What the Belt Sells

The four-part test

Any Belt Export Has to Clear Four Conditions at Once

Most proposals clear one or two and are argued as though clearing one were enough.

Mars has a make-or-import ladder: for any given item, either you manufacture it on the surface or you fly it from Earth, and the crossover is set by mass, cadence, and how many people are waiting for it. The belt does not get that ladder, because the belt is not a settlement with a shopping list. It is a set of industrial sites scattered across an archipelago, and the question is not what it makes for itself but what it can sell to anyone at all.

Four conditions have to hold together. A buyer has to exist. That buyer has to be closer to the belt than to any substitute — not closer in kilometers, which are irrelevant, but closer in Δv and in time. The value has to survive a 30.8-month response lag, made of a 15.3-month wait for the Earth–Ceres launch window plus a 15.5-month transit, because nothing in the belt can react to a market signal faster than that. And the revenue has to amortize a facility built 15.5 months from its nearest spare part.

Four candidate identities have been proposed for the belt, and they are worth taking one at a time rather than arguing about the belt in general. The mine sells raw material to Earth. The foundry sells finished parts to whoever is building in space. The fuel dock sells propellant to traffic passing through. The fourth proposal is that Ceres becomes the place where registry, title, insurance, and assay get settled — a financial center rather than an industrial one. Modeled

The financial-center proposal fails at the first column and is retired here rather than developed. Registry, title, insurance, and assay are functions that run wherever the counterparties and the courts are, and every counterparty is on Earth. Nothing about being 2.77 AU away helps settle a contract; the 14.7-to-31.4-minute one-way light time actively hurts. There is also no crewed presence at Ceres in any near-term scenario, and no reason a legal function would need one if there were.

The mine

The Mine Fails on the Shape of the Demand Curve

Not on freight cost, which is the objection most often raised and the wrong one.

Selling raw material back to Earth splits into two cases, and both close. Bulk metals — iron, nickel, silicon, aluminum — are worth a few dollars a kilogram on Earth and cannot pay for a 15.5-month return trip plus an atmospheric entry at 12.84 km/s. Nobody seriously argues otherwise.

High-value metals are the real proposal, and platinum-group metals are the whole of it. Those do have the value density to justify a return leg, which is why the thesis keeps attracting funding. The failure is elsewhere: the entire world platinum-group market is about 400 tonnes a year and roughly $14 billion, and it does not get larger when the price falls. An operation big enough to matter collapses the price of its own product. The argument is developed in full on Platinum-Group Metals & the Demand Curve, including the iridium case — the one genuine physical bottleneck the belt ever had, which terrestrial engineers dissolved in public in November 2025 by cutting electrolyzer loadings roughly tenfold.

The general rule that falls out of it applies to every future commodity thesis, not just this one. A belt operation cannot respond to a market signal in less than about 30 months, and a terrestrial laboratory can requalify a catalyst or substitute a material in one to three years. Any argument of the form "Earth will run short of X, so mine X in the belt" is a bet that nobody on Earth solves the shortage during the two or three decades it takes to build the supply. That bet has now lost twice in public, on rhodium after 2021 and on iridium by 2025.

One weakened version survives. A buyer worried about supply concentration — South Africa for platinum, rhodium, and iridium; Russia for much of the palladium; China as both the largest consumer and a dominant refiner — is buying insurance rather than metal, and insurance is priced on risk rather than on tonnage. Whether anyone pays interplanetary prices for that insurance has never been tested. Projected

The foundry

The Foundry Works, and Its Only Customer Is the Belt Itself

Import substitution rather than export — which is worth a great deal and is not a business.

The physical case for making things in the belt is the strongest of the four, and it is specific rather than general. Nickel-iron asteroid material is pre-reduced ore: the carbothermic reduction that a terrestrial blast furnace performs was already done by the solar nebula, so the melt energy is 0.34 to 0.40 kWh/kg ideal and about 0.53 realized — parity with an electric-arc furnace running scrap, not a revolution. The revolution, such as it is, is the carbonyl route. Nickel and carbon monoxide combine at low temperature to make nickel tetracarbonyl, a gas, which decomposes on a heated form to deposit metal. Carbon monoxide is a catalyst rather than a consumable, the carbonaceous asteroids supply the carbon the Moon does not have, and because the product is vapor deposited it comes out as a finished part rather than an ingot. The process is also gravity-independent, which matters at 0.0289 g. This is set out on Nickel-Iron & Carbonyl Metallurgy.

So the belt can make structure, pressure vessels, tankage, radiators, truss, and wire. The question is who buys them. Every existing and near-term customer for space hardware sits in Earth orbit or cislunar space, and for those customers the belt is the most expensive supplier in the solar system: 5.4 km/s and 15.5 months against 3.1 km/s and three days from the lunar surface, or an Earth launch that arrives in eight minutes at a price that keeps falling. Mars supplies itself. There is no third party.

What the foundry actually does is cancel imports. Every kilogram of structure made at Ceres is a kilogram removed from the 30.8-month spares queue, and on a site where the governing scarcity is time rather than money, that is the highest-value thing anyone can do. It just does not generate revenue, and a page about what the belt sells has to say so plainly.

The clearest instance is the elevator cable. A Ceres space elevator needs about 12 kilograms of drawn nickel-iron wire per kilogram of payload capacity, and the taper ratio is low enough that no exotic material is required — the surface-to-synchronous potential barrier is 58.6 kJ/kg, against about 60 MJ/kg for Earth. The cable is therefore the foundry's product, and the foundry's customer is the elevator, which then lowers the cost of running the foundry. Details on The Ceres Space Elevator. Modeled

The fuel dock

The Fuel Dock Survives in One Direction Only

Outbound traffic, to a customer base that does not exist yet.

Water is the belt's easiest product. At Ceres it takes roughly 0.63 kWh/kg to liberate water from the surface material and about 6.30 kWh/kg to crack it into hydrogen and oxygen at 70% cell efficiency — ten times as much energy to split it as to get it. The conclusion is to ship water as water and electrolyze it where the sunlight is, which at 177 W/m² is not Ceres. Be careful with the commonly quoted 25% water figure: that is a bulk-body number for Ceres including its interior, while the equatorial surface runs about 5 wt% bound as hydroxyl in phyllosilicates and needs 600 to 1,000 K to drive off. The full extraction case is on Belt Water & Propellant.

Where that water is worth selling is set entirely by direction. Ceres has essentially no gravity well — escape velocity is 516 m/s, about what a rifle bullet does — so leaving costs almost nothing. What it also means is that arriving costs almost everything, because there is no well deep enough for an Oberth maneuver to pay for capture. Both facts push the same way: the belt is a good place to ship from and a bad place to ship to.

For a vehicle headed outward to Jupiter, departing from Ceres costs 2.55 km/s of heliocentric change plus 0.52 to leave the body, about 3.1 km/s all in. The same trip from the lunar surface costs 6.3 km/s, and from Earth's surface 15.7. The belt is the cheapest departure point in the inner solar system for anything going further out, and it is the last place before the outer system where sunlight, water, and metal are all available in the same location — 177 W/m² at Ceres falls to 50 W/m² at Jupiter.

The customer for that is the outer-system traffic, and there is none. The belt's best market is downstream in space and also downstream in time. Projected

The gradient runs outward

Ceres Beats the Moon Outbound by Two to One and Loses Inbound by the Same

One table settles three of the four candidates, because the buyer's location is what decides them.

The relevant comparison is not distance but the velocity change needed to move a kilogram from a candidate source to a candidate destination. Ceres orbits at 2.77 AU with an orbital velocity of 17.90 km/s, against Earth's 29.79 and Jupiter's 13.06. A Hohmann transfer inward from Ceres to Earth costs 4.86 km/s at departure; outward to Jupiter it costs 2.55. The asymmetry is geometric — going outward means slowing down less than going inward means slowing down — and it is the single fact that organizes belt economics.

Lunar and terrestrial departures take their heliocentric change as a perigee burn, which recovers most of the Oberth benefit; Ceres has no well to dive into, so its 2.55 km/s is paid flat. The Ceres figure is still less than half the lunar one and less than a fifth of the terrestrial one.

Lunar and Ceres deliveries both assume aerocapture at Earth. A Ceres return enters the atmosphere at 12.84 km/s, which is the Stardust regime — demonstrated once, in January 2006, on a 46-kilogram capsule, and never qualified at tonne scale. The Earth column is a launch price rather than a Δv cost in practice, and that price has been falling for 15 years.

Read together, the two tables say the same thing from both ends. For every customer that exists today, the belt is the worst supplier available. For every customer beyond Jupiter, it is the best. The mine and the foundry both aim inward, into the column where Ceres is 74% more expensive than the Moon and 150 times slower. The fuel dock aims outward, into the column where Ceres is half the price of anything else and the traffic has not started.

What is actually scarce

The Scarce Thing Is Knowing What the Bodies Are Made Of

The total mass of asteroid material ever returned to Earth is about 127 grams, all of it from carbonaceous bodies.

Every belt business plan written so far rests on a composition that nobody has measured. The 16 Psyche valuation of roughly $10 quintillion is the body's mass multiplied by the terrestrial spot price of metal — about 90,000 times world GDP, which is the tell — and underneath it sits a spectral inference. M-class is a classification made from reflected light. Psyche's bulk density is 4.172 g/cm³, too low for solid metal and consistent with a porous rubble structure of mixed composition. Ground truth arrives when NASA's Psyche orbiter inserts in August 2029.

The returned-sample record is thinner still. JAXA's Hayabusa brought back roughly 1 milligram from Itokawa in 2010; Hayabusa2 returned about 5.4 grams from Ryugu in December 2020; NASA's OSIRIS-REx returned about 121.6 grams from Bennu in September 2023. That is the entire holdings of the human species, and every gram of it is from a carbonaceous body. Nobody has ever held a sample from a known M-class asteroid. Observed

Itokawa's yield was about 1,500 grains too small to weigh conventionally. Tianwen-2's figure is a mission target rather than a result; the spacecraft launched 29 May 2025 and arrived at Kamoʻoalewa on 2 July 2026 at 20 km range, with sampling planned through the end of April 2027 and the Earth-return capsule released in November 2027.

This is where the belt's first product is. On a site with a 30.8-month response lag, arriving and discovering that a body is not what the spectrum suggested is not a setback that gets corrected next quarter — it is a decade. A characterized body is therefore worth something to the second operator that it was not worth to the first, and the characterization is a durable asset in a way that a tonne of platinum is not: nobody on Earth can substitute their way out of not knowing what 16 Psyche is made of.

Whether that constitutes a business anyone gets paid for is an open question, and it is the honest weak point of this page. Survey data has historically been a public good, produced by space agencies and given away. The near-Earth object record shows why it is nonetheless the thing that has to come first: the first commercial wave, Planetary Resources and Deep Space Industries, failed 0-for-2 without characterizing a single target, and the second wave is building cheap replicable spacecraft precisely because you cannot iterate on a body, only across a population. That argument is on Near-Earth Objects. Projected

The strategic position here runs opposite to the Moon, and it is worth stating plainly. China has the only asteroid sample-return mission currently at its target and the only funded main-belt arrival: after releasing the Kamoʻoalewa capsule in November 2027, Tianwen-2 continues to the main-belt comet 311P/PANSTARRS for insertion in January 2035. The United States has one science orbiter in cruise and three venture-funded startups behind a 0-for-2 commercial record. If the belt's first product is a characterized body, China currently operates the only production line.

Built before it is bought

Each Stage's Customer Is the Next Stage

Which is what a bootstrapping regime looks like from the inside, and why revenue arrives last.

Nothing in the four-part test says the belt is uneconomic. It says the belt has no export market at the moment anyone would start building, which is a different claim and a much more ordinary one — it describes the first several decades of most industrial regions. What follows from it is a sequence in which each stage's output is consumed by the stage after it, and the first external sale comes near the end.

The elevator is listed after the foundry because its cable is a foundry product at roughly 12 kg of drawn nickel-iron wire per kg of payload capacity. A mass driver can substitute for the elevator in the launch role and is faster to field — 610 m of track at 30 g, 0.037 kWh/kg — which is why the practical sequence runs launcher first and elevator second, with both retained for redundancy against a single-cable failure.

Reading down the revenue column is the whole finding. Four of the five stages have no external customer, and the belt still has to build all four before the fifth can sell anything. That is a capital structure question rather than a physics question, and it is the reason the belt is a two-generation proposition where the Moon is a one-decade one.

It also explains why the governing scarcity is time rather than energy or mass. Energy is available — 177 W/m² of sunlight, with the crossover to nuclear at roughly 100 to 300 kWe set by the reactor's specific-mass knee rather than by distance. Mass is abundant and already out of a gravity well. What cannot be bought at any price is a shorter feedback loop: the 15.3-month launch window plus the 15.5-month transit is fixed by orbital mechanics, and every design error costs one full cycle of it. The fissile material is the one exception to local abundance — chondritic uranium runs about 10 parts per billion against 1,000 to 20,000 parts per million in terrestrial ore, five to six orders of magnitude, so every gram is an Earth import. That binds the growth rate rather than the possibility, and it is set out on Belt Power & the Fissile Leash.

What would make it a market

Three Developments Would Turn the Belt Into an Export Economy

All three are external to the belt, and all three are testable against published launch and mission records.

What none of the three requires is a technical breakthrough in the belt. The metallurgy works with 1970s process chemistry, the elevator needs no exotic material, the power is available, and the propulsion is off-the-shelf electric. The belt is not waiting on an invention. It is waiting on a customer.

The presence marker on the hub stays Contested for the same reason. Ceres sits on the crewed-viability line and can fall either side depending on whether the supervision case for humans at millisecond latency beats the cost of keeping them alive at 2.77 AU — which is argued on Crewed at Ceres. Nothing on this page depends on which way that falls. A belt built entirely by machines has the same four candidate businesses and the same three falsifiers.

Record

Provenance and Data Sources