Alchemy Is Real: Fusion Startup Turns Mercury Into Gold
Marathon Fusion's October 21 thread on X contains one piece of experimental data: a gamma-ray spectrum that labels mercury-197 and gold-197 signatures measured between 8 and 44 hours after irradiation. The company says the result shows it turned mercury into stable gold in September at ten times the record CERN set in 2025. It has not published how much gold was produced, how pure the gold was, or what the run cost. Those gaps matter because the claim has already moved into discussion of gold prices and fusion economics.
What the thread says
The post makes three statements. In September the company produced stable gold from mercury using fusion neutrons. A week before the post, it enriched isotopes in what it calls the world's most powerful plasma centrifuge, beating a Soviet record that stood for 40 years. And it is now building a fusion reactor in San Francisco to reach record power density.
The thread drew more than 2.4 million views in under a day, and the only experimental evidence in it is the single spectrum.
The timing in that spectrum fits the reaction Marathon described in a 2025 preprint. A fast neutron from deuterium-tritium fusion strikes a mercury-198 nucleus, two neutrons come out, and the resulting mercury-197 decays with a half-life of about 64 hours into gold-197, the only stable isotope of gold. What the spectrum cannot show is the quantity of gold, its purity, or whether the process scales. Glitchwire reports that the company has not disclosed the amount produced, the isotope purity, the recovery efficiency, or the cost, and Runtime Wire reports that the claim has not been independently verified in public reporting.
The CERN comparison
CERN's ALICE experiment reported in May 2025 that near-miss lead collisions at the Large Hadron Collider produced about 86 billion gold nuclei between 2015 and 2018. That amounts to roughly 29 picograms, with a peak rate near 89,000 nuclei per second. Each nucleus broke apart within a fraction of a second. ALICE deputy spokesperson Kai Schweda has said that making a single gram this way would take the collider around 100 billion years.
Marathon's phrase "high-rate alchemy" suggests the tenfold figure compares production rates, but the company has not said whether it means the peak rate or the total yield. The French outlet Cryptoast calculated that applying the factor to total mass would give about 290 picograms. That arithmetic is the outlet's, and Marathon has not stated it. At that scale the quantity is still measured in picograms, and the thread does not say which CERN number is the baseline.
The economics in the 2025 paper
The 2025 preprint, which has not been peer reviewed, models a fusion blanket lined with mercury-198. In its simulations a 1.5 gigawatt thermal tokamak produced 2,953 kilograms of gold a year, which the company markets as about 5,000 kilograms per gigawatt of electricity. Marathon's alchemy page puts annual revenue for a gigawatt-scale plant at $482 million with gold, against $223 million without it. The plant case assumes a $3 billion reactor and a gold price of $100,000 per kilogram.
The paper models revenue and publishes no production cost. Mercury-198 makes up 9.97 percent of natural mercury, and the reference design needs 622 tonnes of it enriched to about 90 percent. The paper estimates the energy cost of one separation stage at about $2.4 per kilogram but concedes that the details of scalable mercury isotope separation still need to be defined. Karl Lackner, emeritus director at the Max Planck Institute for Plasma Physics, described the enrichment as technically demanding.
Radioactivity adds another delay. Traces of mercury-196 produce radioactive gold-195, so the paper says the gold would need about 6.8 years to reach the least hazardous class of low-level waste, 13.7 years before it no longer needs a radioactive label, and 17.7 years to be less radioactive than a banana. Co-founder Adam Rutkowski told the Financial Times in 2025 that storage of 14 to 18 years may be required. Lackner also said mercury-based blankets probably would not appear in the first generation of fusion reactors, and he estimated that all the mercury in the Earth's crust would yield about 600 tonnes of gold a year.
On market scale, mines produced a record 3,672 tonnes in 2025. At two tonnes per gigawatt thermal, the paper itself concedes that 1.65 terawatts of fusion capacity would be needed to match 2024 mine output. Spot gold traded at $4,081.89 an ounce on October 22, against an average miner all-in sustaining cost of about $1,930 an ounce in the second quarter, according to World Gold Council data cited by The Oregon Group. No fusion plant exists yet, and the earliest grid targets the World Nuclear Association lists are Helion's Orion near 2028 and Commonwealth Fusion's ARC in the early 2030s.
The plasma centrifuge result
The centrifuge claim has more outside support than the gold claim. On August 27, Marathon announced it had enriched hydrogen and lithium isotopes using its plasma centrifuge, with development backed by the Department of Energy's ARPA-E VISION OPEN program. The materials testing lab Covalent certified the lithium-6 enrichment, and Dennis Whyte, an MIT professor of nuclear science and engineering, called the isotope separation results impressive and important for the fusion industry.
The device uses a partially ionized plasma in which charged particles collide with neutral atoms and drag the whole gas into rotation. The company says this lets it run at lower temperatures than earlier plasma centrifuges, which heated up until the separation failed. Marathon has not tested the machine on tritium. Its stated use is a process called differential pumping, which the company says could cut tritium flow rates and the size of the associated processing systems by a factor of ten or more. CEO Kyle Schiller told Heatmap that a first full-scale facility would produce tens of tonnes of lithium-6 a year, enough for a new gigawatt-scale plant roughly every two years. Nuclear Engineering International reports the company is targeting 2029 for that facility.
The Soviet record is harder to pin down. Glitchwire notes that Soviet researchers at facilities such as the Kurchatov Institute published plasma centrifuge work through the 1980s, and that Marathon has not specified which benchmark it surpassed. The thread also says "last week," while the documented announcement dates to late August. The sources reviewed for this article do not say whether the record refers to the August experiment or a newer one.
The San Francisco reactor
Marathon's website lists targets for the San Francisco machine of 1 kilowatt of fusion power, a power density of 20 kilowatts per square meter, and a neutron flux of 10^12 neutrons per square centimeter per second. An RF plasma generates deuterium ions that are accelerated at a target of lithium tritide, which can be regenerated in place. BusinessToday reports that the company described a goal of directing a trillion neutrons every second at a target 1 centimeter across, continuously for years, and of doubling power density every month. The outlet characterizes that rate as a development target without evidence that it can be sustained.
One kilowatt is a millionth of a gigawatt, the scale at which the 2025 paper's gold economics apply. In August, Nuclear Engineering International reported that Marathon was not building a pilot reactor to generate electricity. The San Francisco machine is a neutron source, and the nearer revenue lies in medical isotopes such as copper-64 and copper-67 and in fusion fuel processing. Marathon's investors include Divergent Capital, Liquid 2 Ventures, 1517 Fund, Anglo American, and the Department of Energy.
What to look for next
Six items would settle most of the open questions. The first is the measured mass of gold and the assay method used to measure it. The second is the isotopic purity of the product and the recovery efficiency from the irradiated mercury. The third is a statement of which CERN figure serves as the baseline and whether the tenfold factor refers to rate or total yield. The fourth is replication or review by a laboratory outside the company. The fifth is a production cost per ounce, which the 2025 paper does not provide. The sixth is the specific Soviet benchmark behind the centrifuge claim and the conditions of the run that beat it.
Until those appear, the evidence supports three separate statements. A spectrum consistent with mercury-to-gold transmutation has been shown. A lithium-6 enrichment result has outside laboratory certification. A 1 kilowatt neutron source is under construction with stated targets. For readers evaluating fusion economics or gold supply, the gold claim is a reported result awaiting data, and the isotope business is where the company's nearer-term revenue is expected to come from.
