LandSpace landed Zhuque-3. There are now two countries with reusable boosters
A methane-fuelled, stainless-steel first stage came back on landing legs — on the second attempt. In the same week SpaceX flew its 100th mission of the year.
The Chinese startup LandSpace recovered the first stage of its Zhuque-3 rocket, landing it on deployable legs. It is China’s first orbital-class booster recovery, and it came on a second attempt after an earlier failure.
The vehicle’s design choices are worth noting because they are not incidental: methane-fuelled and stainless steel. That is the same pair of bets Starship made, arrived at independently, and for the same reasons.
Why methane and steel keep winning
Methalox
Methane burns cleaner than kerosene, which is the practical argument — RP-1 deposits coke inside engine passages and turbopumps, and cleaning or replacing that hardware is a large part of why kerosene stages are expensive to reuse. Methane leaves the engine substantially readier to fly again. It also runs at a cryogenic temperature close enough to liquid oxygen’s that tank and plumbing design simplifies considerably.
Stainless steel
Steel is heavier than aluminium-lithium or carbon composite per unit strength at room temperature, and it is the right answer anyway. It gets stronger at cryogenic temperatures rather than weaker, it tolerates re-entry heating with far less thermal protection, it is cheap, and it can be welded in a tent rather than cured in an autoclave. For a vehicle you intend to fly repeatedly, manufacturability and thermal margin beat mass fraction.
Two independent programmes converging on the same unusual materials answer is reasonably strong evidence the answer is correct.
What the failure-then-success sequence tells you
The first attempt failed. The second worked. That progression is the actual milestone — more than the landing itself.
Propulsive landing is a control problem before it is a hardware problem: throttling deeply, holding attitude through a boost-back and re-entry burn, and nulling velocity within centimetres of the ground with almost no fuel margin. You cannot fully simulate your way to it, because the residual errors are in exactly the aerodynamic and engine-transient regimes that models get wrong. You iterate on flight data. LandSpace has now demonstrated it can close that loop, which predicts a third and fourth success far better than a first-attempt fluke would.
The cadence gap remains enormous
For scale, in the same week SpaceX flew its 100th mission of 2026 — the third consecutive year it has reached that mark — carrying 24 Starlink satellites from Vandenberg. The constellation now exceeds 11,000 satellites in orbit.
| SpaceX | LandSpace | |
|---|---|---|
| 2026 missions to date | 100+ | Low single digits |
| Booster recoveries | Routine, hundreds cumulative | One |
| Reflight demonstrated | Yes, extensively | Not yet |
| Operational constellation | 11,000+ satellites | None at scale |
One recovery is not a competitive position. But the graph that matters is not the current level — it is the slope, and the slope now includes a second organisation that has proven it can learn from a landing failure. The interesting question is no longer whether China can recover a booster. It is how many flights it takes to go from recovery to routine reflight, which is where the cost curve actually bends.
Recovering a booster proves you can land. Reflying one proves you can afford to.
Also this week, and in a similar vein of the physical world getting harder to ignore: Astranis introduced its Perceptor satellites for monitoring geostationary orbit at 22,236 miles. Space situational awareness becoming a product category is its own signal about how crowded things have got.