Can fusion turn mercury into gold? Marathon Fusion’s claim

The Alchemist's Tweet

Eli: On October twenty-first, a San Francisco startup called Marathon Fusion posted on X: 'Last month, we demonstrated high-rate alchemy: turning mercury into real, stable gold, beating CERN's record by a factor of 10.' The tool, the company said, was neutrons made by nuclear fusion. Mercury went in. Stable gold, it said, came out.

Mara: The listener who wrote in asked it plainly: did a startup really just turn mercury into gold, and could fusion make gold cheap one day? It's Monday, October 26th, and today's angle is how fusion companies revive ancient alchemy to subsidize twenty-first-century nuclear energy.

Eli: Marathon didn't leave it at words. It attached proof to the post, and everything since has turned on what that proof can actually bear.

Mara: And that proof wasn't a bar of gold. It was a set of graphs.

Eli: Gamma-ray spectra. Once the mercury had been hit with neutrons, it gave off gamma rays, and Marathon said the readings showed traces of mercury-197 and gold-197. The measurements ran from about eight hours to nearly two days after the irradiation, that hit with neutrons.

Mara: So every radioactive isotope shines at signature energies, and the peaks act as fingerprints.

Eli: Fingerprints the company chose to lead with. Co-founder Kyle Schiller put that spectrum image in the public release.

Mara: The centuries-old dream of mercury into gold, handed to a fusion reactor.

Eli: And riding on those peaks is the bigger question: whether fusion could turn that old dream into something cheap enough to sell. Mara, the recipe behind it had been public for more than a year.

The Nuclear Recipe

Mara: It's a July 2025 preprint, not yet peer reviewed, by three Marathon researchers, among them Adam Rutkowski and Jason Parisi. The subject is nuclear transmutation: changing how many particles sit in an atom's core, so the atom becomes a different element. The starting point is mercury-198. Hit it with a fast neutron from deuterium-tritium fusion, about 14 million electron volts. Two neutrons get knocked out, and it becomes mercury-197.

Eli: One neutron in, two out. Mercury ends up a neutron lighter.

Mara: And it doesn't stay mercury. Mercury-197 is unstable, with a half-life of about 64 hours. It decays by electron capture: one proton inside turns into a neutron. One proton fewer, and the atom is gold. Gold-197, the only stable kind there is.

Eli: Stable, but not ready. The preprint's own cooldown estimates run long. About seven years to count as the mildest grade of low-level waste. Nearly eighteen to be 'less radioactive than a banana' per troy ounce. Holding newly made gold in radioactive quarantine for nearly two decades means financing and storing it the whole time, a heavy handicap against gold from a mine.

Mara: Which puts all the weight on where that mercury sits.

Eli: In the reactor's walls. A fusion plant would carry a breeding blanket, a layer around the core that soaks up neutrons and uses them to make fresh tritium fuel. Marathon's paper proposes a blanket of liquid mercury. The mercury reactions multiply neutrons, so the plant can still breed its fuel while gold builds up on the side. Marathon says it needn't cost the plant its fuel supply or its power.

Mara: And it needs one particular mercury.

Eli: Nature doesn't supply much of it. Mercury-198 is only about a tenth of natural mercury, and the modeled design wants it refined until nine atoms in ten are the right kind. Karl Lackner of the Max Planck Institute for Plasma Physics said using mercury this way 'probably wouldn't be done in the first generation of fusion reactors.' Enriching it, he said, is 'a technically demanding process.'

Mara: Hundreds of tonnes of refined isotope, before a single flake of gold.

Eli: On paper, the flakes add up. Marathon estimated a one-gigawatt plant could turn out roughly two tonnes of gold a year. Those numbers came from computer simulations, a digital model of the reactor built with a code called OpenMC. But a model like that needs checking against a real commercial fusion reactor, and none exists yet. So the leap from a 64-hour decay to tonnes of sellable gold runs through a plant nobody has built, and before that, through the much smaller machine Marathon actually has.

The CERN Benchmark

Mara: That smaller machine is being built in San Francisco. Marathon lists a one-kilowatt fusion system designed to make neutrons, not electricity. It runs on beam-target fusion: a radio-frequency plasma produces deuterium ions, they're accelerated into a target, and that throws off neutrons. The stated goal is a trillion neutrons every second, all poured onto a spot about the width of a fingertip, without a break for years. And the company says it has already used fusion neutrons on zinc, making radioactive copper for diagnosing and treating cancer.

Eli: One kilowatt, against the gigawatt plant in those simulations. A millionfold gap.

Mara: That's the hardware Marathon has to work with. And in October, Marathon claimed a gold-production rate ten times higher than CERN's benchmark. Except it hasn't defined what that ten-times actually measures. And CERN's figures came from a completely different process.

Eli: A collider, of all places. In 2025, CERN's ALICE experiment reported that lead nuclei, flying past each other at nearly the speed of light, sometimes lose three protons in a near miss. What's left is gold. Over three years of running, the Large Hadron Collider's four main experiments made billions of those gold nuclei. All together, they weighed 29 picograms, each a trillionth of a gram. A speck far too small to see.

Mara: And that's with ALICE at full tilt, peaking around 89,000 gold nuclei a second. Still invisible.

Eli: So even if Marathon's tenfold holds up under replication, ten times CERN over a lab run is still picograms nobody could see, not gold anyone could extract and sell.

Mara: Which changes how that October post reads. It drew six and a half million impressions on X. What's still unknown is how much metal those peaks actually stand for. Eli, the whole claim now rests on what the spectra can show.

The Missing Grams

Eli: So, those October graphs again, with all of that in view. Reporters covering the claim noted Marathon hadn't said how much gold it actually made. Nor has it publicly established the yield, the purity, how much it can recover, or what the process costs. The claim hasn't been independently verified in published peer review. And the tonnes are simulation output. In everything put out so far, there's no weighed flake of gold, and without those figures there's nothing to set against the cost of mined gold.

Mara: A fingerprint, then, with no weight attached.

Eli: And it's read two ways. Kyle Schiller and Marathon say the neutrons triggered real nuclear reactions that made new gold atoms, and the spectra are their evidence.

Mara: Eli, the other reading belongs to independent physicists and commentators: without yield data or replication, the claim is unverified. One outlet put it plainly: 'That does not make the claim false. It makes it unverified.'

Eli: Then there's purity. Natural mercury isn't only mercury-198, and commenters pointed out the other isotopes would likely make unstable gold mixed in with the stable kind. That's a purification problem Marathon hasn't addressed publicly. The worst one is gold-195, made from trace mercury-196. Its half-life, the time it takes for half its atoms to decay away, is about 186 days.

Mara: That's the worst offender behind the preprint's banana clock. And lab gold has always come with a catch. In 1980, at Lawrence Berkeley National Laboratory, Glenn Seaborg's team turned bismuth into gold. Scientific American reported Seaborg's estimate that it 'would cost more than one quadrillion dollars per ounce.' Gold then sold for about 560 dollars.

Eli: A bill no amount of gold could ever pay back.

Mara: And the pattern holds. Demonstrated methods make microscopic amounts of gold at costs far above its market value. Even Marathon's own site says no economically attractive approach has yet been identified. Miners measure themselves by all-in sustaining cost, the full per-ounce bill to dig, process and keep a mine running, and that's the bar lab gold would have to clear. Then there's scale. No commercial fusion plant feeds the grid today. And Karl Lackner estimated that using all the mercury in Earth's crust would make only about 600 tonnes of gold a year.

Eli: About a fifth of yearly mining. Gold 'would therefore not become worthless,' Lackner said.

Mara: Marathon's team reads the economics another way. Its website estimates adding gold would more than double what a gigawatt plant earns from electricity alone, to about 482 million dollars a year. The team says that extra revenue, if achieved at scale, could make fusion more financially attractive. Adam Rutkowski, Marathon's chief technology officer, described using these fast neutron reactions 'to make really large quantities of gold while satisfying the fuel cycle requirements of the system.'

Eli: From Seaborg's bismuth to Marathon's mercury, making gold keeps proving a remarkable piece of physics, and keeps falling short of being a supply of metal.

The Real Currency

Mara: Which leaves the puzzle of why a fusion company would bother with gold at all. Aakash Gupta, writing about the announcement, put it sharply: 'their own math says gold is one of the cheapest things their machine can make. Aim it at a different metal and it makes a cancer drug ingredient worth tens of billions of dollars a gram.'

Eli: Medical radioisotopes: short-lived radioactive atoms for imaging and treating cancer. Like that copper from zinc.

Mara: And Marathon has announced no plans to commercialise the gold. That 482-million-dollar revenue line is still a model. Even if the physics checks out, one assessment says scaling it to make financial sense would be extremely difficult, and unlikely within the next two decades. Gold prices would only feel real pressure if the process became commercially viable and added significant new supply, and that would take decades.

Eli: So the answer the listener asked for: Marathon says it turned mercury into gold, and that hasn't been verified. What it's published isn't enough to compare its gold with mined gold, and the assessments put any market impact decades away.

Mara: The signals to watch start with peer-reviewed mass measurements, with contamination controls, showing gold-197 really formed. Then yield, purity and recovery figures from independent labs. And further out, a licensed gigawatt fusion plant on the grid with a working breeding blanket.

Eli: For listeners wanting to explore the basics, read up on nuclear transmutation.

Mara: Follow Science for more stories as researchers test whether lab-made elements can become practical.

Eli: That's our story for today. I'm Eli.

Mara: And I'm Mara. Thanks for listening to Angle, and until next time.