How Do You Get Water Out of an Asteroid? Methods, Energy and What Nobody Has Done Yet

Almost every plan to make money from asteroids ends up at the same product: water. Split it into hydrogen and oxygen and you have rocket fuel that never had to climb out of Earth's gravity well. So how, exactly, do you get water out of a rock drifting through vacuum?

The short answer: you bake it. The long answer is the subject of Frontier's second research brief, and it turns out to be less about heat than about everything around the heat.

The water is not ice

The first surprise is that carbonaceous asteroids, the dark "C-types" everyone wants to mine, do not contain ice. They contain clay.

Billions of years ago, liquid water flowed through these bodies and reacted with their rock, producing minerals called phyllosilicates: serpentine, saponite and relatives of the clays in your garden. The water is still there, but chemically bound inside the mineral structure as hydroxyl groups and thin layers of molecules between crystal sheets. It does not drip out. It has to be driven out.

Meteorites tell us how much. The most water-rich class, the CI chondrites, hold about 20 wt% water. The far more common CM class averages about 7 wt%, with a range from under 2 to over 10. CR chondrites sit below 9.

Then came the samples. Japan's Hayabusa2 brought back grains from asteroid Ryugu; one gave 6.8 mass% total water, less than the CI meteorites it resembles. NASA's OSIRIS-REx brought back 121 grams of Bennu; the sample is hydrated at 0.84–0.95 wt% hydrogen, built from roughly 80% phyllosilicates by volume, with up to about 2 wt% of loosely held water between clay layers. If every hydrogen atom in Bennu were water, that would be about 8 wt%; some of it is organic matter, so the real recoverable figure is lower.

The honest planning number for a Bennu- or Ryugu-like asteroid is therefore five to eight percent water by mass, not the twenty percent that the best meteorites promise. A hundred tonnes of regolith yields a few tonnes of water: still enormous at orbital prices, but a quarter of what older business plans assumed.

Bake at 400 to 770 degrees

To release bound water you have to heat the mineral until its structure breaks down. Laboratory heating of CM meteorites shows the sequence clearly: adsorbed moisture leaves below 200 °C, iron hydroxides give up water at 200–400 °C, the phyllosilicates themselves dehydrate between 400 and 770 °C, and carbonates break down above that. A process that stops at 300 degrees collects only the easy fraction.

That temperature sets the energy bill. From published constants for the heat capacity of rock, the dehydration of serpentine and the latent heat of water, Frontier's brief works out a thermodynamic floor of roughly 9 megajoules per kilogram of water from CI-like material, and about 16 megajoules from CM-like material, before any losses. In kilowatt-hours that is 2.5 to 4.5 per kilogram. Real hardware does worse. NASA Marshall's microwave experiments on simulated soils measured 2 to 17 watt-hours per gram, which is 7 to 61 megajoules per kilogram.

Why does this matter? Because the Sun delivers 1,361 watts per square metre at Earth's distance. A one-square-metre mirror at 50 percent efficiency yields about 680 watts, so one kilogram of water from CM-type rock takes around six and a half hours of sunlight. A tonne a year is a one-to-two-kilowatt job: roughly what AstroForge's 200 kg DeepSpace-2 carries, and well beyond anything cubesat-sized.

Three ways to apply the heat

Every method proposed so far does the same three things: heat the rock, let the vapour out, and catch it somewhere cold. They differ in where the heat goes.

Concentrated sunlight. TransAstra's "optical mining" wraps a small asteroid or boulder in an inflatable bag and focuses sunlight on it until the surface spalls and releases water from its hydrated minerals. NASA picked the company's 250 kg "Mini Bee" demonstrator for a NIAC Phase III award in 2019, though we can find no evidence it has flown. The most recent ground tests, at Colorado School of Mines in 2025, reached what engineers call TRL 5 and found an awkward side effect: organic vapour boiling out of the rock coats the hardware and slows the process down.

Microwaves. Microwaves heat the bound water directly rather than the whole rock, and they penetrate loose regolith. NASA Marshall engineers extracted water from frozen lunar simulant this way and then repeated the test with a simulated carbonaceous asteroid, recovering all the free water plus an extra 3 millilitres that had been chemically bound in the clay. A published small-spacecraft design built around seven 200-watt microwave probes concluded each craft would cost over $100 million and that more than a hundred would be needed to turn a profit, which shows how far the economics still have to move.

Dig, then cook. The alternative is to scoop or core material and heat it inside the spacecraft. Honeybee Robotics, now part of Blue Origin, built exactly that. Its World Is Not Enough prototype mined simulated asteroid regolith, baked out the water, froze it in a cold trap, turned it to steam and hopped inside a vacuum chamber on New Year's Eve 2018. The simulant was bone dry to the touch; all the water came out of the mineral structure. It also came out full of dissolved carbon dioxide, metals and organics, which is fine for steam but not for drinking.

All three end at a cold trap. In NASA's permafrost tests the trap captured about 99 percent of what sublimated. Keeping it cold and shaded on a rotating asteroid is an attitude-control problem as much as a thermal one.

The hard part is not the heat

The flight record is sobering. The problem is touching an asteroid at all.

When OSIRIS-REx tapped Bennu in October 2020, it sank 48.8 centimetres into the surface before its thrusters fired, leaving an eight-metre crater. The surface behaved like a fluid. Models later showed that regolith with a cohesion of 25 pascals offers essentially no resistance to a sampler. The European Philae lander's harpoon anchors never fired on comet 67P and it bounced across the surface. Hayabusa2's team at Ryugu found the surface so bouldery they "cannot find a 100 percent safe place to touch down".

Every newton of digging force has to be reacted by an anchor. Researchers now favour drilling anchors over harpoons; an ultrasonic cross-drilling anchor held 60 to 250 newtons on a bench, but has never been tried in real microgravity.

Then there is dust. Bennu was caught throwing centimetre-sized particles off its own surface at three metres per second, some of which briefly orbited the asteroid. Any excavation adds more. Optics, radiators and the cold trap are exactly the surfaces that dust ruins.

And spin. Frontier's first brief showed that most asteroids under 100 metres across rotate in minutes, fast enough that nothing can rest on their surface. On those bodies the only option is to bag the whole object. On slow rotators like Bennu or Apophis, surface operations are possible, but the Sun crosses the sky every few hours, so a solar concentrator or shaded trap works at most half the time.

What a small first test could prove

Frontier's first spacecraft is heading for a dry S-type asteroid, Apophis, the only well-understood body reachable in our window. There is no water there to extract. So why think about extraction now? Because the risks that matter are handling risks, and those can be tested on any surface, or on no surface at all. The brief proposes a three-stage programme: vacuum-chamber tests on CI simulant at $95 a kilogram, then microgravity tests on parabolic flights or Blue Origin's New Shepard, whose lockers carry 11 to 23 kilograms, then a 2 to 5 kilogram flight package on whichever Frontier or partner mission first reaches a carbon-rich target.

Five kilograms is not fanciful. Hayabusa2's MINERVA-II rovers weighed 1.1 kilograms each. A heated sample cell, a cooled trap with a microbalance, a contact sensor to measure how the regolith pushes back, and an infrared sensor to confirm what came off: that package could prove bound water can be released and captured in space, measure the real energy per gram, and show whether the surface can be pushed on. It cannot prove throughput, or months of storage in deep space, which one analysis rightly calls "unproven at any scale".

Who is actually working on this

Fewer organisations than the headlines suggest. TransAstra and Colorado School of Mines on optical mining. Honeybee/Blue Origin on dig-and-cook. NASA Marshall on microwaves. Karman+, which raised $20 million to extract water for orbital refuelling, says it will deploy tests on the surface of Apophis around 2030, but has not published how it intends to heat regolith. Nobody has flown anything.

That gap is the opportunity. The water is there, the chemistry says how to free it, and the energy budget fits a one-kilowatt spacecraft. What is missing is a flight test of the unglamorous parts: anchoring, feeding grains without gravity, keeping a trap cold, and surviving the dust. That is a problem a small, focused team can attack.

This is Frontier's Research Brief 02

The full brief, with every figure traced to sample-return papers, meteorite studies, NASA reports and company statements, is published at thefrontier.space/research. We wrote it to be checked: if you find an error or newer data, tell us.

Frontier is community-funded. If you want to help pay for the simulant, the vacuum-chamber time and eventually the flight package this brief describes, see how to contribute at thefrontier.space/rewards.