Advanced Fission

Advanced fission means reactor concepts beyond the light-water design that produces almost all of the world's nuclear electricity: fast neutron spectra, molten salt, high-temperature gas, lead cooling, supercritical water, and the fuel cycles that go with them — thorium, breeding, and closing the cycle by reprocessing. The Generation IV International Forum recognises six coolant families, and this brief takes the same six as its unit of analysis, because arguing about advanced fission in the abstract obscures how differently the six have fared.

Where this brief stops. It is about reactor physics, coolants and fuel cycles. The modular delivery model — factory production, series learning, the economics of building small — is small modular reactors, and this brief does not re-argue it. Several designs are both small and advanced; the two properties are independent, and conflating them is how a coolant novelty gets credited to modularity. What happens to the fuel afterwards is Energy Systems slot 15, Nuclear Waste Solutions, named here in plain text because it is not yet built. Reprocessing appears in this brief only where it is part of a reactor's own fuel cycle — a fast reactor's claim to breed, a molten-salt reactor's claim to process its fuel online — which is unavoidable, because those claims are the entire case for the designs that make them.

One organising question does most of the work, and it is deliberately crude: has this coolant family ever made grid electricity? Sodium has and does. High-temperature gas has, historically, and one unit runs today. Molten salt never has, anywhere, not one kilowatt-hour. Lead has no civil operating precedent at all. Gas-cooled fast has never been built. Supercritical water has never been built and has no prototype planned. That distribution is invisible in coverage that treats “Generation IV” as a single maturing thing.

The second organising claim is about what binds. The physics of fast spectra, salt chemistry and high-temperature gas has been understood since the 1960s; nothing on the critical path is a discovery. What has never been produced is a code-qualified nuclear graphite, a code-qualified nickel alloy for fluoride-salt service, a TRISO performance envelope with closed acceptance criteria, or a fast reactor that closed its own fuel cycle at industrial scale — and the programmes that would produce them run on ten-to-twenty-five-year clocks that began before most of the companies now announcing 2030 deployment dates existed.

A note on method. Where two sources disagree about a date, an event count or a capacity rating, both figures appear below and the brief says it does not resolve them. Where a number could not be verified — an achieved breeding ratio, a lead-coolant temperature ceiling, a 2025 reprocessing throughput — the brief states the absence rather than supplying an estimate.

Established Take the six Generation IV coolant families and ask only whether each has ever made grid electricity. Sodium fast reactors have, and carry more than 400 reactor-years of experience across experimental, prototype, demonstration and commercial units in China, France, Germany, India, Japan, Russia, the United Kingdom and the United States. That figure belongs to the intergovernmental agency and it is stated for sodium-cooled fast reactors specifically; an industry association uses the same number more loosely for fast neutron reactors in general, and this brief cites the narrower attribution because it is the defensible one. High-temperature gas reactors have operated at power historically — Peach Bottom 1, Fort St Vrain, AVR, THTR-300 and Dragon, all now shut — with exactly one operating today. Molten salt reactors never have, anywhere, not one kilowatt-hour. No civil lead-cooled reactor has ever operated. The gas-cooled fast reactor has never been built at any scale, making it the only family with no operating precedent of any kind. Supercritical water has never been built either, and the forum's own portal says plainly that there are none operating commercially or under construction.

Established The family with 400 reactor-years has an outturn record that is almost never printed next to the number. Experience and performance are different quantities, and in sodium they diverge violently.

Established Four hundred reactor-years of learning produced one good plant. That is the single most important sentence in this brief. The programme cost is on the same order as the disappointment: OECD countries spent roughly $50 billion in 2007 dollars on fast breeder research and development between 1974 and 2007. France cancelled ASTRID in 2019; Japan committed Monju to decommissioning in 2016, its regulator having found the operator unfit for purpose in November 2015; South Africa halted PBMR funding in February 2010 and cut about three-quarters of a workforce of more than eight hundred.

Frontier Even the sodium incident record is contested, and this brief does not resolve it. The independent fissile-materials review reports twenty-seven sodium leaks at BN-600 between 1980 and 1997, fourteen of which caused sodium fires. A tertiary source citing the same agency reports twelve water-into-sodium steam-generator leaks in the plant's first twenty-four years, routinely handled by isolating the faulty module and not reported to the agency. The two counts are probably counting different populations — all sodium leaks against steam-generator tube leaks specifically — but nothing in the sources consulted establishes that, and the true event count is recorded here as unresolved.

Established Molten salt's two operating reactors are sixty years old and neither made electricity. The Aircraft Reactor Experiment ran from 3 to 12 November 1954 at a maximum sustained 2.5 MWt, achieving 221 hours of nuclear operation, the last seventy-four of them in the megawatt range, for 96 MWh of total nuclear energy; the fuel was dumped the day after shutdown and dismantling began in February 1955. The Molten-Salt Reactor Experiment reached about 7.4 MWt, went critical on uranium-235 on 1 June 1965 and became the first reactor in the world to run on uranium-233 on 2 October 1968, with plutonium trifluoride added as makeup fuel in its final months. Availability in the final reliability campaigns was 98.6% on uranium-235 and 99.9% on uranium-233, which is genuinely impressive and is the strongest thing in the salt record. Neither reactor had a power conversion system. Neither demonstrated a breeding blanket, online reprocessing at scale, or thorium.

Frontier Even the headline duration figure for the salt experiment is unreconciled across three sources. A national laboratory's training module gives 13,172 hours at full power, split 9,005 on uranium-235 and 4,167 on uranium-233; a tertiary source gives 15,425 critical hours, about 11,555 equivalent full-power hours and 19,405 fuel-pump circulating hours; an industry association says simply “10,000 operating hours.” The differences are almost certainly definitional and no Oak Ridge primary was reachable to reconcile them. Design power is likewise given as 10 or 15 MWt depending on the source; only the achieved 7.4 MWt is consistent.

Established The binding constraint across all of this is materials qualification, and it is measured in decades. There is no ASME code-qualified nuclear graphite. None. A 2022 peer-reviewed perspective states it flatly, and every high-temperature gas design and every graphite-moderated salt design — HTR-PM, Kairos Hermes, Terrestrial's IMSR, China's TMSR-LF1, X-energy's Xe-100 — depends on a moderator and structural material for which zero grades are qualified. Qualifying one under ASME Section III Division 5 requires irradiation-induced dimensional change, elastic modulus, strength, thermal conductivity, coefficient of thermal expansion and creep coefficient across multiple temperature and dose combinations. The Advanced Graphite Creep campaign that would supply them began in 2005, started irradiating in 2009, and is expected to finish in 2030 — twenty-five years, rate-limited by materials test reactor capacity. Frontier The same author benchmarks the cost against Alloy 617, which took twelve years and $15 million of public money and required no irradiation testing at all, and calls $15 million a lower bound for a graphite programme. That is an expert's floor estimate rather than a costed plan, and this brief carries it with that standing.

Established Hastelloy N is still not codified in ASME Section III Division 5, sixty years after the salt experiment it was invented for. Frontier The regulator's technical gap assessment goes further than absence: the original alloy is described as unsuitable, on grounds of severe irradiation embrittlement, strength falling away above roughly 700 °C, helium embrittlement that is lifetime-limiting above 650 °C, and a significant lack of quality irradiation data for many of the candidate metals. A national laboratory assessment in 2018 put completion of Alloy N code qualification at roughly ten years from then — 2028 at the earliest, on a 2018 baseline, and no source consulted for this brief confirms it has been met.

Established And the breeding promise, which is the whole argument for fast reactors, has never been demonstrated at industrial scale. No fast reactor anywhere has taken its own spent fuel, reprocessed it, and returned it as fresh fuel in a sustained industrial loop. The closest anyone has come is kilogram-to-tonne pilot campaigns, and Russia's BN-800 — the most advanced MOX-fuelled fast reactor running — uses plutonium extracted from thermal reactor spent fuel, with recycling of its own discharge described as an aim rather than an outturn.

Established The microreactor case is six different cases that fail for different reasons, and arguing it as one is how the weakest application gets carried by the strongest. The applications routinely bundled are remote community power, forward military bases, off-grid mining, district heat, hydrogen, and industrial process heat. Only the first three compete against diesel, and that is the entirety of their economic case: remote northern communities and fly-in mine sites buy electricity at several times the grid average because they buy fuel by barge and aircraft. Frontier Delivered-diesel costs reported for remote Alaskan and northern Canadian communities run from roughly $0.30 to above $1.00 per kilowatt-hour depending on haul distance and year — a band wide enough to contain almost any microreactor estimate, which is why quoting it proves nothing on its own. The honest statement is that three of the six have a competitor worth beating and three do not, because district heat, hydrogen and process heat are priced against gas at industrial rates.

Established The process-heat case has a temperature problem no coolant family fixes. Industrial heat demand is not distributed where the reactor literature implies: a large share sits below 400 °C, where steam from any reactor, a heat pump or a boiler all serve, and the genuinely hard-to-abate share sits above 1,000 °C — cement clinker near 1,450 °C, blast-furnace ironmaking above 1,200 °C — which no fission coolant reaches. Established HTR-PM, the one high-temperature gas unit in commercial operation, runs a reactor outlet near 750 °C and delivers steam at about 566 °C. The band a reactor can serve is the band with the most competitors and the least decarbonisation value, and the band the announcements point at is the one the materials do not reach.

Established The ceiling on that band is codified and datable. ASME Section III Division 5 admitted Alloy 617 for high-temperature nuclear service to about 950 °C in 2020 — the first new high-temperature structural alloy added to the nuclear code in decades, and the outcome of a qualification campaign of the same order of length as the graphite one described above. Frontier That single code case, rather than any reactor concept, sets the practical outlet ceiling for gas-cooled process heat, and no source consulted for this brief establishes a comparable qualification for the metallic intermediate heat exchanger that would carry 900 °C helium into a chemical plant. Handwave The brief states that absence rather than estimating a date for it.

Almost nothing in advanced fission is open in the sense that a physicist does not know the answer. Most of it is open in the sense that nobody has irradiated the samples, closed the acceptance criteria, or run the loop long enough to find out what fails.

Frontier TRISO fuel is routinely described as already qualified, and the regulators' own joint evaluation says otherwise. The interim report is explicit about what has been qualified: particle layer dimensions within confidence intervals, kernel enrichment below 20% uranium-235, a uranium carbide molar fraction of 30 ± 5%, packing fraction below 40%, an uninterrupted coating process, and irradiation data validated up to a simplified tensile stress metric of 0.810 at the 99th percentile. It is equally explicit about what is not. Four sections are marked “to be completed” — radionuclide release limits, test envelope validation, data measurement methodology and test conditions — and the irradiation campaigns did not establish a correlation between silicon-carbide microstructure and particle performance. The layer that does the containment job has no established microstructure-to-performance link. The report also flags kernel-coating mechanical interaction as possibly a bigger factor at higher burnup — the envelope may not extend to the burnups advanced designs want — and notes that silicon-carbide thermal decomposition above roughly 1600 °C needs further accident analysis. It is interim, and explicitly not subject to either regulator's management and legal review.

Established The operating high-temperature reactor's own fuel was qualified on five pebbles. HTR-PM's fuel elements were validated by irradiating five spherical elements in the High Flux Reactor at Petten in the Netherlands between September 2012 and 30 December 2014, with online fission-gas-release monitoring, and behaviour was reported as very good. Frontier The burnup achieved and the measured failure fractions were not stated in the source consulted, and whether Chinese production pebbles have been qualified beyond those five test elements could not be established. A February 2026 licence for a TRISO fabrication plant — the first Category II fuel fabrication facility ever approved in the United States — authorises a factory, not a performance envelope, and the distinction is routinely lost.

Established No achieved breeding ratio exists in the literature this brief could reach, and the absence is itself the finding. The most comprehensive critical review of fast breeder programmes contains no data on breeding ratios actually achieved, doubling times realised, or confirmed fuel-cycle closure by any operating fast reactor. Handwave Any specific achieved-breeding-ratio figure encountered without a measured provenance should be treated as unsupported. India's peer-reviewed programme paper states a ratio “as high as 1.5” with additional conversion in axial blankets — for metal fuel, which the prototype fast breeder reactor does not use. That is a design claim about a fuel type not in service, and this brief does not carry it as an outturn.

Frontier Fuel-cycle closure exists only in fragments, and every fragment has a documentation problem. Superphénix reprocessed about twenty-five tonnes of its own fuel with plutonium returned into fresh elements — reported by an industry association, with no independent confirmation found. India's CORAL pilot facility, operating since 2003, has reprocessed fast-reactor mixed-carbide fuel at burnup up to 155 GWd/t with short cooling and good decontamination factors, in a conference paper by the programme's own institute that states no quantities in kilograms, no campaign dates, and nothing about whether the material was refabricated and re-irradiated. India's peer-reviewed paper asserts closure without quantifying it. BREST-OD-300's on-site reprocessing module is not scheduled for commissioning until 2030, four years after the reactor's physical launch.

Established Thorium is fertile, not fissile, which is the sentence most thorium advocacy omits: every thorium concept needs a uranium-235 or plutonium driver to start. Two penalties follow. Uranium-232, half-life 68.9 years, is produced unavoidably by parasitic neutron reactions on thorium-232, protactinium-233 and uranium-233; its decay chain reaches thallium-208, which emits a 2.6 MeV gamma with 99.75% probability, requiring remote manipulation for fuel fabrication rather than the glovebox handling used for plutonium, except in a brief window immediately after chemical separation and before daughter ingrowth. And breeding requires extracting protactinium-233 from the neutron flux on a 27-day half-life clock, letting it decay outside the core, and returning the uranium-233. Established An industry association — a pro-nuclear interested party, which is what makes the admission strong — states that this separation process is unproven at any scale.

Frontier China's thorium salt reactor is the only live vehicle for any of this, and what it has achieved is a measurement, not a breeding demonstration. TMSR-LF1 at Wuwei in Gansu is a 2 MWt unit that reached first criticality on 11 October 2023, full power in June 2024, and had thorium loaded online, with the reactor at power, in October 2024. Two announcements followed, neither peer-reviewed: in April 2025 the programme director announced refuelling at full power at a closed Chinese Academy of Sciences meeting, with trade press noting the absence of peer-reviewed publication or third-party verification; on 3 November 2025 the institute formally announced thorium-uranium conversion. Established The reported conversion ratio is about 0.1. Breeding requires greater than 1.0. The thorium inventory is around fifty kilograms and the driver fuel is enriched below 20% uranium-235. The institute's own wording is careful — the first international experimental data obtained after thorium was introduced into a molten salt reactor, and the only operational molten salt reactor to have incorporated thorium fuel. Those are measurement claims. Handwave The framing that reached general readers — one major newspaper headlined it as China breeding uranium from thorium, with no conversion ratio and no independent scepticism anywhere in the piece — is a factor of ten away from the number the institute reported.

Frontier Lead coolant is less characterised than the confident literature suggests, and the gaps are specific. A national laboratory design report on lead and lead-bismuth gives no maximum cladding temperature — Russian structural testing went to 1000 °C, but no design limit is stated — and no maximum coolant velocity for erosion, noting ranges of 0.2 m/s natural convection to 2 m/s forced without correlating either to a limit. Oxygen control is the central mitigation and its numbers are real: control demonstrated across partial pressures from 10⁻²⁷ to 10⁻⁴⁰ atm, with distinct behaviour at 3×10⁻⁸ against 4×10⁻⁶ wt% oxygen. But protective oxide layer formation and stability mechanisms are described as incompletely understood, lead-bismuth's corrosive action on structural materials remains a significant technical issue, and long-term pool-type data are limited. The commonly quoted 550 °C practical ceiling for ferritic-martensitic steels in lead-bismuth is not verified in this brief's sources and is not printed here. The peer-reviewed oxygen-sensor review that would have supplied harder figures was behind a CAPTCHA.

Frontier The gas-cooled fast reactor is the least resolved family of the six. ALLEGRO, the sole demonstrator concept at 75 MWth, has no construction decision; work continues under a European project running 2024 to 2028. Peer-reviewed status as of June 2025 is that silicon-carbide-fibre composite cladding design rules and joining techniques still need to be developed, and that while passive decay heat removal established natural convection within sixty seconds in a test facility, further iteration is needed when the conservative methodology is applied.

Frontier The open question for microreactors is not whether they work but whether anyone will be permitted to run them the way the cost model assumes. Every microreactor economic case assumes staffing far below conventional practice — a handful of people per site, monitoring from a central operations centre, and in the ambitious versions semi-autonomous operation between refuellings. Established Operator licensing, control-room manning and physical-protection rules were written for gigawatt plants and scale with the site rather than the megawatts: an armed response force sized to a fixed site costs the same whether the site makes 5 or 1,000 megawatts. Frontier The decisive unknown is therefore a rulemaking outcome rather than a test result — whether a regulator grants reduced staffing, remote operation and risk-informed security for a small source term. Until that rule exists, every microreactor levelised cost in circulation is conditional on a regulatory assumption its author does not control, and the assumption is almost never stated in the same document as the number.

Frontier The second open question is transport, treated as solved logistics and not solved. A transportable microreactor’s case rests on moving a fuelled or irradiated unit by road, rail or air, and moving fissile or irradiated material requires a certified transport package for the thing being moved. Handwave No source consulted for this brief establishes a certified package for a fuelled microreactor of any design, and the brief records the absence rather than supplying a date. Speculative The plausible resolution is that first-of-a-kind units are fuelled on site and never moved once irradiated — which preserves the safety case and deletes the transportability that most of the defence and mining arguments are built on.

Established The first bottleneck is a queue for irradiation, not a gap in understanding. Qualifying a structural material for nuclear service requires irradiating it to representative doses across multiple temperature and dose combinations and then testing it, and the world has very few materials test reactors. That is why graphite qualification runs from 2005 to 2030 while a non-irradiated alloy took twelve years. No amount of private capital shortens an irradiation campaign, and none of the companies now depending on the queue funds it or can expand it.

Established The second is that the two materials most designs need are the two that are not in the code. ASME Section III Division 5 governs high-temperature nuclear components, and a material not in it cannot go into a safety-related component without a bespoke case. Graphite is not in it — no grade, anywhere. Hastelloy N is not in it, sixty years after the reactor it was designed for. Those two absences constrain more designs than any regulator's schedule does, and they are invisible in coverage that measures progress by licences granted.

Established The third is enrichment, and it is the constraint the 2025–26 legislation did not touch. Commercial high-assay low-enriched uranium is available only from Russia's state exporter. In December 2022 TerraPower announced a minimum two-year delay to Natrium's 2028 in-service date because the invasion of Ukraine had removed what it called the only commercial source of the fuel. The American allocation programme distributed material to five of fifteen requesting companies on 9 April 2025, against a Congressionally directed schedule of three tonnes by 30 September 2024, eight tonnes by 31 December 2025 and ten tonnes by 30 June 2026 — twenty-one tonnes in total. Twenty-one tonnes is a research allocation, not a fleet fuel supply. Established The clearest evidence that developers regard this as binding is a design choice: Terrestrial Energy scoped its IMSR around standard-assay uranium tetrafluoride below 5% uranium-235, explicitly to escape the dependency.

Established The fourth is reprocessing, and it is economic rather than technical. Global commercial reprocessing capacity is roughly 2,000 t/yr against world spent-fuel arisings of roughly 10,000 to 12,000 t/yr — so even a fully utilised world fleet of reprocessing plants would fall behind the pile every year. Japan's Rokkasho plant, begun in 1993 against an original completion target of 1997, reached its twenty-seventh delay in August 2024. The best available economic analysis says the cycle does not pay, and it is twenty-one years old: at a reprocessing cost of $10,000/kgHM the breakeven uranium price for light-water recycling is $3,680/kg U, roughly eight times the then-market price, and reprocessing adds about 1.3 mills/kWh, or roughly $10 million a year for a 1 GW reactor.

Frontier The fifth binds only lead, and it binds because nobody has written the limits down. The corrosion regime that keeps molten lead from dissolving steel has no stated maximum cladding temperature and no stated maximum coolant velocity in the national laboratory report consulted, and the oxide-layer mechanisms that make oxygen control work are described as incompletely understood. BREST-OD-300 is the first civil attempt at power scale, was about 70% assembled in January 2026, and will be operating for four years before its on-site reprocessing module is commissioned.

Established The sixth is that the demonstration is not the demonstration. Criticality is a milestone that can be reached at zero power with no heat removal and no power conversion, and in 2026 four of them were. What has never been done in any advanced fission programme is the industrial-scale closed fuel cycle that justifies the fast reactor, and no schedule in this brief puts it inside the decade.

What is not a bottleneck, stated plainly. Established Reactor physics is not a bottleneck. Fast spectra, salt chemistry and high-temperature gas have been understood since the 1960s, and every one of the six families is a well-posed engineering problem with a known solution shape. Licensing is now, arguably, also not a bottleneck: the American regulator issued a technology-inclusive framework in March 2026, nearly two years ahead of its statutory deadline. The constraint that remains is that the objects these reactors are made of have not been qualified, and qualification is a slow public process that legislation does not accelerate.

Established A further bottleneck is that microreactors intensify the enrichment constraint rather than escaping it. A unit rated in single-digit megawatts that must run five to ten years between refuellings needs a loading that is high in enrichment and high in burnup at once, which is why nearly every design in the American pilot programme specifies high-assay low-enriched uranium. The same supply queue that cost the flagship sodium project at least two years now serves a fleet of small units with less schedule slack each, and a small reactor does not wait more cheaply than a large one. Frontier Whether early units are fuelled from government inventory or from commercial enrichment is a policy choice, and no allocation rule beyond the first competitive round has been published.

Frontier The heat-application bottleneck is coupling rather than generation. Delivering nuclear heat to a chemical plant needs an intermediate loop isolating the reactor from a plant whose operators are not nuclear licensees, a licensing boundary nobody has drawn in production, and a heat exchanger qualified for the duty. Established The demonstrated nuclear-hydrogen projects to date sited low-temperature electrolysers beside conventional light-water plants and bought their electricity, which is a useful result pointing the other way: the hydrogen case was made without the high temperature the high-temperature designs are justified by. Frontier High-temperature steam electrolysis and the thermochemical cycles do use the heat, and Japan’s high-temperature test reactor has operated at 950 °C, but no coupled reactor-to-hydrogen plant runs commercially anywhere.

Established This brief carries no typed dependency edge, and as with its sibling the absence is the finding. Nothing on the critical path for advanced fission is a result that another brief on this map could deliver. The physics of fast spectra, salt chemistry and high-temperature gas has been understood since the 1960s. What advanced fission waits on is a code case for nuclear graphite, a codified nickel alloy for fluoride-salt service, a closed TRISO acceptance envelope with real radionuclide release limits, an accepted method of in-service inspection for a salt primary circuit, and irradiation capacity to produce all of them. Those are standards-body and test-facility constraints. They are institutional and industrial, not scientific.

Established It also depends on an enrichment supply chain that currently has one commercial source outside the West. High-assay low-enriched uranium is the fuel form most advanced designs specify, its only commercial supplier is Russia's state exporter, and a single geopolitical event in 2022 pushed the flagship American programme back by at least two years. That is a dependency on a particular trading relationship rather than on a capability, which makes it both easier to fix and easier to break than a technical one.

Frontier Fast reactors depend, additionally, on a reprocessing industry that does not exist at the required scale. World commercial capacity of about 2,000 t/yr against arisings of 10,000 to 12,000 t/yr is not a bottleneck that a breeding programme can grow into; it is an order-of-magnitude shortfall in the industry that would have to close the cycle. And the economics of closing it, on the best available analysis, do not work at any uranium price the market has seen.

Established What depends on it is stated more carefully than usual. Industrial process heat above the range electricity reaches well, and a closed fuel cycle that would extend uranium resources by orders of magnitude and consume much of the long-lived waste, are the two genuine dependants. Both are assumed generically by several topics on this map, which is why no typed enabling edge is claimed: those assumptions are about firm heat and firm power in the abstract, not about this particular route to them. Advanced nuclear propulsion shares the fuel vocabulary and inherits the same TRISO and HALEU constraints. Nuclear waste solutions, the unbuilt companion slot, inherits whatever this brief's cycles fail to close.

Established The 2026 policy wave produced a measurable result and it is smaller than the headlines. The American reactor pilot programme's stated goal, from the department's own page, was reaching criticality for at least three advanced reactor concepts located outside of the national laboratories by 4 July 2026. Four reactors went critical by the deadline. Every one was zero power — no heat removal, no power conversion, no electricity.

Frontier Three source disagreements sit inside that table and this brief does not resolve them. The count of programme participants is ten according to the professional society and eleven according to a private analyst who computes a 36% success rate from it; the department states no number publicly. The count of companies that hit the deadline is three in the society's checkpoint article and four in the trade press, which quotes the department as having surpassed the ask. The dates for two of the four differ across sources, as shown. And Valar's criticality is arguably two events: a think tank's press release describes cold criticality of the NOVA Core, a subsection of the Ward 250 core, on 17 November 2025 at Los Alamos National Laboratory's criticality facility — explicitly a self-sustaining chain reaction without full operating temperatures or active heat removal — with the June 2026 Utah event reported separately. Which event satisfies the programme goal is not stated anywhere consulted.

Frontier Three of the four went critical at Idaho National Laboratory, against a goal explicitly worded as criticality outside the national laboratories. On the location data, the goal as literally written was met once rather than three times. This is flagged frontier rather than established because the department has published no assessment against its own wording, and the analyst source that makes the count is a private outfit whose standing could not be established — the same source misidentifies a fifth reactor, of which more below.

Frontier A fifth and more substantial criticality followed a month later. Oklo's Groves isotope test reactor is reported to have achieved first criticality on 5 August 2026 in Texas, privately sited and privately funded, claimed as the fastest privately funded, privately sited reactor build in history at less than a year from construction start. That is a vendor disclosure relayed by financial media; the coolant, power level and fuel were not stated in the report consulted. Established It is not Oklo's Aurora-INL, which the private analyst's write-up confuses it with: Aurora-INL is targeted for late 2027 to early 2028 and Aurora-Ohio for early 2030.

Established The one construction start that matters is in Wyoming. An independent status report notes that in the West there is only one design certification and one standard design approval, both for the same American light-water small reactor, and no constructions. The sole Western commercial-scale Generation IV construction start is TerraPower's Kemmerer 1, permitted in early March 2026 and begun on 24 April 2026 — the first American construction approval for a commercial non-light-water reactor in more than forty years. It is a 345 MWe sodium fast reactor, boosted to 500 MWe by molten-salt thermal storage, targeting completion in 2030.

Frontier The natural experiment worth watching in salt is at Wuwei. TMSR-LF1 is the only operating molten salt reactor in the world and the only vehicle testing thorium conversion in a salt. The decisive readout is not another announcement but a published conversion ratio, from a peer-reviewed source, above the 0.1 the institute has reported — and the intermediate readout is whether any of this reaches peer review at all, since both announcements to date came through a closed academy meeting and an institutional release.

Established A negative result worth recording as an experiment in its own right. The graphite creep campaign is a twenty-five-year irradiation programme whose result will be either a code case or no code case, and it will report around 2030 regardless of what anyone announces in the meantime. It is the cleanest scheduled test on this map: nothing that depends on nuclear graphite can be finally licensed against a qualified reference until it concludes.

Established The high-temperature gas story contains a design retreat that is rarely reported, and it was made three separate times by three separate programmes. The concept's efficiency case rested on a direct helium turbine — helium expanding through a closed Brayton cycle, no steam, thermal efficiencies in the forties. The one operating plant uses a steam cycle instead: the helium heats water and a conventional turbine does the work. That is a sound engineering decision, it gives up a large part of the thermodynamic advantage that justified the coolant, and it was not made once but repeatedly.

Established The best documentation of why is a vendor assessment that is adverse to the design it assessed. A gas-turbine manufacturer's preliminary assessment of the GT-MHR power conversion system, produced for the American next-generation plant programme, found the following. Electromagnetic bearings were rated high risk and outside current world experience: a thirty-five-tonne rotor at 4,400 rpm exceeded established bearing manufacturers' capability, and on bearing failure the catcher bearings would undergo significant crushing loads and frictional heat build-up, risking thermal damage without oil lubrication. The recuperator was rated very high risk: one unit needed 350 tonnes of material, roughly 796,000 weld operations and about 50 km of total weld length, sited inside the power conversion unit where maintenance is difficult. The power electronics alone for a 4,400 rpm, 286 MW generator were costed at roughly $50 million. Helium's high specific heat ratio demands far more turbine stages than a conventional gas turbine, lengthening and adding mass to the shaft, and assembly required seven annular seal rings from 1 m to 3.2 m in diameter installed blind with a thirty-five-tonne rotor. The recommendation was a combined cycle — a 66 MWt gas turbine plus a conventional steam cycle — eliminating recuperator risk, cutting electromagnetic bearing loads by about 70%, and using commercial off-the-shelf steam equipment.

Established That is one of three abandonments. A national laboratory's lessons-learned report records that the 1986 modular high-temperature gas reactor design rejected the direct gas turbine on economic grounds, given the smaller cores needed for passive decay heat removal, and that post-2007 concepts moved to an indirect cycle on an independent review group's recommendation for reasonable development risk. South Africa's PBMR abandoned the direct Brayton cycle in 2009, moving from 400 MWt and 165 MWe direct-cycle to 200 MWt and 80 MWe conventional Rankine. Frontier The reasons for that third decision were not stated in the source consulted and are recorded here as unestablished. An earlier paper from the Chinese programme identified the core obstacle as the absence of an integrated test with a real primary system environment, with specific concerns about radioactivity deposition on turbine blades and turbine shaft overspeed; HTR-PM's cycle was then still an open choice, and it went with steam. Frontier Whether direct cycle is permanently dead is open; that it has been rejected three times on four independent grounds is not.

Established Graphite is the other high-temperature gas constraint and it is physical rather than economic. Under neutron irradiation graphite first shrinks and then swells, and the turnaround point sets component life. Without a code case, each design must argue its graphite from first principles to a regulator with no qualified reference to compare against. The historical fleet shows what graphite actually does in service. Peach Bottom 1, 115 MWt, running 1966 to 1974, suffered 45 to 84% fuel particle failure in its first core from the amoeba effect — kernel migration across pyrolytic carbon coatings under thermal gradients — which was fixed by moving to BISO coatings and led onward to TRISO. Fort St Vrain, 842 MWt, 1976 to 1989, took in water from water-lubricated circulator bearings, absorbed it into the pressure vessel insulation and released it on startup and shutdown, causing fuel particle hydrolysis and corrosion of graphite support posts; helium pressurisation lines plugged with carbon steel corrosion product; core temperatures fluctuated from 1977 to 1980 at 30 to 70% power from bypass flow and component motion; and eighty-four region constraint devices had to be retrofitted.

Established The single most important negative result in pebble-bed history is a graphite result, not a fuel result. AVR, a 46 MWt pebble bed running 1967 to 1988, generated between 46 and 200 kg of graphite dust over twenty-one years through notching, spalling, pitting, fracture and air-ingress peeling. That dust became the principal carrier of mobilised radioactivity — caesium-137, caesium-134, strontium-90, silver-110m and iodine-131 — accumulating uncontrollably through the primary circuit. Any pebble-bed design proposal that does not address dust transport is not engaging with the operating record. Frontier THTR-300, 750 MWt, 1985 to 1991, is attributed by that same national laboratory report to sociopolitical reasons; the report is from a laboratory with a programmatic interest in high-temperature gas reactors, other literature emphasises pebble jamming and cladding problems, and no independent German account was consulted, so the sociopolitical framing is recorded here as an interested reading rather than a neutral one.

Established Molten salt's engineering problems are chemical and they are documented by a regulator. Tellurium, a fission product, caused intergranular cracking in all metal surfaces exposed to the fuel salt at Oak Ridge. Frontier The regulator's position is that tellurium embrittlement remains a concern if salt redox chemistry is not well controlled — mitigation through uranium(IV)/uranium(III) ratio control exists in principle and is not demonstrated over reactor lifetimes. Established Tritium production was calculated at about 54 Ci/day, of which roughly 6 to 10% diffused into containment and a further 6 to 10% reached air through heat removal; Frontier lithium-7 enrichment is the mitigation and is not demonstrated at scale. Frontier Compatibility and corrosion data are described by the regulator as scattered and poorly controlled, with chloride-salt data especially sparse. And there is no accepted way to inspect a molten-salt primary circuit in service: the regulator's high-priority recommendation is for an in-situ passive materials surveillance programme, which is a recommendation to invent one.

The nearest neighbour is small modular reactors, and the boundary between them is a source of persistent confusion that this brief tries to hold firmly. Modularity is a delivery model — factory production, series learning, the economics of building small. Advanced fission is a set of coolants and fuel cycles. Many designs are both, and the standing example is instructive: TerraPower's Natrium is frequently discussed as a molten-salt design and is nothing of the kind. It is a 345 MWe sodium fast reactor whose molten salt is thermal storage only, boosting output to 500 MWe on demand. The storage is a grid-integration innovation and belongs to the modularity and market-design argument; the sodium fast core belongs here.

Nuclear waste solutions, Energy Systems slot 15, is the necessary companion and is not yet built. It matters more to this brief than to most, because the reprocessing question sits exactly on the seam: whether spent fuel is a waste stream to be disposed of or a fuel stock to be recycled is simultaneously a reactor-design question and a waste question, and the answer that the economics currently give — roughly 25% of about 448,000 tonnes of heavy metal reprocessed, the rest in storage — is the reason the seam has not been resolved in sixty years.

Advanced nuclear propulsion shares the fuel vocabulary, the TRISO particle and the high-assay enrichment constraint, and optimises for thrust-to-weight where this brief optimises for cost and lifetime. Anything that closes the TRISO acceptance envelope helps both.

Commercial fusion is the useful contrast rather than a competitor: fusion's binding constraint is physics and engineering at the frontier, while fission's is paperwork and irradiation queues on materials whose behaviour has been studied for decades. Anyone who believes fission is nearly here and fusion is far away should be able to say which of those two constraints they think is easier to remove. Geothermal megaprojects are adjacent as the competing firm-power claim, and are the honest alternative to compare against on capital cost per delivered megawatt-hour rather than on elegance.

Established The institutional pattern is that legislation moves faster than qualification, and the two are routinely confused. The ADVANCE Act and Part 53 are real changes to how quickly a licence can be obtained, and the regulator delivered the second nearly two years ahead of its statutory deadline — the first new initial licensing regulations since 1989 and the first major framework update since 1956. Neither shortens an irradiation campaign by a day. A licensing milestone reported as a technology milestone is the central category error in this field's coverage, and it is committed most often by people who are not being dishonest, because the licensing news is genuinely new and the qualification news has not changed in ten years.

Established The standards bodies are the actual gatekeepers and they are slow by design. ASME Section III Division 5 governs high-temperature nuclear components, and a material not in it cannot go into a safety-related component without a bespoke case argued from first principles. Graphite is not in it — no grade, anywhere. Hastelloy N is not in it. Those two absences constrain more designs than any regulator's schedule. Neither absence has a lobby, a press release or a funding line proportional to its importance.

Frontier There is a live question about where qualification capacity comes from, and it has no owner. Materials test reactors are public infrastructure, largely built decades ago, and the graphite campaign is explicitly rate-limited by their availability. The current wave of private advanced-reactor companies depends on a queue it does not fund and cannot expand, and no institution in any country has a mandate to expand it on their behalf. This is the clearest instance in the corpus of a shared input that everyone needs and nobody is responsible for.

Established The regulator's own gap assessments are the most useful documents in this field and are almost never cited in coverage of it. The molten salt technical gap assessment and the joint TRISO evaluation both say, in plain regulatory language, that materials the industry describes as ready are not. A regulator saying a material is unsuitable is stronger evidence than a vendor saying it is suitable, because the regulator has no product to sell and a strong incentive against being wrong in the permissive direction. That asymmetry is the most useful evidentiary rule available to a reader of this subject.

Frontier The institution that does not exist is an independent operator-data channel for the plants that matter most. The two most informative operating reactors in advanced fission — HTR-PM and CFR-600 — publish essentially nothing. An independent status report states there is little to no information available on the operational experience of the two HTR-PM modules, and no load factor, availability, breeding ratio or burnup data for CFR-600 could be found anywhere. Frontier That opacity has a concrete consequence. HTR-PM's 2024 load factor is widely quoted as 20.70%, but the same report states that nominal capacity was cut by a quarter, from a combined 200 MW to 150 MW, for unknown reasons, while a tertiary source gives 210 MWe. A load factor is only as good as its denominator, and 20.7% of 150 MW supports a very different argument from 20.7% of 210 MW. This brief does not resolve which rating the figure was computed against, and does not quote it as self-explanatory.

Established Who buys, and who is not in the room. The buyers of advanced fission today are governments and government-adjacent programmes: a department of energy running a criticality-deadline pilot, a defence-adjacent isotope market, a state atomic energy department pursuing a three-stage plan since 1954, and a Chinese academy institute. No utility is a volume purchaser of an advanced fission design anywhere in the West; the one Western commercial-scale construction start has a public power-purchase structure behind it. The institution most conspicuously absent is a standards body with a funded mandate and a schedule for advanced-reactor materials — which is to say, the one institution whose existence would change every date in this brief.

Frontier One further absence, recorded because it is the honest state of the record. This brief could not establish the capital cost of either of the world's two largest reprocessing plants. Neither the independent panel's reporting on Rokkasho nor the trade press consulted gave a figure, and the widely circulated lifecycle number for Rokkasho was not verified. A brief that argues reprocessing does not pay ought to be able to state what the plants cost, and this one cannot.

Frontier The distinctive ethical question is proliferation, and it varies sharply by fuel cycle rather than by reactor. Reprocessing separates plutonium, and thorium cycles produce uranium-233, which is weapons-usable in principle and protected in practice only by the hard gamma of the uranium-232 that accompanies it — a safeguard that better remote handling erodes, and which is absent in the brief window immediately after chemical separation before the daughter products grow back in. That window is a technical fact reported by the same tertiary source that gives the 2.6 MeV figure, and it is the reason the self-protection argument for thorium should be treated as a delay rather than a barrier.

Evidence quality is the second ethical issue and it is unusually acute in this field. Established A large share of what is publicly known about advanced fission comes from parties with a direct interest: an industry association that owns the main trade news outlet, national laboratories with programmatic stakes in the technologies they assess, vendors announcing their own milestones, and government programme offices reporting against their own targets. This brief marks every such source in its reading list and states what it relies on each for. The pattern it applies is that interested parties are reliable on facts they would be embarrassed to get wrong — dates, tonnages, licence numbers — and unreliable on framing. The strongest single instance runs the other way and is worth naming: an industry association conceding that protactinium separation is unproven at any scale is powerful precisely because it is against interest.

Established Public money and public accountability are entangled here in a way that deserves stating. The materials test reactors, the graphite irradiation campaign, the code qualification work and the enrichment programmes are all public. The reactor companies whose schedules depend on them are largely private. That is a defensible division of labour, and it becomes a disclosure problem when a company's announced 2030 deployment date depends on a qualification campaign scheduled to report in 2030 and the announcement does not say so. Where public money is allocated against the announced date rather than the qualification date, the mismatch is not a forecasting error but a disclosure one.

Established Opportunity cost is the question the field asks least about itself. Roughly $50 billion in 2007 dollars went into fast breeder research in OECD countries between 1974 and 2007 and produced one reactor with a good operating record. Monju alone consumed about US$8.5 billion and will consume roughly $3.2 billion more to demolish, for 250 days at power. The argument is not that the research should not have been done; it is that the same money spent on irradiation capacity, code qualification and a single repeated design would plausibly have produced more, and that the current wave is structured to repeat the mistake.

This brief's own unresolved questions, stated as an obligation. Four things it would need to make its case cleanly and could not verify. First, an achieved breeding ratio from any fast reactor ever built — none exists in the comprehensive critical review consulted, and the brief therefore rests its breeding scepticism on an absence rather than on a measurement. Second, whether India's fast test reactor cycle closure included re-irradiation of refabricated fuel, and in what quantity; the programme asserts closure and the supporting conference paper gives a burnup and no masses. Third, operational data of any kind for China's CFR-600 — load factor, availability, grid connection date after October 2023, burnup, breeding — none of which is published anywhere the pack behind this brief could find, which means the second-largest operating fast reactor programme in the world is invisible to outside assessment. Fourth, the denominator of HTR-PM's load factor, discussed below, which the brief could not resolve because the intergovernmental database renders only as a JavaScript application. Each of these is a place where a reader should discount this brief's confidence, and saying so is cheaper than being caught.

Frontier The civilisational case is real and it is worth stating at full strength before discounting it. A closed fast-reactor fuel cycle would turn a few decades of uranium resource into several thousand years of energy and would consume much of the long-lived waste it produced, converting the most politically intractable feature of fission into a fuel stock. High-temperature gas would supply industrial process heat in a range that electrification serves badly — cement, steel reduction, chemical feedstocks — which is a decarbonisation problem with no other good answer on this map. These are among the largest prizes anywhere in the corpus.

Established The honest assessment is that none of it is blocked by physics and all of it is blocked by qualification. That is a better position than fusion is in and a worse one than the announcements imply, because qualification clocks are long, publicly funded, and indifferent to capital. A private company can raise a billion dollars and cannot raise a materials test reactor. The single most useful thing anyone could do for advanced fission is build irradiation capacity, and almost nobody is proposing it — because it produces no reactor, no ribbon-cutting and no revenue, and it takes a decade to matter.

Established The general principle this case illustrates is about what learning curves actually require. Four hundred reactor-years of sodium experience across eight countries and roughly $50 billion of public research produced one plant with a good operating record. That is a real result about industrial learning: experience accumulated across dissimilar prototypes, in different countries, under different regulators, with each programme cancelled before its successor was designed, does not compound. The learning curve is a property of repetition, and the fast reactor programme was never allowed to repeat anything. The same warning applies directly to the current wave, in which a dozen companies are each building one first-of-a-kind unit.

Speculative And there is a scenario in which the qualification constraint is simply out-waited by someone else. China has the only operating high-temperature gas plant, the only operating molten salt reactor, an operating sodium fast reactor whose data are not published, and a state willing to fund twenty-year materials programmes without quarterly justification. If the code cases that Western designs wait on are ultimately written against Chinese operating experience, the institutional geography of this technology changes in a way that no Western licensing reform addresses. Nothing in the sources consulted establishes that this is happening; it is a reasoned extrapolation from who is currently operating what, and it is flagged accordingly.

Established What already happened, because the timeline for this subject usually starts sixty years too late. The Aircraft Reactor Experiment ran for nine days in November 1954. The Molten-Salt Reactor Experiment ran from 1965 to 1969 and was the first reactor in the world to operate on uranium-233. Peach Bottom 1 ran from 1966, AVR from 1967, Fort St Vrain from 1976, THTR-300 from 1985, Superphénix from 1985; all are shut. The graphite creep campaign began in 2005 and started irradiating in 2009. OECD countries had spent about $50 billion in 2007 dollars on fast breeder research by 2007. France cancelled ASTRID in 2019.

Established 2021 to 2026: what was actually delivered. HTR-PM's two units went critical in September and November 2021, reached the grid on 20 December 2021, full power on 9 December 2022, and commercial operation in December 2023. TMSR-LF1 went critical on 11 October 2023 and loaded thorium at power in October 2024. The ADVANCE Act was enacted in July 2024, the first Category II TRISO fabrication licence issued on 13 February 2026, and the Part 53 framework — the first new initial licensing regulations since 1989 — on 25 March 2026, nearly two years early. Construction began at Kemmerer on 24 April 2026. India's prototype fast breeder reactor reached first criticality in April 2026, more than fifteen years late.

Frontier 2026 to 2027: BREST-OD-300 and Rokkasho. Russia's lead-cooled demonstrator was about 70% assembled in January 2026, with physical launch scheduled for 2026 and grid connection targeted for the first half of 2027. No source consulted compares that schedule against its original dates, so the slip cannot be quantified here. Rokkasho's current target is fiscal 2026, ending March 2027. Frontier As of May 2026 the prefectural governor stated that the project will definitely be delayed again, while government officials maintained the deadline was firm; the operator had missed a November 2025 documentation milestone and a March 2026 approval milestone. Whether a twenty-eighth delay has been formally declared is not established. The associated MOX fabrication plant has already slipped from the first half of fiscal 2024 to fiscal 2027.

Handwave 2029 to 2030: dates that have already moved. Kairos's Hermes test reactor — 35 MWt, fluoride-salt-cooled, solid TRISO fuel, not a power reactor — had its construction completion deadline extended by the regulator from 31 December 2026 to 30 April 2029 in May 2026, a slip of about 2.3 years, on grounds of first-of-a-kind construction demonstrations and site preparation that required significantly more work than planned. India's fast reactor fuel cycle facility, originally due in 2014, is now expected in December 2029 — a fifteen-year slip on the reprocessing half of the cycle. BREST's reprocessing module is targeted for commissioning in 2030. Kemmerer targets completion in 2030. Terrestrial Energy's stated first plants are early 2030s with no construction and no licence. Every date in this paragraph belongs to a programme that has already moved one.

Established Around 2030: the graphite creep campaign concludes and either produces a code case or does not. This is the one scheduled milestone on the map that is not a company's target, and it gates everything graphite-moderated.

Handwave 2035 and beyond: the thorium demonstration and the three-stage programme. China targets a 100 MWt thorium salt demonstrator by 2035, with no construction start verified. India's own peer-reviewed programme literature conditions the introduction of thorium into its fast reactor fleet on that fleet reaching about 220 GWe; four months after first criticality the fleet stands at 0.5 GWe. Stage three is not decades behind. On the programme's own stated threshold, it is not scheduled in any meaningful sense.

One microreactor clock is worth separating from the reactor dates above, because it is set by rulemaking and by materials codes rather than by criticality:

• 10 yr: Frontier The binding item is a staffing and physical-security rule for small source terms rather than a first criticality, because pilot units can be built and still be uneconomic to operate under gigawatt-plant manning. Established The process-heat ceiling inside the same window is set by a 2020 code case rather than by anything a demonstration will show.
• 25 yr: Speculative Either a certified transport package for a fuelled unit exists and the transportable case survives, or first-of-a-kind units are fuelled in place and the defence and mining arguments quietly become site-built ones. Handwave Rulemaking schedules are not technology schedules, and this brief does not forecast them.

• Depends on Nothing on this map. The physics of fast spectra, salt chemistry and high-temperature gas has been understood since the 1960s; nothing on the critical path is a discovery, and everything on it is a qualification. That is a stronger position than fusion is in and a considerably weaker one than the announcement volume implies, because qualification clocks are long, publicly funded and indifferent to private capital.
• Requires (not on this map) An ASME Section III Division 5 code case for nuclear graphite, of which zero grades are currently qualified. A codified nickel alloy for fluoride-salt service, Hastelloy N having been described by a regulator as unsuitable and estimated in 2018 as ten years from qualification. A closed TRISO acceptance envelope, four of whose sections are marked “to be completed” in the regulators' own joint evaluation. Materials test reactor capacity to produce all three, which is the physical queue behind the twenty-five-year graphite campaign. High-assay low-enriched uranium enrichment capacity outside Russia, whose absence cost the flagship American programme at least two years. And commercial reprocessing capacity roughly five times what exists, if the closed cycle is ever to be more than a pilot campaign. All six are standards-body, test-facility, supply-chain or industrial capabilities rather than discoveries. A seventh constraint sits one level above the materials, and it arrived with the microreactor fleet rather than with any coolant. Operator licensing, control-room manning and armed-response requirements scale with the site rather than with the power, so a fleet of small units inherits a large plant’s fixed regulatory overhead per site, and every microreactor cost estimate in circulation assumes relief that no regulator has yet granted. It is a rulemaking rather than a result, which puts it with the other six and not with the physics.
• Enables Industrial process heat above the range electricity serves well, and a closed fuel cycle that would extend uranium resources by orders of magnitude while consuming much of the long-lived waste. Several topics assume both generically, so no typed enabling edge is claimed — those assumptions are about firm heat and firm power rather than about this route to them.
• Adjacent Small modular reactors, which shares hardware and argues a different question; nuclear waste solutions, the unbuilt companion slot that inherits every cycle this brief fails to close; advanced nuclear propulsion, which shares the TRISO and HALEU constraints and optimises for thrust rather than cost; and commercial fusion, whose constraint is physics where this one's is qualification.

“Generation IV is nearly here.” Handwave Of six coolant families, three have never been built at all, one has never generated a kilowatt-hour, and the one with 400 reactor-years produced Superphénix at under 7% lifetime load factor. The claim mistakes a licensing milestone for a materials qualification. The binding constraint is not reactor physics; it is that the structural materials these reactors are made of are not code-qualified, and the qualification clock runs in decades.

“TRISO is already qualified.” Frontier The regulators' own joint evaluation carries four sections marked “to be completed” — radionuclide release limits, test envelope validation, data measurement methodology and test conditions — and records that the irradiation campaigns did not establish a correlation between silicon-carbide microstructure and particle performance. The operating high-temperature reactor's fuel was qualified on five pebbles irradiated in the Netherlands over twenty-seven months, with burnup achieved and failure fractions not stated in the source consulted. The February 2026 fabrication licence authorises a factory, not a performance envelope.

“Molten salt reactors were proven at Oak Ridge.” Established Two reactors ran, in 1954 and from 1965 to 1969, and neither had a power conversion system. No molten salt reactor has ever generated electricity, anywhere, ever. What the experiments did demonstrate — operation on uranium-233, high availability in the final campaigns — is real and is not the same as demonstrating a power plant. What they did not demonstrate is a breeding blanket, online reprocessing at scale, or thorium.

“Five companies are building molten salt reactors.” Established Two of the names usually on that list are not molten salt reactors in the fuel sense. TerraPower's Natrium is a sodium-cooled fast reactor in which the molten salt is thermal storage only. Kairos's Hermes is a fluoride-salt-cooled test reactor with solid TRISO fuel and no power conversion — and its construction completion deadline moved from December 2026 to April 2029, a 2.3-year slip on a 35 MWt test unit. Only Terrestrial Energy, Moltex, ThorCon and China's TMSR-LF1 are salt-fuelled. Lumping them overstates salt-fuel maturity considerably, and it is the single most common structural error in coverage of this field.

“China has bred uranium from thorium.” Established This is the most commonly mis-stated fact in the subject. The institute's own reported conversion ratio is about 0.1; breeding requires greater than 1.0, so the claim is off by a factor of ten. What was achieved is real and worth respecting: online thorium loading with the reactor at power in October 2024, and measurement of thorium-uranium conversion announced on 3 November 2025. Neither is peer-reviewed; the first was announced at a closed academy meeting in April 2025 with no third-party verification. The institute's own wording claims a measurement, not a breeding demonstration. The gap between that wording and the headlines is the story.

“Thorium is a cleaner, simpler fuel.” Established Thorium is fertile, not fissile, and every thorium concept needs a uranium-235 or plutonium driver to start. The cycle requires protactinium-233 extraction on a 27-day half-life clock — a process an industry association concedes is unproven at any scale — and produces uranium-232, whose decay chain reaches thallium-208 and a 2.6 MeV gamma at 99.75% probability, forcing remote fabrication rather than gloveboxes. India's own peer-reviewed programme literature puts thorium's entry at about 220 GWe of fast-reactor capacity; India has 0.5 GWe, and has had it for four months.

“Sodium fast reactors have 400 reactor-years, so they are the mature option.” Established The figure is correct, it is an experience number rather than a performance number, and it belongs to sodium-cooled fast reactors specifically rather than fast reactors generally — cite the intergovernmental agency, not the industry association that uses it loosely. Beside the outturn it means something different: exactly one plant, BN-600, combines a good record with commercial scale.

“India's prototype fast breeder reactor is now in operation.” Established It reached first criticality — at 20:25 on 6 April 2026 per India's own atomic energy department, on 7 April per an independent panel, almost certainly an Indian-evening event reported next day, and this brief prints both. It is not in commercial operation and no such date has been verified. It is more than fifteen years behind its original 2010 target and its cost roughly doubled, from ₹3,492 crore to ₹8,181 crore. Its fuel cycle facility, originally due in 2014, is now expected in December 2029. And it does not contain thorium: the department describes the design as eventually using thorium-232 in the blanket.

“Rokkasho has been delayed twenty-seven times.” Established Twenty-seven is correct as of August 2024 and comes from an independent panel that states it explicitly. Frontier A twenty-eighth is in progress: as of May 2026 the prefectural governor stated the project will definitely be delayed, government officials maintained the deadline was firm, and the operator had missed both a November 2025 documentation milestone and a March 2026 approval milestone. The honest formulation is “twenty-seven confirmed, a twenty-eighth expected.” The trigger for the twenty-seventh was a regulator's demand for more work on the ground model for earthquake-resistant design, with the regulator expressing concern that the project was schedule-driven — a regulator saying on the record that a plant thirty years into construction is being run to a calendar rather than a safety case.

“About 30% of the world's spent fuel has been reprocessed.” Established Two figures circulate and this brief prefers the intergovernmental one. The agency's June 2026 inventory, built on 2025 convention reporting, gives roughly 448,000 tonnes of heavy metal cumulative with about 25% reprocessed and 75% in storage, split 41% wet and 31% dry. An industry association gives roughly 400,000 tonnes and about 30%. The five-point spread is about 22,000 tonnes. And the stock figure is the less damning one: the flow is world capacity of about 2,000 t/yr against arisings of roughly 10,000 to 12,000 t/yr.

“Reprocessing has become economic.” Handwave No analysis consulted supports that, and the best available rebuttal is twenty-one years old, which is a weakness worth stating rather than hiding. The 2005 peer-reviewed assessment — academic authors, no vendor interest — put the breakeven uranium price for light-water recycling at $3,680/kg U against a reprocessing cost of $10,000/kgHM, roughly eight times the then-market price; called that cost conservative, with privately financed facilities possibly exceeding $20,000/kgHM; put MOX fabrication at $1,500/kgHM against historical prices spanning $1,900 to $24,000/kgHM; costed the penalty at about 1.3 mills/kWh, roughly $10 million a year for a 1 GW reactor; and concluded once-through would stay significantly cheaper for at least fifty years. No 2025 or 2026 peer-reviewed replacement was found, and the obvious current government source was blocked by a robots file. Treat a 2026 claim of MOX viability as unsupported — and the rebuttal's vintage as a real caveat.

“High-temperature gas reactors give you a helium turbine.” Established The one operating unit drives a conventional steam cycle. The direct cycle that justified much of the efficiency case has been abandoned three separate times by three separate programmes — on economic grounds in 1986, on the vendor assessment's grounds of magnetic bearing immaturity and recuperator manufacturability in the 2000s, and by PBMR in 2009, which moved from 400 MWt direct Brayton to 200 MWt conventional Rankine. This is not a technology awaiting its moment. The steam cycle is the settled answer, and it caps thermal efficiency well below what the direct-cycle promise rested on.

“Four advanced reactors went critical in 2026, so deployment is under way.” Established All four were zero power — no heat removal, no power conversion, no electricity — and three were at a national laboratory, against a goal specifying criticality outside the national laboratories. Seven of the roughly ten or eleven selected companies missed the deadline entirely. Criticality is a physics milestone; the deployment milestones are heat removal, power conversion and a grid connection.

“The 2024–26 legislation unblocked advanced nuclear.” Established It unblocked licensing, which was a genuine constraint and is now substantially less of one. The ADVANCE Act of July 2024 required streamlined licensing, an expedited qualification procedure, frameworks for fusion and microreactors, a lower hourly fee rate for advanced applicants and pre-applicants, and assessment of licensing at former fossil and brownfield sites, and it rewrote the regulator's mission statement to include enabling the safe and secure use and deployment of civilian nuclear technologies. Part 53 followed on 25 March 2026, technology-inclusive and delivered nearly two years ahead of its statutory deadline. None of that qualifies a graphite grade, codifies a nickel alloy, closes a TRISO acceptance envelope or enriches a kilogram of uranium.