Starship's Ride Home Gets Made From Mars Air And Ice

In 2016, SpaceX test fired its first Raptor engine. It burns liquid methane and liquid oxygen. SpaceX's 2017 Mars plan doesn't carry the propellant for the trip home from Earth. It gets that propellant from Mars itself.

The refueling stop

Mars has carbon dioxide in its air and water frozen in the ground. A factory can take those two materials, use electricity to process them, and make methane and oxygen, which are the two things the engine needs. Elon described that approach in his 2017 Mars presentation.

Why do it this way? Because you are sending equipment that makes the return supply at the destination. That takes a whole load of propellant out of the cargo you have to deliver from Earth.

I call it the return factory. Its job is very specific: fill the tanks that let the ship leave Mars. The catch is that a gas station on Earth just stores fuel somebody else made. On Mars the station has to do the manufacturing too. It has to collect its ingredients, turn them into the right liquids, and keep those liquids ready for the ship. There's a lot of equipment sitting behind that pump.

Most of the load is oxygen

Why does a rocket need so much oxygen? Because methane needs oxygen to burn. The ship carries the two in separate tanks, and Raptor brings them together in the engine. A full methane tank does nothing for you if the oxygen tank is empty. "Propellant" covers both: methane is the fuel, oxygen lets it burn. When somebody talks about making rocket fuel on Mars, the factory has to deliver the whole pair.

And the pair is very lopsided. In its 2021 environmental filing, SpaceX described Raptor using 3.6 parts oxygen for every one part methane by weight. That puts oxygen at about 78% of the propellant's weight. More than three quarters of what you're trying to make is oxygen. Make that oxygen locally and you take a huge amount of mass off the delivery list.

Where does it come from? Mars's air is about 95% carbon dioxide, so oxygen is already sitting in the air, locked to carbon. A machine has to pull it apart before an engine can use it. Water is a second source of oxygen. It also carries the other thing methane needs: hydrogen.

That's why ice matters so much in this plan. Methane is carbon plus hydrogen. The carbon comes from the Martian air. The hydrogen comes from the water. The factory takes those ingredients apart and rebuilds them into something useful.

Splitting water

The machine that does it is called an electrolyzer. Water goes in, power goes in, and two separate streams come out, hydrogen and oxygen. The oxygen can go to storage. The hydrogen goes on to the equipment that makes methane. So the factory is already producing something the rocket needs before the methane step even starts.

Where does the water come from in a working plant? You dig up the ice, get it out of the dirt, melt it, and clean it. Then it can go into the splitter. Each step feeds the next. If the digging equipment stops delivering ice, the water supply stops growing. If the splitter stops, the methane equipment loses its hydrogen. The whole factory has to run as one system.

Does that mean Mars hands you free fuel? No. You still have to make it. The plant spends energy turning those raw materials into the chemicals the engine burns, and that energy comes from a power supply you brought and set up yourself. Electricity runs the splitter and other equipment around the site. The return trip starts consuming power while the rocket is still sitting on the ground.

The Sabatier step

Now the hydrogen has to meet the carbon dioxide. That's the Sabatier reaction. Carbon dioxide from the air and hydrogen from the splitter go into a reactor, and under the right conditions they produce methane and water. The reactor uses a catalyst, a material that helps the reaction happen. Nickel is one example used in these reactors. You also control temperature and pressure so the reaction runs properly.

What happens to the water the reactor makes? You send it around again. The methane goes to the fuel supply. The water goes back to the splitter, electricity separates it again, and the hydrogen returns to the reactor for another shot at becoming fuel.

Then why does the factory still need fresh water? Because hydrogen leaves the plant inside the methane you're storing, so you have to replace what leaves. Recycling helps the plant use its ingredients well. It doesn't create more of them.

The best part shows up when you look at the whole loop. Carbon dioxide and water go in. Methane and oxygen come out. By weight, the complete recipe yields four parts oxygen for every one part methane. Raptor's 3.6 to 1 from the 2021 filing needs a bit less oxygen than that. So making enough methane also makes enough oxygen, with some left over. The two outputs match the engine very well.

It already worked off Earth

Has any of this worked away from Earth? Yes, in pieces.

The International Space Station first ran its Sabatier reactor in October 2010. It took hydrogen from the station's oxygen generator, which splits water and leaves hydrogen behind, and carbon dioxide breathed out by the crew. The reactor made water and methane. The station kept the water and vented the methane into space. Same reaction, opposite priorities. A Mars plant would keep the methane, because that's what the ship burns.

The oxygen side got its test on Mars itself. In April 2021, NASA's Perseverance rover made oxygen on Mars with a device called MOXIE. It pulled in carbon dioxide from the air and used electricity to separate out oxygen, with carbon monoxide as its other output. In September 2023, NASA reported a total of 122 grams of oxygen, about 4.3 ounces. That's roughly what a person breathes in a few hours.

So why send a machine all that way to make so little? Because the point was to prove it could work with the air it actually found there. Its compressor had to pull in thin air. Its filter had to keep dust out. Its oxygen cells had to keep working as conditions changed. The test runs covered a Martian year, by day and by night, and it made oxygen through all of it.

Getting the factory there

Getting all that equipment to Mars starts with a different refill. In SpaceX's 2017 Mars plan, the ship refuels while it's still in Earth orbit. Reaching orbit burns nearly all the propellant it launched with, so tanker ships bring up more, and then it can leave Earth carrying the factory. Super Heavy, the big booster underneath, stays on Earth. The ship makes the trip, and that's the vehicle the Martian plant has to fill for the return. (I covered the tanker side and the heat shield in "Why Elon Bet Starship on Hexagons," so I'll stay on the factory here.)

Doesn't all this equipment add weight? Of course it does. The useful comparison is equipment against the propellant it makes. In a 2017 NASA study, the production equipment weighed about 1.7 metric tons, roughly 3,700 pounds. It was sized to make about 30 metric tons of propellant, about 66,000 pounds. That design included processing soil with very little water, and it needed a separate power supply. The machinery stays put while local material flows through it, so a small mass shipped from Earth turns into a much bigger supply at the destination.

On the transfer side, SpaceX tested part of the process on March 14, 2024, during Starship's third flight. The company posted: "Propellant transfer demo complete." That test moved liquid oxygen from the header tank in the ship's nose to the main tank farther back, using a pressure difference to push it. Both tanks were inside the same ship.

Moving propellant between two ships is a harder job. Two spacecraft have to meet, connect, and move liquid across. As of November 2026, SpaceX had yet to demonstrate methane and oxygen moving between two Starships in orbit, and it now targets that test for early 2027. In a NASA video published in February 2025, astronauts Matthew Dominick and Don Pettit used water bottles on the space station to show why: in space, getting a liquid to settle where it has to flow out is a real problem. A tanker has to solve that with the extremely cold liquids a rocket uses.

So the Earth refill gets the factory to Mars. Then the factory has to turn whatever is at the landing site into the ship's way home. Which means where it lands decides how hard its job is.

Ice you can reach

Has a spacecraft actually touched ice on Mars? Yes. In June 2008, NASA's Phoenix lander exposed bright chunks under the soil, and over the next few days those chunks disappeared. That was part of the evidence the material was water ice. In July 2008, Phoenix heated a soil sample and confirmed water.

The plant needs ground where a ship can land and where water is accessible at the same spot. A deposit under rough terrain doesn't give you a usable factory site. In May 2025, Elon named Arcadia Planitia as the leading landing candidate. It's a broad plain, and studies have found buried ice there. The appeal is the combination: landing terrain on top, water underneath.

How much dirt comes along changes everything. Picture two sites holding the same amount of water. One has a thick ice deposit under a cover of dirt. The other has a little water spread through a lot of soil. The second site means processing much more material for the same water. Same rocket, same chemistry, far more mining work for every tank you fill.

The 31-foot cube

How much ice does a ship-sized plant need? Work backward from full tanks. SpaceX's vehicle page listed ship capacity at 1,600 metric tons of propellant in 2026, about 3.5 million pounds. Run that through the 3.6 to 1 mixture from the 2021 filing, which listed a smaller 1,500 ton tank, and you can size the ice.

For that much methane and oxygen, the water needed, as pure ice, would fill a cube about 31 feet across, about 9.5 meters. That wide, that tall, that deep. Big, but you can picture it, and once you can picture it you can ask the right questions: what machine reaches it, how much dirt sits on top, how the water gets back to the plant.

That cube is the water requirement. It isn't the hole. The hole includes every bit of material around the ice that has to move, and the deposit decides how big that is. This is why concentrated ice is so valuable. A cleaner supply means less digging and processing for the same output. For the crew, that difference shows up as time spent waiting for tanks to fill. My read: the landing site is part of the fuel system.

Thin air, dust, and boil-off

The other ingredient surrounds the plant, but collecting it takes work too. Average surface pressure on Mars is about 0.6% of Earth's at sea level, so there's very little gas in each breath of air. The plant has to pull in a lot of it and compress it to feed the reactor.

Dust comes in with it, so the intake needs filters that keep working as more air flows through, plus a way to clear the dust that piles up. The carbon dioxide also carries small amounts of nitrogen and argon. The Sabatier reaction doesn't use them, so the plant has to separate or vent them so they don't build up.

Then the gases have to become liquids. That takes refrigeration. Mars is cold, but liquid oxygen and liquid methane have to be far colder than the average Martian surface, so the surroundings still leak heat into the tanks. Some liquid turns back to gas. That's boil-off, and a supply that waits months needs protection from it. Insulation slows the heat coming in. Powered refrigeration pulls it back out. Cooling keeps drawing electricity even after the fuel is made.

Power and the clock

Which brings everything back to power. In May 2025, Elon said the Mars propellant plants would run on solar. That makes sunlight part of the production schedule. Panels stop at night. Dust storms cut the light. To keep running through those stretches, the system needs stored energy, or it has to throttle production to match the power available while still keeping the stored propellant cold.

What does a schedule look like? A NASA study published in January 2024 sized a plant to make 300 metric tons of methane and oxygen, about 660,000 pounds, over 20 months. That's less than a fifth of the full tank example. The study also started with water already delivered to the plant and specified nuclear power. It needed continuous electricity for production and liquefaction.

That points to one conclusion: the factory needs a head start on the people. Mars Direct built that sequence in back in 1990. Robert Zubrin and David Baker's plan sent the return vehicle and its propellant equipment ahead of the crew, and the crew left Earth only after the tanks were filled and checked. That's a very different decision from landing people and hoping the plant starts.

Earth and Mars line up for low-energy trips about every 26 months, a little over two years. The factory works in the gap between those windows, which is why sending it early is so useful. If the supply isn't ready for the planned departure, you lose that window. Starting early gives you time to find a problem while the crew is still on Earth.

A trip with a return leg

So what would get you on that ship? A working reaction is useful evidence. A tank actually filled by equipment at your actual destination is the result you need, because the mission can measure propellant that exists. And a working plant keeps producing after the first ship leaves, so the next arrival can count on the same factory.

That Raptor firing from 2016 shows the engine burning the two liquids. Making those liquids on Mars gives that engine a supply for another departure. In my view, that's what makes this whole effort worth it: the ship can carry people home and stay part of a transport system that gets used again. The ice under the ground becomes part of someone's way back to Earth. Build the place to fill up, and a trip to Mars can include a way home.