Open ScoutGold sits around $4,127 an ounce right now, close to the highest it has ever been, up more than 25% since the start of 2025, because central banks keep buying and people are nervous about debt and currencies. Then Marathon Fusion put out a paper saying a fusion reactor could make gold out of mercury, and in October the company claimed on X that it had done it at ten times the rate CERN ever managed. So the question going around is simple. If you can print gold, does gold survive?
To answer it you have to know what these companies actually do and run the numbers. Here is the real version, Marathon Fusion on the energy side and AstroForge on the asteroid side, and then what the economics imply.
What Marathon Fusion actually does
The boring part is the real company, so start there.
A fusion reactor running on deuterium and tritium (two heavy forms of hydrogen) has a fuel problem. Tritium is radioactive, barely exists in nature, and costs around $30,000 a gram. A plant has to breed its own by wrapping the reactor in a layer called a breeding blanket, usually full of lithium, so the neutrons flying out of the reaction hit the lithium and make fresh tritium. The problem is that most of the tritium you feed in never fuses. It comes straight back out in the exhaust, and if you cannot catch and recycle it fast, you have to keep a huge, expensive, dangerous amount on site just to keep the machine running.
Marathon’s main product attacks exactly that. They built a plasma centrifuge that spins charged gas at supersonic speed using crossed electric and magnetic fields, which lets them pull tritium selectively out of the exhaust and feed it back in. They say it cuts the tritium a plant has to hold, and the flow it has to handle, by a factor of ten or more. The same machine enriches lithium-6, the isotope the blanket needs, fast enough to supply a new gigawatt plant’s blanket every two years. Dennis Whyte, who runs nuclear science at MIT, called the separation results impressive and important.
This part is sellable today. Every company building a D-T fusion plant has the tritium problem, so Marathon holds letters of intent from Commonwealth Fusion Systems and Helion, the two biggest developers, and it is building a pilot fuel-processing and isotope-separation factory it wants running by 2028, commercial by 2029. The money behind it is still modest, about $18 million including a $3.63 million ARPA-E grant. Marathon never has to build a reactor. It sells the plumbing to everyone who does.
The gold, and how the physics works
The gold is a side idea bolted onto that blanket, and the mechanism is real.
A D-T reaction throws out a neutron carrying 14 million electron-volts, which is a lot of energy. Marathon’s move is to put mercury-198 in the blanket next to the lithium. When one of those fast neutrons hits a mercury-198 atom, it can knock two neutrons loose instead of one. That reaction, called (n,2n), does two jobs at once. It multiplies neutrons, which actually helps the plant breed its tritium, and it turns mercury-198 into mercury-197. Mercury-197 is unstable and decays in about 64 hours into gold-197, the only stable form of gold there is. So the same neutrons the plant already makes, and needs, throw off gold as a byproduct without stealing from the power output. That is why serious people read the paper instead of laughing at it.
The numbers on paper are large. The simulation shows a 1.5 gigawatt (thermal) reactor making about 2,950 kilograms of gold a year, roughly 2 tonnes per gigawatt. At $4,127 an ounce, 2,950 kilograms is about $390 million of gold a year from one plant, on top of the electricity. Marathon’s own, more conservative figures take a plant from $223 million a year in revenue to $482 million. Either way you roughly double what a fusion plant earns. For an industry whose whole problem is that the economics do not close, that is a real claim.
Then the full picture. The blanket needs 622 tonnes of mercury-198 enriched to about 90%, and mercury-198 is only about 10% of natural mercury, so you have to enrich it, and the paper admits scalable mercury enrichment still needs to be fully defined. Enriched mercury-198 today exists only in research quantities. The gold also comes out radioactive, because stray mercury isotopes make unstable gold alongside the good stuff. It takes about 6.8 years before it counts as low-level waste, 13.7 years before it loses the radioactive label, and 17.7 years before, in the paper’s words, it is less radioactive than a banana. Marathon’s answer is that you could sell it forward on paper while it cools, which is fair, but you are still warehousing bars for over a decade.
And all of it assumes a commercial gigawatt fusion plant running for years. None exists. Marathon’s own reactor in San Francisco is rated at 1 kilowatt, a million times smaller than the one in the simulation. The claim of beating CERN deserves context too. CERN’s collider made about 29 picograms of gold over three years, a rate of roughly 100 billion years per gram, so ten times that is still almost nothing. The comparison that matters is a gold mine, and matching one year of global mine output would take about 1.65 terawatts of fusion, several times the entire planet’s current nuclear fleet.
So the gold is real physics, genuinely clever, and more than a decade from mattering, sitting on top of a fuel-cycle business that earns money now.
What AstroForge actually does
AstroForge wants to mine metal asteroids for platinum-group metals and sell them back on Earth. Platinum, palladium and the rest of that family go into catalytic converters and hydrogen fuel cells and little else, and they cost what they cost because they are rare in the Earth’s crust. They are not rare in M-type asteroids, which are basically leftover metal cores, and some are dense with exactly these metals.
The plan is to bring the refinery to the rock. The spacecraft is designed to vaporize asteroid material and sort it into separate elements in orbit, using the metals’ magnetic behavior to pull the valuable fraction out, then return only the refined metal. Founded in January 2022 by Matt Gialich and Jose Acain, it has raised around $55 million.
It has flown twice and failed twice, for reasons worth knowing because neither was about the core idea. Brokkr-1, in April 2023, was a refinery test in Earth orbit on a chunk of asteroid-like rock. It never ran, because the satellite had deployment and solar-panel trouble, the refinery’s own magnetic field interfered with the spacecraft’s orientation, and it got lost among about 50 other rideshare satellites. Odin, also called Brokkr-2, launched in February 2025 to fly by a near-Earth asteroid called 2022 OB5 and check whether it is actually metallic. The team rebuilt it in-house in under a year for around $3.5 million after the original failed vibration testing, then lost it to ground-station failures and never established control. The next one, DeepSpace-2, flies on Intuitive Machines’ IM-3 in the second half of 2026 and aims to return to that metallic asteroid and land on it.
The unit economics, on paper, are the appealing part. Platinum is about $2,150 an ounce. A tonne is roughly 32,150 ounces, so one to two tonnes returned is about $70 million to $140 million of metal, which is exactly the range AstroForge quotes. The company wants missions under $10 million each. Ten million in and seventy million out is a very good trade. The reason nobody has it yet is that every hard step, grabbing material off a spinning rock in microgravity, refining it in space, getting it home, surviving reentry, has never been done, and AstroForge has not yet reached an asteroid at all. Its real edge so far is speed and cost. It builds spacecraft in-house in months for single-digit millions and treats failures as cheap lessons, which is how it has flown three times while richer predecessors like Planetary Resources went under.
You cannot crash a fear asset by making more of it.
So will gold crash
This is where the question actually lives, and it comes down to one difference. Platinum and gold look alike and behave nothing alike.
Platinum is an industrial metal. The market is thin, around 180 tonnes mined a year, worth maybe $12 to $13 billion, and supply moves the price. A few tonnes a year from space is a percent or two of supply, survivable. To pay back a decade of R&D and a fleet of missions you would have to return far more than that, and then you are a real share of the market and the price falls. That is what the main academic study found: returning platinum to Earth pencils out only under unlikely conditions, because the market reacts to the supply you add. The more you bring back, the less it is worth.
Gold works the other way. Only a small share of it gets consumed by industry. Most of it is held, as bars, reserves and jewelry, and the price is set by fear, interest rates, the dollar and central-bank buying, not by how much is dug up. There are already around 200,000 tonnes above ground, and mines add about 3,000 tonnes a year, so new supply is already only about 1.5% of the pile. Even the fantasy fusion build-out, 1.65 terawatts of reactors, would trickle gold into that enormous stock, and the gold would show up radioactive and locked away for a decade. So manufactured gold does not crash the gold price on any timeline worth planning around. People have been trying to make gold for 2,600 years, and the people who buy it have never cared whether it was easy to make.
Two metals, opposite machines
| Gold | Platinum | |
|---|---|---|
| Price now | ~$4,127/oz | ~$2,150/oz |
| Mined per year | ~3,000 t | ~180 t |
| Already above ground | ~200,000 t | a sliver by comparison |
| Priced on | fear, rates, the dollar, central banks | industrial supply and demand |
| Moves if supply jumps? | barely | yes, sharply |
| Who is aiming at it | fusion (Marathon) | asteroids (AstroForge) |
There is a pattern across both sides. The thing that is valuable because it is scarce is the hardest thing to get rich from by making it common, so the sharper companies are already selling something else. Marathon’s real business is the fuel cycle. Karman+ raised $20 million to mine water from asteroids and sell it in space as propellant, where it is the only supplier and there is no Earth price to collapse. TransAstra earns money today selling orbit-tracking software, and its founder says plainly that shipping platinum back to Earth will not pay for a long time. Interlune signed a $300 million-plus deal for lunar helium-3 before mining a gram. The money is in the picks and shovels, and in selling where you are the only seller.
What to actually watch
None of this is a reason to write any of it off. The physics is sound, the engineering is moving, the people are serious. Track it by milestones instead of headlines.
- Marathon’s pilot fuel plant, 2028. If it runs and a real fusion developer buys the output, Marathon wins whether or not a single gram of gold ever gets made.
- The first independently verified gram of transmuted gold. Until the data is published, the alchemy claim is a claim.
- AstroForge’s DeepSpace-2, second half of 2026. Landing on the asteroid would be the first time a private company has touched one, and that happens before any mining.
- The first kilogram of asteroid platinum returned to Earth. That is the moment the whole field stops being theoretical.
The gold and the platinum are the least interesting part of what either company is building. The machines underneath are the story.
Sources
Gold and platinum prices: Fortune (gold, Oct 21 2026), Fortune (platinum).
Marathon Fusion fuel cycle: NEI Magazine. Transmutation paper, Rutkowski, Harter & Parisi: arXiv:2507.13461. Gold rate, enrichment, radioactive cooling and revenue figures: The Oregon Group.
AstroForge missions and funding: Wikipedia, dot.LA. Asteroid mining economics, Hein, Matheson & Fries: arXiv:1810.03836. Karman+, TransAstra, Interlune: The Space Review, AVEVA, The Oregon Group.