Cold fusion's next announcement will not require a press conference. It will require three laboratories.
In March 1989, Martin Fleischmann and Stanley Pons claimed that electrolysis had produced more heat than chemistry could explain inside palladium loaded with deuterium. The expected nuclear products did not appear in proportion to the heat, other laboratories could not reproduce the result and cold fusion became a warning about discovery conducted in public before it had survived replication.
Thirty years later, a Google-funded collaboration returned with better calorimetry, better materials control and permission to find nothing. It found nothing. Its 2019 report said that the programme had produced no evidence of cold fusion, while identifying highly loaded metal hydrides and low-energy nuclear reactions as an underexplored scientific territory. The Google-funded re-examination
That territory has now produced a result. A 2025 benchtop experiment showed that electrochemically adding deuterium to palladium increased beam-driven fusion by 15%. In 2026, a second team drove the collision energy down to 0.25 kiloelectronvolts and found that fusion inside palladium and titanium stopped falling towards zero. The measured yield reached a plateau more than a quintillion times above the bare-nucleus expectation. The metal was participating in the nuclear event. The 2025 palladium experiment The 2026 sub-keV experiment
ParallaxSee forecasts that by the end of 2032, three independent laboratories will reproduce an accelerator-free nuclear reaction inside a hydrogen-loaded metal, establishing the scientific discovery of cold fusion. House confidence: 48%.
The discovery will not power a house. It will prove that a solid can do more than hold nuclear fuel: its electrons, defects and moving hydrogen can help determine whether nuclei meet. Cold fusion will arrive first as a materials effect, then as a scientific instrument and only much later—if the reaction can be multiplied—as an energy technology.
01 — The metal offers fusion a shorter wall.
Two deuterium nuclei repel one another because both carry positive electrical charge. The closer they come, the steeper that repulsion becomes. The strong nuclear force can bind them only after they reach a distance unimaginably smaller than the normal spacing between atoms in a solid.
Classical physics would leave the nuclei outside this wall. Quantum mechanics gives them a small probability of tunnelling through it. Hot-fusion machines raise that probability by giving nuclei enormous kinetic energy. A metal hydride offers a different starting point: it can pack deuterium at densities approaching one atom for every metal atom, place mobile electrons around the positive nuclei and create defects, strain and narrow pathways through which deuterium moves.
No electron cancels the nuclear barrier. The useful question is whether the collective material can lower its effective width, replenish reacting deuterium near favourable sites or create brief local conditions that make an otherwise invisible tunnelling probability measurable.
The 2026 experiment confined the observed reactions to a near-surface region less than one-tenth of a micrometre deep. Palladium and titanium behaved differently at higher energies, as their hydrogen solubility and transport properties suggested they should. Below approximately 2 keV, both developed the same unexpected floor in fusion yield. The sub-keV measurements and experimental controls
That shared plateau is the article's scientific foundation. Cold fusion no longer needs a metal that performs a miracle. It needs engineers to identify and reproduce the microscopic neighbourhood in which the probability stopped behaving normally.
02 — The failure of 1989 created the correct discovery test.
Fleischmann and Pons measured heat. Fusion normally leaves fingerprints: deuterium–deuterium reactions predominantly produce either a neutron and helium-3 or a proton and tritium. Enough conventional D–D fusion to create substantial heat would also create a radiation field too large to overlook.
The mismatch between the claimed energy and the missing nuclear products became fatal. Calorimetry at the edge of instrument accuracy is vulnerable to calibration drift, recombination, changing gas flow and assumptions about how heat escapes. A warm cell alone cannot establish a nuclear reaction.
The Google-backed programme rebuilt the investigation around this lesson. Researchers developed demanding calorimeters, prepared highly hydrided materials and tested several proposed regimes. They reported no anomalous effect that survived ordinary explanation. The result did not prove that every arrangement of hydrogen and metal is inert. It established the evidential price of reopening the case. The 2019 multi-institution investigation
A qualifying 2032 experiment will therefore begin with nuclear products. It will measure their energies, directions and timing while simultaneously measuring heat. It will substitute ordinary hydrogen for deuterium, vary loading, cycle the stimulus on and off and publish the raw detector and calibration data.
Cold fusion will be rediscovered by behaving unlike cold fusion did in 1989: the reference experiment will come before the revolution.
03 — Thunderbird proved that chemistry can turn a nuclear dial.
The 2025 Thunderbird Reactor is small enough to fit on a laboratory bench and too large to hide what drives it. A microwave plasma supplied deuterium ions. A 30-kilovolt potential accelerated them into a 300-micrometre palladium membrane. On the opposite side, an electrochemical cell loaded additional deuterium from heavy water into the same metal.
With the beam operating, the reactor produced neutrons at energies expected from D–D fusion. Turning on electrochemical loading increased the rate by an average of 15%, with a reported uncertainty of two percentage points. Ordinary water instead of heavy water sharply reduced the rate. The experiment repeated the effect across multiple targets and separated neutrons from gamma-ray background using pulse-shape discrimination. Electrochemical loading enhances deuterium fusion rates
This was not cold fusion. The incoming ions carried 30 keV and the apparatus consumed 15 watts to sustain its plasma. Its fusion yield was equivalent to only one billionth of a watt—approximately fifteen billion times below energy break-even.
Its importance lies elsewhere. Electrochemistry operates at the electronvolt scale; the detected nuclear products carried megaelectronvolts. Changing the chemical state of the palladium changed the nuclear reaction rate. The cell had become a control surface for fusion.
Thunderbird separated a defensible claim from the old promise. It did not say that chemistry supplied fusion energy. It demonstrated that chemistry changed the target in a way a nuclear detector could count.
04 — In 2026, the falling reaction rate found a floor.
The Berkeley Lab and UC Davis experiment removed almost all of Thunderbird's collision energy. Thin palladium and titanium foils separated an electrochemical cell from an ion-beam chamber. Deuterons entered at centre-of-mass energies between 6.5 and 0.25 keV. Detectors counted the proton and neutron branches independently.
Above roughly 2.5 keV, familiar screening models described much of the change. Below about 2 keV, those curves continued down while the measured yield flattened. At 0.25 keV the enhancement exceeded 10¹⁸ relative to the calculated reaction probability for bare, unscreened nuclei. More than 350 hours of background measurement remained stable. Both reaction branches followed the same energy dependence, and the complete dataset was released for scrutiny. Enhanced nuclear fusion in the sub-keV energy regime The experiment's open dataset
A quintillionfold enhancement sounds like an engine. It is a scientific magnifying glass. The bare-nucleus prediction at that energy is so close to zero that multiplying it by a quintillion still produced fewer than one detected proton an hour at the lowest setting.
This is exactly why the result matters. A countable event has appeared where the conventional extrapolation left almost nothing to count. The task has changed from searching blindly for unexplained heat to measuring a nuclear rate across materials, loading methods, defect populations and particle energies.
An accelerator created the incoming deuterons. The next discovery will come when the metal preserves a measurable part of this enhancement after that beam disappears.
05 — The number 10¹⁸ measures mystery, not proximity to a power plant.
The lowest incident energy in the 2026 experiment was 0.25 keV, or 250 electronvolts. Thermal motion at room temperature is associated with approximately 0.025 electronvolts. The experiment therefore remained around ten thousand times above the room-temperature thermal scale for each reacting pair.
The beam-off control is even more decisive. Electrochemically loaded palladium and titanium produced no excess proton or neutron counts in the measured channels when the ion beam was absent. Whatever caused the plateau, electrochemistry alone did not produce detectable fusion in this apparatus. Beam-off controls in the 2026 study
The fitted screening potential should also be read as an effective description rather than a literal voltage supplied by electrons. At the surface of an irradiated foil, deuterium concentration changes with time. Carbon and oxygen collect. The beam creates defects and drives atoms through the lattice. A 2008 investigation showed how contamination, inhomogeneous loading and deuterium-density dynamics can imitate unusually large screening values. Experimental pitfalls in metal-target fusion
The new work is stronger because it finds the plateau in two reaction channels, two metals, multiple loading states and long background runs. Its next test is also clear: other laboratories must obtain the same absolute rates while independently measuring the deuterium profile and surface composition.
The great numerical enhancement has earned a programme of research. The small absolute rate determines how much work remains.
06 — A newly measured reaction channel widens the search.
The old cold-fusion puzzle was expressed as missing radiation. If the claimed heat came from ordinary D–D fusion, where were the expected neutrons, protons and tritium? One answer is that the heat was not nuclear. A more radical possibility is that very-low-energy fusion can leave through a channel that becomes important only near the reaction threshold.
In 2025, researchers at the University of Szczecin reported high-energy electrons, positrons, 511-keV annihilation photons and bremsstrahlung during beam-driven D–D reactions at energies down to 5 keV. They interpreted the measurements as internal electron–positron pair creation through a threshold state in helium-4. Their inferred partial width for the pair channel was at least ten times the proton channel at the measured low energies. The electron–positron reaction-channel experiment
The paper does not explain historical claims of silent heat. It used an accelerator, detected energetic radiation and did not demonstrate a room-temperature reaction. It does something more useful: it shows that the catalogue of low-energy D–D signatures may be incomplete. Detectors designed only for neutrons could miss the channel that changes fastest as energy falls.
A preceding theoretical study had proposed that a narrow helium-4 threshold resonance could make pair creation important at extremely low energies. The authors also acknowledged the materials problem: the required deuterium movement cannot be assumed freely inside an ordinary lattice. The threshold-resonance model
The 2032 reference experiment will therefore surround its sample with a complete nuclear accounting system. Neutrons, charged particles, gamma rays, annihilation photons, isotopes and heat will be measured together. The reaction will identify itself by conserving energy across all of them.
07 — Three external triggers are already approaching the boundary.
The ion beam is only one way to create a small population of energetic deuterons inside a dense lattice. NASA's lattice-confinement experiments use deuterated erbium exposed to photons energetic enough to split some deuterons. The resulting neutrons and protons transfer energy to neighbouring deuterons, while the metal screens their later collisions. The fuel and lattice remain near room temperature; the particles that fuse do not. NASA's lattice-confinement fusion programme
Laser, acoustic and electrical stimulation offer a more interesting next step. A pulse can concentrate energy into defects, collective vibrations or transient electric fields without maintaining an ion beam. The scientific target is not free energy from a gentle tap. It is a repeatable route by which low-energy input creates a rare energetic configuration within an already dense deuterium population.
The US Department of Energy placed $10 million across eight teams to search for exactly this boundary. The projects include MIT's laser-stimulated metal hydrides, Stanford's phonon-stimulated deuterated nanoparticles, the University of Michigan's precision calorimetry and nuclear-product detection, and Berkeley Lab's sub-500-eV measurements. The programme asked teams to establish a reproducible reference experiment—or determine why prominent claims fail. ARPA-E's low-energy nuclear-reaction awards The eight project descriptions
Europe spent another €5.62 million on the CleanHME collaboration between 2020 and 2025. Its public record contains work on calorimetry, screening and hydrogen–metal systems, including reported anomalous-heat experiments. Those reports have not yet supplied the independently reproduced reference reaction this forecast requires. They have supplied materials, apparatus and investigators prepared to test one. The European CleanHME project record
Cold fusion has moved from scattered claims into a small international instrument-building race. The winning laboratory will be the one that removes the external nuclear trigger without removing the signal.
08 — The six-year path runs from a plateau to a switch.
During 2027 and 2028, laboratories will reproduce the sub-keV plateau with new targets and detectors. They will manufacture palladium, titanium, nickel and complex hydrides with controlled grain boundaries, vacancies and deposited layers. Ion-beam analysis, microscopy and in-situ diffraction will map where deuterium sits while the nuclear counters run. The result will become a property of a defined material rather than of one apparatus.
During 2028 and 2029, the best-performing microstructure will be tested under several forms of stimulation. Beam energy will fall towards tens of electronvolts. Short electrical pulses, lasers and driven lattice vibrations will be compared against quiet controls. Most combinations will produce nothing. One will reveal that the plateau depends on a transient material state that can be created repeatedly.
During 2030 and 2031, the energetic ion source will disappear. A sample will produce a small but unmistakable burst of nuclear products after a low-energy stimulus. The rate may be only a few events a minute. The detector energies will match a known or newly established reaction channel; ordinary hydrogen controls will remain quiet; independent calorimetry will close the chemical ledger.
By 2032, two outside laboratories will reproduce the same protocol. Their absolute rates will vary because defects are difficult to copy. Their on-and-off response, isotope dependence and nuclear spectrum will agree. The phrase cold fusion will return in headlines, while the papers prefer materials-driven fusion or solid-state nuclear reactions.
The timetable is short because the search no longer begins with an undefined warm cathode. Researchers now possess a measured low-energy plateau, open data, multiple nuclear channels and programmes designed around replication. Six years is enough to turn a surprising curve into a controlled switch.
09 — Discovery will arrive decades before useful power.
The first machine built from the effect will count particles rather than sell electricity. A material that increases fusion probability at low voltage could shrink neutron sources used for isotope production, materials analysis, cargo inspection, planetary instruments and some forms of medical imaging and therapy. Berkeley Lab identifies compact neutron generation as the immediate application of its work. Berkeley Lab on materials-driven fusion
Energy demands a different order of achievement. The reaction must accelerate by many additional powers of ten, direct most released energy into recoverable heat, avoid destroying or poisoning its active sites and operate without producing an intolerable neutron or tritium burden. A watt is not a paper. A power station must deliver billions of billions of nuclear events in a controlled stream while every component survives them.
ParallaxSee assigns only 8% probability to net-energy solid-state fusion by 2050 and 3% to commercial electricity from it by that date. Those are secondary estimates, not the published forecast. The 2032 claim concerns discovery: an accelerator-free reaction that other laboratories can turn on and count.
The forecast resolves true if, by 31 December 2032, at least three institutionally independent laboratories have reproduced the same peer-reviewed protocol for a nuclear reaction inside a hydrogen-isotope-loaded solid held below 500°C. The experiment may use chemical, electrical, optical or acoustic stimulation. It may not use an incident ion beam, an external neutron source, photons above 100 keV or a hot fusion plasma. At least one identified nuclear product or isotope must rise at least five standard deviations above a characterised background, track repeated stimulus cycles and disappear under an appropriate ordinary-hydrogen or unloaded-material control. Excess heat alone does not qualify. Net energy is not required.
The causal estimate is explicit: an 80% chance that the sub-keV plateau survives independent reproduction and becomes a controllable material property; a 75% conditional chance that a non-nuclear stimulus can replace the beam by 2032; and an 80% conditional chance that the resulting protocol reaches three independent laboratories. Multiplication yields 48%.
Cold fusion will not be discovered when a metal becomes warm. It will be discovered when a nuclear detector clicks on command, the control remains silent and a second laboratory hears the same rhythm. By 2032, it will.