Seven caveats in the claim about turning mercury into gold with nuclear fusion

A San Francisco-based company named Marathon Fusion announced on October 21 that it had ‘converted’ mercury into gold in an experiment in September and that it had achieved a production rate tenfold greater than a record reported by CERN in 2025.

The company has also proposed a programme in which mercury-198 is exposed to high-energy neutrons produced by nuclear fusion — in an existing power plant — converting it into radioactive mercury-197, which subsequently decays into stable gold-197. In a preprint paper uploaded to the arXiv repository in July 2025, Marathon scientists had predicted this process could produce roughly 2 tonnes of gold annually for every gigawatt (GW) of thermal fusion power.

This said, the company has not said if its recent claim has been independently verified — there are no public records indicating that — nor has it disclosed how much gold it actually produced. Overall, there are at least seven caveats around the claims and the hype that has erupted around it.

1. The process involves nuclear fusion tangentially

In nuclear fusion, light nuclei, such as deuterium and tritium, are forced together, releasing high-energy neutrons and energy. The proposal is to have these neutrons strike mercury nuclei and eventually produce gold nuclei. The latter is called nuclear transmutation, not nuclear fusion. Neither the gold nor the mercury nuclei undergo any fusion of their own. Nuclear transmutation is well-known in nuclear physics. In other words the claim about “using” fusion to turn mercury into gold can be misleading.

2. The claim is not backed by enough evidence

On its website, Marathon Fusion has published data showing the energy signatures associated with mercury-197 and its decay into gold-197. It is not possible to infer how much gold was produced from this information, nor its purity or how efficiently the process works. Marathon has also not placed a detailed experimental methodology or independent verification in the public domain. Its prior research, described in the preprint paper, was based on computer simulations.

3. The claim about beating CERN’s record is misleading

In 2025, CERN had reported that interactions between lead ions at the Large Hadron Collider had produced around 86 billion gold nuclei from 2015 to 2018. This is just around 0.029 billionths of gram — about the amount of gold you can buy with 0.0000414 paise. These nuclei existed for mere fractions of a second before disintegrating further. CERN was also just conducting particle-physics research, not attempting to manufacture gold on a commercial scale. And Marathon Fusion has not explained exactly how it calculated its tenfold improvement.

4. There is a gap in scale

In the 2025 preprint paper, the authors estimated that a suitable nuclear fusion plant could produce around 2,000 kg of gold every year per gigawatt of thermal power. The figure is rooted in certain advanced computer simulations of neutron transport and nuclear reactions. It is not based on the workings of a functional, commercial reactor. Marathon Fusion has also described plans for a reactor with a fusion power target of just 1 kW, which is one-million-times smaller than 1 GW. Long story short, even if Marathon Fusion’s claimed experiment in September demonstrated converting mercury to gold, it cannot be used to validate the company’s projections of producing gold on an industrial scale.

5. The proposed process requires lots of mercury

Only around 10% of the mercury that occurs naturally is mercury-198, which is the isotope Marathon Fusion proposes to use. However, its computer simulations assumed mercury enriched to 90% mercury-198. This will require a programme to enrich mercury isotopes on an industrial scale. Second, mercury is extremely toxic, so the facilities will have to have specialised containment materials and protocols and environmental safeguards. Marathon Fusion has acknowledged these difficulties and has proposed technological solutions but their practicality and costs have yet to be determined at scale.

6. The gold could be radioactive for years

The Marathon Fusion way to make gold is not implausible but the resulting gold will be radioactive. Other nuclear reactions happening at the same time will also create other radioactive isotopes, including gold-195, which has a half-life of around 186 days. And according to Marathon’s own simulations, the resulting gold mix will need around 6.8 years to meet one low-level radioactive waste threshold, 13.7 years to fall below a specified radioactive material labelling threshold, and 17.7 years to meet a more conservative benchmark. It will be possible to purify the isotopes to reduce these timelines but not without adding to the costs.

7. The claim about power plants’ revenue is speculative

The Marathon Fusion website says producing gold in this way could roughly double the revenue of fusion power plants without compromising electricity output. To this end, the preprint paper modelled a 1.5-GW thermal fusion power plant that cost $3 billion to build and which operates at 100% capacity for three decades while converting 40% of its thermal output to electricity. In this scenario the plant will generate around $263 million a year from electricity sold at $50 per megawatt-hour. And producing three tonnes of gold a year and selling it at $100,000 per kilogram would add $300 million in revenue. The problem? Commercial fusion power does not yet exist and these figures are all based on expectations of the future.