Marathon Fusion says it has turned mercury into stable gold, and it has not said how much. The San Francisco start-up’s 21 October post on X claimed a rate ten times CERN’s record, a record that stands at 29 picograms. Against that, the Reserve Bank of India holds 880.52 tonnes. The Marathon Fusion gold claim is a physics story with an Indian footnote: nothing in it touches the value of that vault any time soon.
Gold desks reacted anyway. Within hours the question was whether lab-made gold could do to bullion what lab-grown stones did to diamonds. The arithmetic in the Marathon Fusion gold preprint says it cannot. What follows separates the claim from the evidence, then follows the numbers into India’s vaults, households and price screens.
Marathon Fusion Gold: Signatures Shown, Quantity Not Disclosed
On 21 October Marathon Fusion posted gamma-ray spectra that, it says, show mercury-197 and gold-197 after irradiation with fusion neutrons. It described the result as ten times CERN’s rate. The company has not disclosed how much gold it made, and press coverage notes the result has not been independently validated. As it stands, this is a company claim, not an established result.
The physics behind it is conventional. A deuterium-tritium plasma releases neutrons of 14 MeV (million electron-volts). When one strikes mercury-198 in a reactor’s breeding blanket, the layer that makes tritium fuel, it can knock out two neutrons and leave mercury-197. That isotope decays with a half-life of about 64 hours into gold-197, the only stable form of the metal.
The founders set the Marathon Fusion gold idea out in a preprint posted to arXiv on 17 July 2025. Their simulations point to roughly 2 tonnes of gold per gigawatt of thermal power a year. The same paper says fresh gold would carry radioactive isotopes and need 7 to 17 years of storage before sale, a wait that shortens if the mercury feedstock is purified first. Press reports add that Marathon’s own reactor is rated at 1 kilowatt, a millionth of a gigawatt.
“We’ve reduced alchemy to a series of difficult but tractable engineering problems,” the company wrote. The word to hold on to is “difficult”.
CERN’s 29 Picograms Is the Yardstick, and Ten Times That Is Still Dust
CERN’s ALICE collaboration reported in May 2025 that near-miss collisions of lead nuclei at the Large Hadron Collider made about 86 billion gold nuclei during Run 2, from 2015 to 2018. Their combined mass was 29 picograms. CERN notes that such nuclei survive a tiny fraction of a second before fragmenting in the beam pipe.
If Marathon’s “ten times” meant ten times CERN’s mass, the haul would be about 290 picograms, roughly three ten-billionths of a gram. That reading is TES arithmetic, because the company has published no figure. Marathon Fusion gold is said to be stable gold-197, which is why the claim deserves testing, and why a disclosed mass is the first thing physicists will ask for.
Every earlier demonstration has produced atoms, not ingots. Glenn Seaborg’s team at Berkeley turned several thousand bismuth atoms into gold in 1980. CERN’s collider has done it by the billion. Neither came near a gram.
Lab-Grown Diamonds Do Not Predict Lab-Made Gold: 2 Tonnes Against 3,672
A diamond is carbon in a different lattice, which factories can arrange with heat, pressure and plasma. Gold is an element, and making it means changing a nucleus. On the Marathon Fusion gold figure of about 2 tonnes per thermal gigawatt a year, matching the 3,672 tonnes miners produced in 2025 would take roughly 1,800 such plants.
The World Gold Council (WGC) puts the above-ground stock at 220,700 tonnes at the end of 2025, so a year of mining adds about 1.7 per cent to it. Fifty plants making Marathon Fusion gold would add around 100 tonnes, under 3 per cent of one year’s mine output (TES arithmetic). Each tonne would also wait years in shielded storage before it could be sold.
The Marathon Fusion gold preprint does make an economic case, though not against bullion. It argues that gold revenue could double a fusion plant’s income. That is an argument for financing fusion reactors, and it needs reactors that do not yet exist at that scale.
India’s Gold Exposure Sits in the RBI’s 880.52 Tonnes and the Household Vault
The Reserve Bank of India’s annual report records 880.52 tonnes of gold at 31 March 2026, up 0.94 tonnes on the year. About 680 tonnes, or 77 per cent, is stored in India, against 38 per cent in March 2023. Gold stood at 16.70 per cent of forex reserves on 31 March 2026, according to the Press Information Bureau (PIB).
At about ₹1.47 lakh per 10 grams, the 20 October rate, that stock is worth roughly ₹13 lakh crore (TES arithmetic, not an RBI valuation). One plant-year of Marathon Fusion gold, at 2 tonnes, would be worth about ₹2,900 crore. TES has tracked the RBI’s gold repatriation and the way gold now accounts for about 17 per cent of forex reserves in earlier reports.
Households move far more metal. WGC data show India’s gold demand at 710.9 tonnes in 2025, worth ₹7,51,490 crore. In April-June 2026 demand fell 6 per cent to 131.4 tonnes while its value reached a record ₹1,98,100 crore. The WGC has said the rise in import duty from 6 per cent to 15 per cent could trim 50 to 60 tonnes from 2026 demand, while record household recycling has added supply from old jewellery.
For Indian buyers and policymakers, the levers on the gold bill are the price, the customs duty and the RBI’s reserve strategy. Fusion is not among them this decade.
India also has its own yardstick for what gold costs the ground. Kolar Gold Fields in Karnataka yielded roughly 800 tonnes between 1880 and 2001, and its Champion Reef shaft reached about 3.2 km below the surface before Bharat Gold Mines Limited shut the mines in 2001. The RBI’s vault now holds more than Kolar ever produced. At 2 tonnes a year, one Marathon plant would need around 400 years to match it.
The gap is acknowledged in policy. Estimates put India’s mine output below two tonnes a year, with imports and recycled gold meeting the rest of demand. Union Mines Minister Pralhad Joshi directed the Mineral Exploration Corporation Limited (MECL) to begin exploration at Kolar in September 2020, and the old shafts remain closed.
Gold Is Slipping for Reasons That Predate Marathon’s Post
Retail 24-carat gold in India quoted about ₹1.47 lakh per 10 grams on 20 October, a day before Marathon’s post, against about ₹1.60 lakh in late August, according to published rate tables. Spot gold was near $4,124 an ounce on 20 October, per Kitco. The slide began before the lab claim, so any causal link remains unproven.
TES noted in July that gold was trading roughly 30 per cent below its January record. Official buying explains why the floor has held. The WGC reports that central banks reported 170 tonnes of purchases through August, after heavy first-quarter selling by Turkey, Russia and Azerbaijan. For all of 2025 the figure was 863 tonnes, and in the first seven months of 2026 purchases ran near 130 tonnes against about 160 a year earlier.
Poland leads with 98 tonnes bought this year and 648 tonnes in total, closing on its 700-tonne target. The People’s Bank of China added 20 tonnes in August, its 22nd straight month of buying, and press reports of a further 21 tonnes in September would make it 23. The wider case for central bank gold buying is in TES’s earlier analysis.
The nearer risks for Indian investors are the ones already on the screen: price direction, duty policy and the pace of official buying. Marathon Fusion gold, in an unstated amount, is not on that list.
What to Watch Before Lab-Made Gold Becomes a Market Story
Is the Marathon Fusion gold result independently verified?
No. Press coverage of the 21 October post reports no disclosed quantity and no independent validation. Peer review of the preprint was also pending when that coverage appeared.
Could Marathon Fusion gold crash prices the way lab-grown diamonds did?
Not on the preprint’s figures. About 2 tonnes per thermal gigawatt a year, a 7 to 17 year storage wait and 3,672 tonnes of annual mine output leave no route to a price crash.
What would change this assessment?
Three markers: a disclosed mass of gold, an independent replication, and a peer-reviewed yield that includes the cost of enriched mercury. The WGC’s next Gold Demand Trends report will also show whether Indian demand follows the price down.
Until a mass is disclosed and replicated, the Marathon Fusion gold claim belongs in the physics column. India’s gold column will be written by the rupee price of the metal, by duty policy, and by how many of the RBI’s 880.52 tonnes stay at home. Kolar’s miners went 3.2 km down for their tonnes. The next test is whether anyone can show a gram.
Disclaimer: This article is analysis for information only and is not investment advice. Gold prices move with global and domestic factors, and readers should consult a registered adviser before investing.
Source note: Marathon Fusion’s arXiv preprint and public posts, CERN’s ALICE collaboration, the World Gold Council, the RBI Annual Report 2025-26, PIB, and published press reports. Figures labelled TES arithmetic are TES calculations.
