On 10 July 2026, China's new Long March 10B rocket lifted off and, a few minutes later, did something no Chinese rocket had done before: it flew its first stage back down onto a net-like recovery structure on a ship in the Yellow Sea, intact.
It's easy to undersell how hard this is. SpaceX spent years and multiple exploded boosters getting propulsive landings to work reliably, and even now its Super Heavy booster catches at the launch tower are a genuinely difficult piece of engineering. China's approach is different again: rather than legs or tower arms, the booster descends onto a floating net-like platform sized to catch it within a few metres of accuracy, on a rocking ship, on the very first flight of a new vehicle.
That last part is worth sitting with. This wasn't a booster that had already flown and landed a dozen times before anyone trusted it near a recovery structure. It was the maiden flight of Long March 10B, and China stuck the catch first time. Whether that's luck, unusually thorough ground testing, or some of both isn't clear from the outside, and a single success on one flight is not the same as routine, repeatable reuse. SpaceX's own early landing attempts were followed by plenty of failures before reuse became boring.
China National Space Administration hasn't announced a date for Long March 10B entering regular service, or how many recovery attempts it plans before declaring the system operational. But the timing isn't incidental. Long March 10B and its crewed sibling are the rockets China is counting on for its own lunar landing programme, targeted for before 2030. A reusable heavy-lift booster changes the maths on how many launches that programme can afford, the same way it changed the maths for SpaceX's Starlink constellation.
Beyond the immediate mission, this is the clearest sign yet that booster reuse isn't a SpaceX-only capability anymore. The US, historically the sole country with a demonstrated recovery system, now has company. What happens next, whether China turns one good catch into a reliable operational cadence, is the actual test.