Reusable rockets
Falcon 9 landed its first stage in December 2015 and now flies the same booster dozens of times; Starship aims for full reuse.
Open in the interactive tree →Before 2015 rockets were thrown away after one use. SpaceX landed a Falcon 9 first stage on 21 December 2015, reflew one in March 2017, and flew 165 Falcon 9 missions in 2025, more than all other orbital rockets together. Blue Origin's New Glenn landed its booster in November 2025 and re-flew and landed it on 19 April 2026, although that flight left its satellite in a useless orbit.
As of October 2026
Falcon 9 booster B1067 flew its 37th mission on 25 August 2026, the 100th Falcon 9 launch of the year. Starship Flight 14 on 28 September 2026 reached orbit for the first time and released 26 Starlink V3 satellites; the booster splashed down in the Gulf of Mexico and the ship in the Pacific, and a ship has never been caught. Ship-to-ship propellant transfer, needed for Moon and Mars missions, has not been demonstrated and had no date as of September 2026.
Open steps
- Learned control for landings Medium AI leverageDevelop RL or neural-network control for booster descent and landing, reducing engineering tuning and enabling rapid iteration.
- Reusable TPS system design Medium AI leverageUse ML and simulation to optimize thermal protection (tiles, coatings, ablators) for repeated launches.
- Sub-24-hour turnaround for relaunch Low AI leverageDevelop autonomous inspection and checkout procedures to shorten turnaround from weeks to <1 day.
Where AI could help
Low AI leverage. The next steps, catching ships and refuelling in orbit, are engineering and flight-test problems; AI helps guidance, inspection and simulation at the margins.
- Learn landing and catch guidance in simulation before real flights
- Inspect heat-shield tiles and engines automatically between flights to cut turnaround
- Use surrogate models of re-entry heating and engine combustion to shorten design cycles
- Predict component wear to set safe reuse limits