Natrium's specific configuration: a renewable-heavy grid that will pay for storage-backed load-following, in a jurisdiction willing to license sodium. Don't pick it for cheap baseload, where LWRs win. If you're running a state program, choose the SFR when you want fuel-cycle infrastructure (breeding, actinide burning, plutonium management) and the treasury can hold a multi-decade position, as Russia, China, and India do. Either way, treat a secured HALEU or plutonium startup inventory as the gating item, because it's the binding constraint on every Western schedule.

Electric output 300–800 MWe per unit, 345 MWe base and 500 MWe peaking on Natrium · liquid sodium coolant at atmospheric pressure, no moderator · outlet 500–550 °C · MOX or metal fuel, needing HALEU up to 19.75% or plutonium · design burnup roughly 100–150 GWd/t, several times what a light-water reactor reaches · breeding ratio around 1.0–1.2 · about 20 units have operated worldwide since 1951.

Russia's BN-800 (operating on MOX, the only commercial fast fleet); TerraPower's Natrium (345 MWe, metal fuel, plus a molten-salt heat store for 500 MWe peaking) — construction underway at Kemmerer, Wyoming, targeting ~2030; India's PFBR at Kalpakkam approaching operation; historic Phénix, Superphénix, Monju, EBR-II, FFTF.

Natrium is the West's flagship advanced build. Its salt-store peaking design is a genuinely clever answer to renewable-heavy grids, and its HALEU dependence (waived initially via DOE stockpiles) is the weak point in the schedule. Russia and China (CFR-600) treat SFRs as strategic fuel-cycle infrastructure rather than merchant plants, which is a reminder that fast-reactor economics have always been a state's long game rather than a market's.

14 minThe Genius Behind Bill Gates' New Natrium ReactorDr Ben Miles · 500k+ views

20 minSmall Nuclear Reactors - NatriumIllinois EnergyProf · 100k+ views

3 minThe Natrium™ Reactor and Energy Storage SystemTerraPower, LLC · 50k+ views

Fast Neutron Reactors (World Nuclear Association) · Sodium Fast Reactor (Generation IV International Forum)

The LFR gives you fast-spectrum physics with a coolant that doesn't burn. Molten lead (or lead-bismuth eutectic) is chemically inert with air and water, boils at 1,749 °C, and shields gamma radiation well enough that the vessel needs less biological shielding. The Soviet Alfa-class submarines ran lead-bismuth reactors in the 1970s, the only LFRs ever operated. They proved the concept and also documented its problems: coolant freezing at 327 °C, polonium-210 activation, and corrosion of the steel.

There's no sodium-style fire risk, no intermediate loop is needed, natural circulation is strong enough for passive safety, and the temperature potential is high. Weaknesses: lead corrodes and erodes structural steels unless oxygen chemistry is held in a very narrow window; the coolant freezing solid can destroy the plant; pumps and materials run in territory with almost no civilian operating data; and the submarine heritage is the only heritage there is, which is as much a warning as a validation.

Right now this is an investment thesis rather than something you can buy. Back an LFR if you believe corrosion control in lead is an oxygen-chemistry engineering problem that BREST-OD-300 and loop programs will solve, and you want fast-spectrum benefits without paying for sodium-fire mitigation in the plant. A newcleo-style venture is fundamentally a European plutonium-disposition play, so diligence the fuel-cycle politics as hard as the metallurgy. If you need a fast reactor on a committed schedule, go with sodium instead, since it has the operating fleet and the reactor-years.

Electric output 200–300 MWe on current designs, 700 MWt at BREST-OD-300 · molten lead or lead-bismuth coolant at atmospheric pressure, no moderator · lead melts at 327 °C and boils at 1,749 °C · outlet around 535 °C · mixed nitride or MOX fuel carrying plutonium · the only operating history is the Soviet Alfa-class submarine reactors of the 1970s and 1980s.

Russia's BREST-OD-300 under construction at Seversk (the world's first purpose-built civilian LFR, tied to an on-site closed fuel-cycle complex); newcleo (UK/France/Italy, lead-cooled mini-reactors with MOX ambitions); Westinghouse's LFR concept; Belgium's MYRRHA accelerator-driven lead-bismuth research machine.

The pitch is SFR benefits without paying for sodium fire protection, in exchange for materials risk that only reactor-years can settle. BREST's completion will be the field's proof point. Private ventures (newcleo has raised close to a billion) are effectively selling European plutonium-disposition policy plus lead physics, so they're fuel-cycle businesses more than reactor businesses.

7 minThe LeadCold RevolutionJanne Wallenius · 5k+ views

1 minIntroducing newcleo’s LFR-TL-40 reactornewcleo · 10k+ views

Lead Fast Reactors (Generation IV International Forum) · LFR System Safety Assessment 2020 (GIF Risk and Safety Working Group)

The GFR is the hardest entry on the Gen-IV roadmap. It combines helium coolant (transparent, inert, no activation) with a fast spectrum and an 850 °C outlet, which on paper gives you HTGR temperatures and breeder fuel economy at the same time. The hard part is decay heat. A fast core has high power density and helium has almost no thermal inertia, so losing pressure means losing cooling within seconds. Nobody has solved depressurized decay-heat removal without water or sodium's natural circulation, and design iterations have been trying since the 1960s.

The coolant is chemically inert and could drive a direct-cycle turbine, you get the full fast-spectrum fuel-cycle benefits, and there are no sodium fires or lead corrosion to manage. Weaknesses: the decay-heat problem requires either massive passive injection systems or fuel that survives 1,600 °C+ uncooled, which means carbide fuels that barely exist; no GFR has ever been built anywhere; and every practical compromise pushes the design back toward the thermal HTGR it was supposed to beat.

Don't plan deployments or venture bets here. Nothing has ever been built, and depressurized decay-heat removal has beaten sixty years of design iterations. The GFR belongs in a program portfolio only as low-cost research optionality with a defined revisit trigger, and the signal to watch for would be refractory carbide fuels maturing in space-reactor work. If you want fast-spectrum fuel economics, use sodium. If you want high-temperature helium, use a thermal HTGR.

Reference design 2,400 MWt and roughly 1,200 MWe, with ALLEGRO's 75 MWt demonstrator as the only concrete proposal · helium coolant at about 70 bar, no moderator · outlet 850 °C · refractory carbide fuel in SiC cladding that has to survive above 1,600 °C uncooled · no unit has ever been built.

Purely programmatic: Euratom's ALLEGRO demonstrator concept (Czech/Slovak/Hungarian consortium, perpetually pre-decision), historic GA and UK GCFR studies, Gen-IV Forum working groups.

There's no credible commercial vehicle. The GFR survives as a research placeholder for "what if HTGR safety met breeder economics." Its value is as a boundary case: it shows that coolant thermal inertia, rather than design elegance, decides which advanced reactors are buildable. Only watch this space if refractory carbide fuels mature somewhere else (space reactors may qualify them as a side effect).

Gas-Cooled Fast Reactor (Generation IV International Forum)

In a thermal MSR the fuel and the coolant are the same liquid: uranium (or thorium) fluorides dissolved in molten salt, circulating through a graphite-moderated core at ~700 °C and atmospheric pressure. Oak Ridge's MSRE proved the concept over four years in the 1960s. Liquid fuel buys you several things at once: no fuel fabrication, on-line refueling and cleanup, a strong negative temperature coefficient, and freeze-plug drain tanks that passively dump the core to safety.

You get atmospheric pressure, walk-away drain-tank safety, 700 °C heat, fuel that can't melt down because it's already molten, and unusual fuel-cycle flexibility including thorium. Weaknesses: the salt is very corrosive and radioactive everywhere it flows, so every pump, valve, and heat exchanger becomes a licensed nuclear component; tritium production (from lithium salts) permeates hot metal; off-gas systems have to handle mobile fission products continuously; and graphite lifetime limits core life.

Back a thermal MSR when the thesis is a much cheaper plant (no pressure vessel, no fuel fabrication, 700 °C heat) and the venture has a credible answer to salt chemistry as an operations problem. Today that means an IMSR-style sealed replaceable core, pursued in a jurisdiction like Canada that will engage with novel designs. If you're planning a program, treat China's TMSR line as the pacing demonstration and calibrate your timelines to it. If you need salt-heat revenue this decade, the FHR's solid-fuel compromise is the lower-risk route; liquid fuel pays off more but takes longer.

Thermal output 2–440 MWt across operating and licensed designs, up to about 195 MWe · fuel and coolant are the same liquid, uranium or thorium fluorides in molten salt, with a graphite moderator · outlet roughly 700 °C at atmospheric pressure · FLiBe carrier salt melts near 460 °C, so the loop has to stay hot · LEU to HALEU at up to 19.75% enrichment, plus thorium blends · on-line refueling and cleanup, with the IMSR core sealed and swapped about every 7 years · MSRE ran about 13,000 hours critical between 1965 and 1969.

MSRE (Oak Ridge, 1965–69 — the founding demonstration); China's TMSR-LF1 at Wuwei (2 MWt, achieved criticality 2023 and thorium-in-salt operation — the world's only operating MSR); Terrestrial Energy's IMSR (LEU, sealed replaceable core units, deep into Canadian licensing); Denmark's Copenhagen Atomics and Seaborg pursuing mass-manufactured salt units.

The promise is a cheap plant (no pressure vessel, no fuel fabrication, high-grade heat). The cost is a licensing and materials problem that has taken every MSR venture roughly a decade. China's state program doesn't face Western regulatory economics, and it's now the pacing demonstration. Western ventures live or die on whether "sealed core, swap every 7 years" (IMSR) convinces regulators to treat salt chemistry as an operations problem rather than a design-basis one.

14 minMolten Salt Nuclear Reactors – Visually ExplainedHiggsino physics · 50k+ views

21 minThe Molten-Salt Reactor ExperimentOak Ridge National Laboratory · 1m+ views

13 minAre Molten Salt Reactors the Future of Nuclear Power?Curious Droid · 100k+ views

Molten Salt Reactors (World Nuclear Association) · Molten Salt Reactor Program (Oak Ridge National Laboratory)

These reactors run molten salt without a moderator. Chloride salts (rather than fluoride) dissolve enough heavy metal to sustain a fast spectrum, so you get liquid fuel's safety and chemistry benefits together with a fast reactor's ability to breed and to burn actinide waste. There's no graphite, so there's no core-life limit, and chloride salts run even hotter than fluorides. In theory this is the endpoint of the fuel cycle, a reactor that could eat spent LWR fuel indefinitely. Almost none of it has ever been built.

The concept combines drain-tank passive safety with waste transmutation, needs no fuel fabrication, and needs no graphite replacement. Salt expansion gives a large negative feedback. The weaknesses: chloride salts require isotopically enriched chlorine-37 to avoid neutron poisoning, which means standing up a brand-new supply chain. Corrosion data is thinner than for fluorides, fast-spectrum liquid cores raise novel criticality-safety questions, and startup inventories of fissile material are large (HALEU or plutonium in tonnes rather than kilograms).

This is the longest-dated bet in the salt family. Back it only if you have decade-plus patience and believe a credible waste-burning reactor changes nuclear politics enough to be worth owning early. Gate any commitment on MCRE's results and on three specifics: chlorine-37 enrichment supply, tonnes-scale fissile startup inventory, and chloride corrosion data. The reactor physics is not the risk. If you want molten-salt heat sooner, look at the FHR or thermal MSRs instead. If you're a state that wants actinide burning sooner, look at sodium fast reactors.

Fuel and coolant are the same chloride salt, typically sodium chloride carrying uranium or plutonium chlorides · no moderator, fast spectrum · outlet temperature roughly 600–750 °C at atmospheric pressure · chlorine enriched above 99% Cl-37, against 24% natural abundance · fissile startup inventory in tonnes of HALEU or plutonium rather than kilograms · online refueling, so no cladding burnup limit · conversion ratio above 1 in breeder configurations.

TerraPower's Molten Chloride Fast Reactor program with its MCRE experiment planned at Idaho National Lab (with Southern Company) — the first fast-spectrum liquid-fuel criticality since the 1960s; Moltex's Stable Salt Reactor (UK/Canada — salt fuel in static pins, a deliberate simplification); Elysium and TerraPower's earlier chloride studies.

This is the furthest from revenue of the salt family, and it carries the biggest strategic prize: a credible waste-burning reactor changes the politics of nuclear everywhere. MCRE's results this decade will determine whether the concept graduates from viewgraph to engineering. Put your diligence weight on chlorine-37 supply, fissile startup inventory, and salt chemistry control. The reactor physics is sound.

8 minElysium Molten Chloride Fast Reactor (MCSFR): LFTR Alternative | Ep. 8Rock Logic With Sean Kenny · 5k+ views

2 minRevolutionising Nuclear: Meet the Molten Chloride Fast Reactor (MCFR)CORE POWER · 1k+ views

1 minSouthern Company Services and TerraPower Build World’s Largest Chloride Salt SystemU.S. Department of Energy · 1k+ views

Molten Chloride Fast Reactor Technology (TerraPower) · First Fuel Salt Batch for the MCRE (US Department of Energy)

Thorium is a fuel strategy rather than a reactor design. Th-232 is fertile, breeding fissile U-233 under neutron irradiation. It is 3–4× more abundant than uranium, it is the only fertile material that breeds in a thermal spectrum rather than a fast one, and its waste chain is shorter-lived. The canonical embodiment is the Liquid Fluoride Thorium Reactor (LFTR), a two-fluid molten-salt breeder that continuously separates protactinium. India holds vast thorium and little uranium, and has pursued a three-stage program toward thorium for seventy years.

The fuel is abundant, breeding works in a thermal spectrum, and transuranic waste is reduced. U-233 also comes contaminated with U-232, whose decay chain emits strong gammas, and that makes weapons diversion genuinely harder. Weaknesses: thorium needs a fissile starter (enriched uranium or plutonium), so it solves no startup problem. The same U-232 gammas that deter proliferation also make fuel handling difficult. Continuous protactinium separation is unbuilt chemical engineering. And the economic case is weak while uranium stays cheap, which has been thorium's perennial problem since the 1960s.

Commit to thorium only when your resource endowment forces it. India's uranium-poor, thorium-rich position is the one honest national case. Otherwise treat it as a far-horizon option layered onto an already-funded molten-salt or reprocessing program. It is never the reason to start a program: every thorium route still needs fissile startup material and unbuilt separation chemistry, and the economics don't close while uranium stays cheap. If you're an investor, the only meaningful signals to track are China's TMSR thorium operation and India's AHWR commitment. For everyone else, the default is the conventional LEU cycle.

Th-232 is 3–4× more abundant in the crust than uranium · the fertile chain runs Th-232 to Pa-233 (27-day half-life) to U-233 · thermal-spectrum breeding ratios come out near 1.05, so the margin is thin · U-233 carries parts-per-million U-232, whose Tl-208 daughter emits a 2.6 MeV gamma · LFTR designs target roughly 700 °C outlet at atmospheric pressure · China's TMSR-LF1, the only operating thorium-in-salt reactor, is 2 MWt.

India's program: KAMINI (U-233 fueled research reactor), thorium blankets in its PHWR fleet, the planned Advanced Heavy Water Reactor; China's TMSR-LF1 running thorium in salt (2023 — the first molten-salt thorium operation since Oak Ridge); historic Shippingport light-water thorium breeding demonstration (1977–82, quietly successful).

Thorium's advantages are real but modest, and they are chronically oversold. It is a second-order improvement that gets marketed as a revolution. No thorium route avoids first mastering either molten salt technology or reprocessing chemistry, and each of those is a decade-scale program. The investable signal to watch is China's TMSR scale-up and Indian AHWR commitment, not the level of enthusiasm online.

18 minThorium Nuclear Reactors Visualized & ExplainedHiggsino physics · 50k+ views

6 minThorium Fuel Cycle IntroductionFlibe Energy · 50k+ views

17 minWhat You Need to Know: Thorium Nuclear PowerIllinois EnergyProf · 100k+ views

Thorium (World Nuclear Association)

This is the pragmatist's molten-salt reactor. The fuel stays solid (TRISO pebbles, already qualified) and clean fluoride salt (FLiBe) is used only as coolant. That sidesteps liquid-fuel licensing entirely, because the salt is hot and low-pressure but barely radioactive, and you still get 650 °C output and pebble-bed walk-away safety. Kairos Power built its company culture around iterative hardware: non-nuclear salt test loops, then the Hermes demo, then power reactors, with each unit treated as deliberately disposable learning.

It runs at atmospheric pressure, uses a qualified fuel form, has a much simpler source term than liquid-fuel MSRs, and delivers high-grade heat. Kairos also has the sector's most disciplined build-test-iterate execution. Weaknesses: FLiBe means beryllium toxicity and lithium-7 enrichment supply, which is another boutique supply chain. TRISO and salt costs stack. Pebble handling in salt is novel. And the design intentionally forfeits liquid fuel's fuel-cycle advantages, because it is a heat plant rather than a breeder.

Choose the FHR path if you want advanced-reactor heat on the nearest credible timeline and you weight execution evidence over fuel-cycle ambition. Permits granted, concrete poured, and hyperscaler offtake signed make it the de-risked entry in the salt family. It fits data-center campuses and TVA-style utilities that want 600 °C-class output without liquid-fuel licensing novelty. Skip it if the prize you're after is breeding or waste burning, which this design deliberately forfeits; that's sodium or chloride-salt territory. Check lithium-7, beryllium, and HALEU supply before underwriting fleet scale-up.

Hermes demo is 35 MWt, with commercial KP-FHR units targeting roughly 140 MWe · FLiBe coolant at near-atmospheric pressure, 650 °C outlet · solid TRISO pebbles carrying HALEU just under 20% enrichment · lithium-7 enrichment above 99.99% to hold down neutron absorption and tritium · the Google offtake covers 500 MW of fleet output · overnight capex projected in the $3–7k/kW band.

Kairos Power: Hermes (35 MWt) under construction at Oak Ridge — the first non-water US power-reactor construction permit in 50 years — with Hermes 2 approved, a Google power-purchase agreement for 500 MW of fleet output, and TVA partnership; ORNL's decades of FLiBe research underneath.

On execution evidence this is the most credible near-term advanced-reactor program in the US: permits in hand, concrete poured, hyperscaler offtake signed. The bet is that manufacturing iteration (Kairos casts its own vessels) plus a boring licensing story beats more ambitious physics. Lithium-7 and beryllium supply plus HALEU are the scaling dependencies to watch.

5 minUnderstanding the Hermes Low Powered Demonstration Reactor (Oct 2022)Kairos Power · 5k+ views

4 minBuilding the Engineering Test UnitKairos Power · 1k+ views

3 minDelivering a Molten Salt MilestoneKairos Power · 1k+ views

KP-FHR Technology (Kairos Power) · Kairos Power Starts Construction of Hermes Reactor (US Department of Energy)

Microreactors treat nuclear as a product rather than a project: 100 kWe–20 MWe units designed for factory assembly, truck transport, and unattended or minimally-staffed operation. Most use TRISO fuel and either heat pipes (solid-state cooling with no pumps, as in Westinghouse eVinci) or compact gas cycles. The target customers aren't buying $/MWh. They're buying power where the grid isn't: remote mines, military bases, disaster response, Arctic communities, and increasingly data centers that want dedicated on-site generation.

Factory series production is finally plausible at this size, the passive safety case is small enough to argue for tiny emergency-planning zones, and deployment takes weeks. Weaknesses: cost per kW is the worst in the industry ($15–30k+/kW), so the pitch only works against diesel logistics at $0.30–1/kWh. Security and staffing rules built for gigawatt plants have to be rewritten for boxes in the field. HALEU supply gates everyone. And the customer set is real but thin; USNC's bankruptcy showed how fast micro-nuclear burns cash.

Buy a microreactor only where the true competitor is diesel logistics: remote mines, Arctic communities, forward bases, and island grids paying $0.30–1/kWh landed. In those places a sealed multi-year core clears its $15–30k/kW price. Defense procurement is the anchor customer, so treat data-center colocation as upside rather than base case. Don't buy one on grid-scale economics, because a 300 MW-class SMR beats it by an order of magnitude on $/kW. And diligence vendor runway hard, since fuel qualification burns cash for years before first revenue.

Output 100 kWe to 20 MWe, so a unit moves by truck or in standard containers · TRISO or metal fuel at HALEU enrichment just under 20% · sealed cores run 5–10 years, and several designs never refuel on site at all · overnight capex $15–30k/kW or higher, the worst in the industry · that price only clears where diesel lands at $0.30–1/kWh · site setup measured in weeks rather than years.

Westinghouse eVinci (heat-pipe, 5 MWe, test unit at INL planned), Oklo's Aurora (liquid-metal, INL site, NRC application famously denied once and refiled; Sam Altman-backed IPO), Radiant Kaleidos (1 MWe, helium; first fueled test slated for the DOME facility at INL), BWXT's Project Pele (DoD transportable prototype), Last Energy's containerized PWR.

Unit economics look more like aerospace than utilities: high margin, low volume, logistics-driven willingness to pay. Defense demand (Pele, installation resilience) is the anchor tenant, and data-center colocation is the volume dream. The honest comparison in every deal is a diesel genset fleet plus fuel convoys. If a sealed 8-year core beats that, the price premium clears.

21 minModular Micro-Reactors – The Future of Nuclear Energy?Illinois EnergyProf · 100k+ views

5 minIntroducing Kaleidos, a Portable Microreactor from RadiantRadiant Nuclear · 100k+ views

2 minHow the Westinghouse eVinci™ Microreactor Can Help Remote Communities, Mining CampsWestinghouseNuclear · 10k+ views

What Is a Nuclear Microreactor? (US DOE Office of Nuclear Energy) · eVinci Microreactor (Westinghouse)

These are reactors for places solar can't reach: kilowatt-to-megawatt units for lunar bases, Mars missions, and deep space, plus nuclear thermal propulsion for fast transit. Modern designs are solid-state. Heat pipes passively wick heat from a compact core to Stirling engines or thermoelectrics, so there are no pumps or coolant loops to fail. NASA's Kilopower/KRUSTY test (2018) was the first new US space-reactor demonstration since 1965's SNAP-10A, the only US reactor ever flown.

These systems supply continuous power through lunar nights and dust storms, where solar dies. Heat pipes eliminate nearly every classic failure mode, and RTGs give the approach decades of deep-space heritage. Weaknesses: launch approval for fissile material is politically hard, mass budgets leave little room for shielding, and testing ground-based analogs of space conditions is expensive. Programs in this field (NASA's Fission Surface Power, DARPA/NASA's DRACO nuclear-thermal demo) also have a long history of cancellation before flight, with DRACO's 2025 wind-down the latest.

This is a mission-driven government market. Specify a fission surface-power or heat-pipe system when the mission profile rules out solar (lunar night, Mars dust, the outer planets) and RTG power levels of a few hundred watts fall short. If you're an investor, you get exposure through component, fuel, and shielding suppliers rather than mission primes, and the commercial payoff is the terrestrial microreactor crossover (eVinci came out of the Kilopower lineage). If you're planning a program, price cancellation as the dominant risk, because the field's history, DRACO included, is programs dying before flight.

Power class runs from 1 kWe (KRUSTY) to about 40 kWe (NASA's Fission Surface Power target), against a few hundred watts electric from an RTG · heat pipes with alkali-metal working fluid replace pumped loops entirely, so there are no moving parts in the cooling path · Stirling conversion runs roughly 25–30% efficient against 6–7% for thermoelectrics · KRUSTY ran 28 hours at full power in 2018 on a highly enriched uranium-molybdenum core · NASA's lunar specification is 40 kWe under 6 t launch mass with a 10-year surface life.

KRUSTY (1 kWe ground demo), Fission Surface Power program (40 kWe-class lunar unit, industry teams incl. Westinghouse and IX/Intuitive Machines), the Soviet TOPAZ and RORSAT flight heritage, DRACO nuclear-thermal (paused), China/Russia announced lunar-station reactor plans.

This is a government program market rather than a commercial one, but it is strategically load-bearing. Heat-pipe cores, refractory fuels, and compact shielding developed for space flow directly into terrestrial microreactors (eVinci is Kilopower's industrial cousin). For investors the exposure is via component and fuel suppliers rather than missions. For nations it's Artemis-era prestige infrastructure.

3 minKilopower: A Gateway to Abundant Power for ExplorationNASA Video · 50k+ views

13 minThe Problem of Power in Space. NASA's New Kilopower ReactorFraser Cain · 500k+ views

2 minHeat Pipe Basics and Demonstration on How a Heat Pipe WorksAdvanced Cooling Technologies Inc. · 100k+ views

KRUSTY Nuclear Ground Test Results and Lessons Learned (NASA NTRS) · Fission Surface Power Project (NASA Glenn)

By operational record this is the most successful nuclear program in history: compact PWRs driving submarines and carriers, with the US fleet logging 5,700+ reactor-years and zero radiological accidents. US cores use weapons-grade HEU in robust cermet fuel and reach 30+ year lifetimes with no refueling, so refueling never comes up over the life of the boat. France runs LEU boats (refueling every ~10 years), and AUKUS is exporting the US/UK HEU approach to Australia.

These reactors have extreme power density, shock tolerance, decades-long cores, silent natural-circulation operation at low power, and an institutional safety culture (Rickover's) that civilian nuclear still studies. Weaknesses: everything is classified, gold-plated, and cost-insensitive, so almost nothing transfers directly to civilian economics. HEU fuel is a proliferation problem that pushes future designs toward LEU at real performance cost. And the industrial base (two US yards, one UK, one French) is a chronic bottleneck.

There is only one procurement scenario here. You are a navy (or an icebreaker fleet) for which power density, endurance, and stealth justify defense-budget pricing, and no civilian buyer faces that menu. For civilians the value is evidence and lineage: proof that small PWRs can be serial-built to immaculate quality, the workforce behind Rolls-Royce's SMR credibility, and Russia's RITM-derived floating and land units, which are the one direct naval-to-civilian product line. Draw cost lessons carefully, because nothing in these programs was ever optimized for $/MWh.

The US fleet has logged 5,700+ reactor-years with zero radiological accidents · US cores run HEU above 90% enrichment in cermet fuel and last 30+ years, so the boat is never refueled · France runs LEU under 10% and refuels roughly every 10 years · thermal output is commonly quoted around 150 MWt for an attack submarine and several hundred MWt per unit on a carrier · pressurized-water design with natural circulation at low power for quiet running · the whole Western industrial base is two US yards, one UK, and one French.

US Virginia- and Columbia-class (life-of-ship cores), UK Astute/Dreadnought (Rolls-Royce PWR3), France's Suffren class (LEU K15 lineage), China's expanding fleet, Russia's icebreakers (the civilian cousin — RITM-200 reactors now spun into floating and land SMR offerings).

Naval reactors are militarily invaluable and commercially instructive. They prove small reactors can be serial-built with immaculate quality, though only at defense budgets. The civilian spillover is real but indirect: Rolls-Royce's SMR credibility, Russia's RITM-based SMRs, and the trained workforce. AUKUS-driven HEU expansion versus LEU conversion research (a US congressional perennial) is the fuel-policy tension to watch.

14 minSubmarine Nuclear Power | Engineering behind it Nuclear Reactor How it WorksAiTelly · 1m+ views

14 minHow Aircraft Carrier Works? US Nuclear Power Ship Nimitz Class #shipAiTelly · 1m+ views

3 minThe Naval Nuclear Propulsion Program: An OverviewNaval Nuclear Laboratory · 1k+ views

Powering the Navy — Naval Nuclear Propulsion Program (US DOE NNSA)

Tokamaks are fusion's front-runner by every metric. A torus of plasma is confined by strong toroidal magnets plus a current driven through the plasma itself, which holds 150-million-degree deuterium-tritium fuel away from every wall. Tokamaks hold all the confinement records (JET's 69 MJ shot, ITER's design goal of Q=10). Since REBCO high-temperature superconducting tape enabled 20-tesla magnets, the machines have shrunk from ITER's cathedral scale to warehouse scale, which is the insight behind Commonwealth Fusion Systems.

Tokamaks have fifty years of accumulated physics, the best confinement performance, and a mature international knowledge base. Weaknesses: the plasma current makes the machine prone to disruptions (sudden collapses that can damage it), and steady-state operation requires continuously driving that current. D-T operation means breeding tritium in lithium blankets, which has never been demonstrated at scale, and it means neutron-degraded first walls that have to be replaced robotically. The engineering gain (electricity out over electricity in, including magnets and systems) also remains far below the plasma physics gain in the headlines.

Back tokamaks if you want fusion exposure with the least plasma-physics risk and the most independent validation. This is where the confinement data, the supply chain, and the capital sit, and SPARC is the field's nearest scheduled Q>1 test. A national program buying one decadal magnetic-fusion effort should default here for the same reason. Diversify elsewhere if you judge the binding risks to be the engineering ones tokamaks carry structurally (disruptions, current drive, tritium breeding), which is the stellarator's case for the endgame plant. If you want higher-variance, higher-ceiling exposure, the FRC and pulsed camps are the complement.

Plasma runs at roughly 150 million °C · REBCO high-temperature superconductors reach 20 T, against ITER's 11.8 T peak field on niobium-tin · JET's record D-T shot released 69 MJ · ITER targets Q=10, meaning 500 MW of fusion power from 50 MW of heating, and SPARC targets Q>1 around 2026–27 · EAST has held H-mode plasma past 1,000 s · a self-sufficient plant needs a tritium breeding ratio above roughly 1.05, and no blanket has shown it at scale.

JET (retired 2024 after record D-T runs), KSTAR and EAST long-pulse records, CFS assembling SPARC in Devens with ~$2B+ raised, the flagship private fusion program.

Tokamaks attract most of fusion's capital. CFS alone holds a large share of all private fusion funding, plus a signed 200 MW Google offtake. The bet decomposes into three things that all have to work in sequence: SPARC's Q>1, tritium self-sufficiency, and an availability-and-maintenance story competitive with anything on the grid. Even bulls should model first commercial electricity no earlier than the early-to-mid 2030s and fleet relevance in the 2040s.

10 minConstruction and Working Principle of TokamaksEnergy Encyclopedia · 10k+ views

3 minWhat is a tokamak? And is a spherical tokamak different?Tokamak Energy · 100k+ views

5 minOur first completed magnet for SPARC could lift an aircraft carrierCommonwealth Fusion Systems · 100k+ views

The ITER Machine (ITER Organization) · Overview of the SPARC Tokamak (Journal of Plasma Physics, open access)

A stellarator is the tokamak's close relative. Instead of driving current through the plasma, it builds the confinement twist entirely into very complicated external magnets. With no plasma current there are no disruptions and operation is inherently steady-state, so the two problems that most threaten tokamak power plants don't exist here. The price was always that the 3D coil shapes were unbuildable. Germany's Wendelstein 7-X (designed by supercomputer, built over 15 years) proved they no longer are.

Stellarators are intrinsically steady-state, disruption-free, and quiescent, which is the most reactor-like operating mode in magnetic fusion. Weaknesses: confinement historically trailed tokamaks (W7-X is closing the gap). The coils are the hardest magnets ever manufactured, and getting maintenance access through the tangle of them is a serious plant-design problem. The field also sits a machine-generation behind tokamaks, and no stellarator has ever run D-T fuel.

Back stellarators on the thesis that the hard part of fusion is running a plant 8,000 hours a year rather than reaching first gain. Steady-state, disruption-free operation is what tokamaks have to engineer around, and stellarators get it from geometry. The bet requires two things to be true: that HTS magnets plus modern optimization close W7-X's confinement gap, and that coil manufacturing and maintenance access yield to the Type One/Proxima generation of designs. The value case is that entry prices run an order of magnitude below the tokamak camp's. If you need a near-term Q>1 catalyst, be in tokamaks instead.

W7-X runs 2.5 T on axis in a 30 m³ plasma, held by 50 non-planar and 20 planar superconducting coils · discharges have reached 8 minutes, with no plasma current and therefore no disruptions · the machine took about 15 years and roughly €1B to build · coil geometry has to hold millimeter tolerances, which is the manufacturing constraint · no stellarator has ever run D-T fuel, so there is no measured Q for the concept.

Wendelstein 7-X (Greifswald, record stellarator triple products and 8-minute plasmas, the field's flagship); Type One Energy (Infinity One design, TVA partnership at a retired coal site, the leading private effort); Proxima Fusion (Munich, W7-X heritage, HTS coils); Thea Energy (planar-coil simplification); Japan's LHD heritage.

The thesis is that tokamaks win the race to Q>1 and stellarators win the race to a power plant that runs all year. Modern optimization plus HTS magnets could leapfrog a machine generation. Private funding is an order of magnitude below the tokamak camp's, which makes stellarators the value bet in fusion portfolios. The odds on timing are long, but the odds that the physics choice is right for the endgame are strong.

8 minConstruction and Working Principles of a StellaratorEnergy Encyclopedia · 10k+ views

6 minWendelstein 7-X fusion device – visualized from construction dataMax-Planck-Institut für Plasmaphysik · 50k+ views

11 minWendelstein 7-X – from concept to realityMax-Planck-Institut für Plasmaphysik · 100k+ views

Wendelstein 7-X (Max Planck Institute for Plasma Physics) · Stellarator Research Opportunities (National Stellarator Coordinating Committee, via DOE OSTI)

ICF fuses fuel by implosion. Converging laser energy crushes a peppercorn-sized D-T capsule until it ignites, and it burns in nanoseconds before it flies apart, so the only thing confining it is its own inertia. The National Ignition Facility made history in December 2022 with the first laboratory ignition (3.15 MJ out, 2.05 MJ laser in), and has repeated it since with gains above 2. The catch is that NIF's wall-plug lasers consumed ~300 MJ per shot and fire a few times a day, while a power plant needs gain of ~50–100 at ten shots per second, forever.

This is the only approach with demonstrated ignition, it needs no magnets and no steady-state plasma to sustain, and its subsystems (laser, target, chamber) are separable enough to iterate independently. Weaknesses: the gap from scientific to engineering gain is a factor of ~100 in laser efficiency and repetition rate. Targets have to cost cents rather than the thousands of dollars of hand-built NIF capsules, at millions per year. The final optics stare directly at fusion neutrons. And NIF itself is a weapons-stewardship facility whose geometry (indirect drive) is knowingly inefficient for energy.

Back laser inertial fusion if you weight demonstrated ignition above all else and you believe diode-pumped laser efficiency and cents-per-unit target manufacturing can each improve by the required factor of ~100. The bet rides on those two cost curves rather than on plasma physics. The defense overlap is a genuine de-risker, since stockpile-stewardship revenue can carry ventures across the long gap to power. If you like the inertial physics but not the laser economics, pulsed-magnetic drivers (Pacific Fusion) run the same wager on capacitors at a tenth the driver cost. If you need power-plant engineering plausibility now, look at MTF or the tokamak camp.

NIF's 192 beams put 2.05 MJ on target for 3.15 MJ of fusion yield in December 2022, and later shots have passed gain 2 · wall-plug draw is roughly 300 MJ per shot at a few shots per day · a power plant needs gain of 50–100 at about 10 shots per second · the D-T capsule is about 2 mm across and burns in nanoseconds · targets cost thousands of dollars each today and have to reach cents at millions of units per year · diode-pumped lasers have to hit roughly 10% wall-plug efficiency, against under 1% for NIF's flashlamps.

NIF (LLNL, ignition achieved repeatedly), Xcimer Energy (Denver, excimer lasers at 10× NIF energy for cheap direct drive), Focused Energy (Texas/Darmstadt, proton fast ignition), Germany's Marvel Fusion, historic Laser Mégajoule (France) and Omega (Rochester).

Ignition transformed the field's credibility and fundraising overnight, but every link in the commercial chain (diode-pumped lasers at ~10% efficiency, mass-produced targets, chamber survivability) is its own venture-scale program. The near-term revenue is defense work. Inertial fusion expertise is stockpile-stewardship expertise, and several startups quietly monetize that overlap while the power plant slips into the 2040s.

5 minHow NIF WorksLawrence Livermore National Laboratory · 100k+ views

14 minInertial Confinement FusionEnergy Encyclopedia · 10k+ views

7 minNew Nuclear Fusion Record at NIF! Can laser fusion break even?Sabine Hossenfelder · 100k+ views

Achieving Fusion Ignition (LLNL National Ignition Facility) · Basic Research Needs for Inertial Fusion Energy — Summary (US DOE)

MTF sits between magnetic and inertial fusion. You form a magnetized plasma (typically a compact toroid), then mechanically crush it to fusion conditions with imploding liquid-metal liners, pistons, or pulsed fields. The magnetization slows heat loss during compression, so the implosion can be a thousand times slower and cheaper than laser ICF: steam-driven pistons instead of megajoule lasers. General Fusion's design compresses plasma inside a rotating cavity of liquid lithium-lead, which serves as neutron blanket, tritium breeder, and heat-capture medium all at once.

Plant integration is unusually clean, because the liquid-metal liner solves first-wall damage, tritium breeding, and heat extraction simultaneously, and the compression hardware is industrial machinery rather than exotic optics. Weaknesses: the target plasma has to survive compression without instabilities destroying it, and that core physics risk is still unproven at scale. Repetition-rate engineering (a new plasma every second, in liquid metal) is barely explored. And the field is effectively a handful of companies, so a single company's stumble reads as a verdict on the whole category.

Back MTF if your read is that engineering, rather than physics, is what kills fusion projects. Here the plant-killer problems (first-wall damage, tritium breeding, heat capture) are all solved at once by the liquid-metal liner, which leaves a single concentrated physics question: whether the target plasma survives compression. That inversion suits investors who prefer one binary technical risk over a long chain of engineering problems that all have to go right. Size the position for a category that is one or two balance sheets deep. If you want diversified, physics-validated fusion exposure, you belong in tokamaks.

Compression runs roughly a thousand times slower than laser ICF, which is what lets pistons and pulsed fields replace megajoule lasers · repetition rate has to reach about one shot per second · a rotating liquid lithium-lead liner handles neutron shielding, tritium breeding, and heat capture in one part · the 14.1 MeV D-T neutrons deposit in that liquid rather than in a solid first wall · General Fusion's LM26 targets 100 million °C and scientific-breakeven-equivalent conditions, after 20+ years of development.

General Fusion (Vancouver, 20+ years, Bezos-backed; its LM26 machine demonstrated plasma compression milestones in 2025 amid deep funding difficulties), Helion arguably borrows MTF physics in pulsed form, historic LANL FRX-L and Soviet MAGO programs.

On paper this is the cheapest path to a plant that sells power rather than another experiment, because everything expensive about tokamaks and lasers is replaced by pistons and pumps. In practice the category's fortunes track one company's balance sheet. For diligence, price the plasma-compression physics as the whole risk, because the engineering story, uniquely in fusion, is already plausible.

1 minGeneral Fusion’s MTF: See How It WorksGeneral Fusion · 5k+ views

2 minGeneral Fusion: What is Magnetized Target Fusion?General Fusion · 1k+ views

2 minGeneral Fusion's LM26: Built and OperatingGeneral Fusion · 1k+ views

Magneto-Inertial Fusion (Journal of Fusion Energy, open access) · Magnetized Target Fusion with a Spherical Tokamak (Journal of Fusion Energy)

An FRC is a self-organized smoke-ring of plasma that confines itself with its own internal currents inside a simple cylindrical chamber, with no toroidal field coils threading the plasma at all. The geometry gives extreme plasma pressure relative to magnetic field (high beta), which makes it the natural home for advanced fuels. Helion accelerates two FRCs to collision at a million mph, compresses the merged plasma, and then takes electricity straight out of the expanding plasma as it pushes back on the magnetic field, skipping steam entirely (no other fusion approach does this). TAE aims further still, at aneutronic proton-boron fuel.

The geometry is simple and linear, Helion's direct electricity recovery tolerates lower gain, and the fuel cycles (D-He3, p-B11) produce few or no neutrons, which sidesteps tritium logistics and wall damage. Weaknesses: FRC stability and confinement at reactor scale remain the great unknown, since the configuration historically decays in milliseconds. Advanced fuels demand plasma temperatures 10–50× more demanding than D-T. Helion's D-He3 requires breeding helium-3 from D-D reactions in-machine. And the field's boldest commercial promises sit furthest ahead of published physics.

Back the FRC camp for the highest-ceiling outcome in fusion. Direct electricity conversion and low-neutron fuels would obsolete the steam-cycle plant everyone else is designing, provided the self-organized plasma holds together at reactor scale, which remains unproven. Underwrite on published triple products and on dated, penalty-backed commitments like Helion's Microsoft PPA rather than on announcements; the gap between the two is precisely the risk being priced. This is venture-shaped exposure where commercial promises lead the physics. If you want validation-first fusion, you belong in tokamaks. If you want cheap-hardware optionality, look at the pulsed camp.

Plasma beta runs 0.8–0.9, against a few percent in a tokamak, which is what makes advanced fuels thinkable · Helion accelerates two FRCs to collision at about a million mph · D-T ignites near 100 million °C, D-He3 near 1 billion °C, and p-B11 near 3 billion °C · Helion claims up to 95% electrical recovery direct from the expanding plasma, against 33–40% for a steam cycle · the Microsoft power agreement is 50 MW starting 2028 · TAE has raised about $1.2B.

Helion (Everett, WA): Polaris machine targeting first fusion electricity, a signed 2028 Microsoft power agreement (fusion's first PPA) and Sam Altman as chairman; TAE Technologies (California, ~$1.2B raised, p-B11 roadmap via its Copernicus machine); University of Washington and PPPL FRC research lineage.

This is the venture-capital shape of fusion: the highest ceiling (direct conversion, clean fuels, small machines), the widest error bars, and commercial commitments that lead the physics rather than trail it. The Microsoft PPA converts credibility risk into a dated, penalty-backed test. Track published triple-product data rather than announcements; the gap between the two is the investment.

3 minHelion's approach to fusion: How it worksHelion · 100k+ views

15 minAn Exotic Approach to Nuclear Fusion (Helion Energy BTS)Plasma Channel · 500k+ views

3 minIntroducing ‘Norm’: A First-of-its-Kind Fusion MachineTAE Technologies · 10k+ views

Field Reversed Configurations — the classic review (Nuclear Fusion) · Overview of C-2W Beam-Driven FRC Plasmas (Nuclear Fusion)

This is the oldest fusion idea, now rehabilitated. You run an enormous current through plasma and let its own magnetic field pinch it to fusion density, with no external confinement magnets at all. Classic Z-pinches died of instability in the 1950s, and the modern revival adds stabilizing physics. Zap Energy's sheared axial flow keeps the column intact, while Pacific Fusion and Sandia's MagLIF compress fuel-filled liners with precisely-timed megaampere pulses. That is inertial fusion driven by capacitors instead of lasers, at perhaps a tenth the driver cost.

The hardware is very simple and cheap: capacitor banks and electrodes rather than superconductors or megajoule optics, and small unit sizes are conceivable. Weaknesses: instability suppression has to hold at fusion gain, not only at breakeven-adjacent conditions. The electrodes sit very close to the plasma and have to survive millions of pulses. Repetition rate and liner/target replacement carry the same rep-rate problem as laser ICF. And the approach's history is full of pinches that worked right up until they scaled.

Back pulsed-magnetic approaches if you want fusion's cheapest iteration loop. Capacitor banks and electrodes ride commodity power-electronics cost curves, so each experiment (and each eventual unit) costs a fraction of superconducting magnets or megajoule optics. The risk profile is the tokamak's inverse: nearly all of it is whether instability suppression and electrode survival hold at gain conditions, and almost none of it is capital cost. That makes this the natural high-variance satellite position in a fusion portfolio. If you need demonstrated confinement scaling or a near-term gain catalyst, hold tokamaks or laser ICF instead.

Sandia's Z machine delivers about 26 MA in roughly 100 ns, the largest pulsed-power drive anywhere · MagLIF preheats the fuel with a laser inside a 10 T axial field, then crushes a centimeter-scale liner with about 20 MA · Zap's sheared-flow Z-pinch runs at a few hundred kA in a column tens of centimeters long, with no external confinement coils · driver hardware costs roughly a tenth of an equivalent laser · electrodes sit close to the plasma and would have to survive millions of pulses in a plant.

Zap Energy (Seattle, sheared-flow Z-pinch, FuZE-Q targeting scientific breakeven-relevant conditions), Pacific Fusion ($900M raised at founding, pulser-driven inertial fusion with NIF-ignition alumni), Sandia's MagLIF program and Z machine (the world's largest pulsed-power facility), UK's First Light Fusion (projectile-driven inertial, a related approach; pivoted to supplying targets and amplifiers).

This is the capital-efficiency wing of fusion. If confinement physics cooperates, the driver hardware is commodity power electronics riding its own cost curve. Pacific Fusion's founding raise signals investor appetite for "NIF physics with affordable drivers." The risk concentration is inverted from tokamaks: the machine is cheap and the physics uncertainty is expensive, which makes this the highest-variance, lowest-capex corner of the field.

1 minHow It Works: Sheared-Flow StabilizationZap Energy · 5k+ views

4 minJust A Little Z-Pinch: Zap Energy's Game-Changing Approach to Fusion PowerARPA-E · 10k+ views

8 minFusion: Pulsed Power Research at Sandia National LaboratoriesSandia National Labs · 50k+ views

Z-Pinch Fusion — review by U. Shumlak (Journal of Applied Physics, via DOE OSTI) · Z Pulsed Power Facility (Sandia National Laboratories)