Starship orbital refueling explained: why the Moon landing depends on it

You have probably read that NASA’s return to the Moon “depends on refueling Starship in orbit” and wondered whether that is a real engineering requirement or a talking point. It is a real requirement, and as of 20 November 2026 the central step has not been shown to work: no one has yet moved large amounts of super-cold propellant from one spacecraft to another in space. SpaceX has done a smaller version inside a single ship, and its own executives point to early 2027 for the first true ship-to-ship attempt. Every number you see about how many tanker flights a landing needs is an estimate, and the estimates disagree by a factor of three or more.

Why a Moon lander needs a gas station

The trouble is the rocket equation, which punishes you for carrying fuel to carry fuel. Starship as NASA’s Human Landing System is a very large vehicle that has to fly to the Moon, land, and lift off again. To launch with all of that propellant on board, the whole stack would be too heavy for the booster to lift into orbit. So the plan is to launch the lander with only enough propellant to reach low Earth orbit, then top up its tanks there before it leaves.

That is not unusual in spirit. Airliners refuel on the ground, and military aircraft refuel from tankers in flight. The difference here is the fuel: Starship burns liquid methane and liquid oxygen, which are stored at temperatures of roughly minus 160 and minus 183 degrees Celsius. They want to boil the moment anything warms them, and in orbit, where everything floats, you cannot rely on gravity to keep liquid at the bottom of a tank.

The tanker concept

The approach NASA and SpaceX have described is a chain. Tanker versions of Starship launch from Earth, reach orbit, and meet another ship. Reports of a NASA graphic published in early November 2026 describe the Artemis IV sequence as starting with a depot ship placed in low Earth orbit, which collects propellant from the tankers. The lander then docks with the depot, fills up, and departs. The tanker ships are the same basic vehicle as the lander with a different payload. Because the booster and ship are meant to be reused, SpaceX’s argument is that these flights are cheap compared to disposable rockets. Whether that is true depends on reuse working at high speed, which we discuss in our pieces on launch costs and the Flight 15 ship catch.

There is a second route NASA has kept open. Blue Origin’s Blue Moon lander is a smaller vehicle that also depends on in-space fuel handling, although its plan uses several New Glenn launches. Reports of the November graphic say NASA could pick Blue Moon for Artemis IV if SpaceX cannot show that refueling works. That is a hedge, not a forecast.

Has anyone done it yet?

Here the answer needs care, because two different things get mixed together in headlines.

What has been done: on Starship Flight 3, on 14 March 2024, SpaceX moved propellant between two tanks inside the same ship while in space. NASA-linked technical papers describe it as several metric tons of liquid oxygen, and some press accounts say about 10 tons. It used a gentle, controlled acceleration to push the liquid to the right end of the tank so the pumps would not suck in gas. This was a real result, funded in part by a NASA technology contract. It was one vehicle, with a pipe between two of its own tanks.

What has not been done: moving cryogenic propellant between two separate spacecraft that have flown up independently and docked. NASA technical guidance notes that this has never been demonstrated. The planned Starship Propellant Transfer Demonstration involves two ships, with the second launching some weeks after the first and rendezvousing with it. The original target was early 2025, and it has slipped repeatedly. Reporting of SpaceX’s late-October earnings call has Gwynne Shotwell naming early 2027 for an orbital refueling attempt, framed as a development goal and not a published schedule. A Wikipedia entry lists it as no earlier than the first quarter of 2027. One newsletter claimed two ships had already completed a transfer on Flight 12; none of the other results we found support that, and the more recent reporting still treats the demonstration as pending, so we do not count it.

The vehicle has also been changing underneath all this. Flight 12 on 22 May 2026 was the first launch of the upgraded Starship V3, and it was only partly successful, with the booster losing an engine. Flight 15 on 13 November brought a ship back from orbit. A lander design that has never been in orbit for weeks, or sat through a long cold soak, is one thing; a lander that must wait while tankers arrive is another.

How many launches? Estimates only

This is the number everyone asks for, and nobody official has fixed it. The published figures we found are best read as a spread:

  • Elon Musk has said in public that a handful of tanker flights would do, with figures such as four to eight in older statements and six to eight in more recent coverage.
  • A SpaceX executive in 2024 called it “ten-ish,” and said it could change.
  • NASA’s Lakiesha Hawkins said in November 2023 that the count was in the high teens.
  • The Government Accountability Office described SpaceX’s baseline proposal as 16 launches in total, which it broke down as one depot, 14 tankers and the lander.
  • In October 2026, reporting quoted Charlie Precourt of NASA’s Aerospace Safety Advisory Panel as estimating up to about 22 flights for Artemis IV, with the higher figure attributed to reserves and propellant that evaporates.

These do not conflict as much as they look. The low numbers assume tankers that carry a lot, little boil-off, and a fast cadence. The high numbers assume the opposite. The honest reading is that the count depends on tanker payload that has not yet been demonstrated in orbit, and the count is likely to be revised as soon as there is real data. If someone gives you a single confident figure, ask them which assumptions they used.

Boil-off and settling, the two physics problems

Boil-off. Every watt of heat leaking into a cold tank turns liquid into gas. Sunlight on a big steel ship in low Earth orbit is a lot of heat. NASA lists boil-off reduction, tank stratification and pressure control among the open technology gaps for long-duration cryogenic storage. The reason it hits Starship harder than most is the waiting time: tankers do not arrive all at once, so the first propellant delivered sits in orbit for weeks before the last. Every day costs some of it, which is part of why higher estimates add flights. Better insulation, reflective coatings, or active cooling can help, but those add mass.

Settling. On Earth, liquid sits at the bottom and pumps draw from there. In free fall, liquid floats in blobs and clings to walls, and a pump can pull gas instead. SpaceX’s Flight 3 test used small thrust to settle the liquid. Doing the same with two ships docked together is harder, because both tanks have to be managed, the pressures have to be balanced, and the connection has to stay sealed while the ships push and drift.

What this means for Artemis

This is the reason Artemis III no longer lands. NASA rewrote it this year as an Earth-orbit rendezvous and docking test, and the crewed landing moved to Artemis IV, currently targeted for 2028. Coverage we read says Artemis III does not include a refueling test, and that NASA’s Aerospace Safety Advisory Panel named Starship lander refueling as a major risk for Artemis IV. Our Artemis explainer covers the dates and the dispute over them in detail. For scale comparisons with the rocket that took Apollo to the Moon, see Starship versus Saturn V.

Questions people ask next

Could they just build a bigger rocket instead? Apollo did, with one giant launch. NASA’s Space Launch System flies rarely and cannot lift a lander of this size, and a bigger new rocket would be its own multi-year development. Refueling is the bet that a reusable rocket flown often is cheaper than one huge disposable launch.

What would count as proof? A two-ship transfer of a large fraction of a tanker’s load, with measured amounts in and out, followed by a second attempt that shows it is repeatable. One success would be a milestone, not a system.

What should I watch? The date of the two-ship demonstration, the amount transferred, and whether NASA’s own review of it signs off. A slip of that demonstration past mid-2027 would be a stronger signal about 2028 than any launch-count estimate.

We are a game studio and have no inside view; in our browser game we skip fuel logistics entirely, which is a luxury real engineers do not have. Our spaceflight hub has more explainers like this one.

Key sources: NASA’s Artemis III page, NASA Technical Reports Server (cryogenic transfer papers), SatNews on the NASA refueling graphic.