Beyond chemical rockets
Nuclear thermal or nuclear-electric propulsion could cut travel time or fuel mass for deep-space missions, but no nuclear rocket has flown.
Open in the interactive tree →Chemical rockets are limited by exhaust speed. A solid-core nuclear thermal rocket heats hydrogen in a reactor and is expected to reach roughly 850-1,000 s of specific impulse, about double hydrogen-oxygen engines. Nuclear-electric systems use a reactor to power electric thrusters with far higher exhaust speed but very low thrust. The US tested nuclear thermal engines on the ground from 1955 to 1973 (Rover and NERVA), and none has flown.
As of October 2026
DARPA ended its DRACO nuclear thermal flight demonstration in 2025 after NASA's fiscal 2026 request zeroed it; DARPA's deputy director cited falling launch costs. At its Ignition event in March 2026 NASA announced Space Reactor-1 Freedom: a roughly 20 kW-electric HALEU reactor feeding electric thrusters of up to 48 kW, launching to Mars in December 2028 on a trip of about a year, described by NASA as a '70 percent solution' pathfinder. On 24 February 2026 JPL fired a lithium magnetoplasmadynamic thruster at up to 120 kW, over 25 times the power of Psyche's thrusters, with a goal of 500 kW to 1 MW per thruster while a crewed Mars mission might need 2-4 MW.
What is missing
- Fuel that survives hydrogen near 2,700 K and thermal cycling; graphite fuel eroded and cracked in the NERVA era
- A flight-qualified space reactor with shielding and heat rejection; SR-1 Freedom is the first planned, for December 2028
- Electric thrusters at hundreds of kilowatts to megawatts with very long life; a Mars mission could need more than 23,000 operating hours
- Ground-test methods for nuclear thermal engines that do not release radioactive exhaust
- A stable funding path: DRACO was cancelled and the fiscal 2026 request had no nuclear thermal money
Becomes possible once solved
- Shorter crewed trips to Mars with less time in space radiation
- Heavy cargo to the outer planets
- Reactor power and design experience for lunar and Mars surface bases
Open steps
- Fuel that survives hot hydrogen Medium AI leverageFind reactor fuel and coatings that withstand hydrogen near 2,700 K and thermal cycling without cracking or eroding, the failure seen in NERVA-era graphite fuel.
- Reactor design with fast surrogates High AI leverageUse machine-learned emulators of neutronics and heat transfer to optimise compact reactor cores and control-drum logic for start-up and restart in space.
- Megawatt-class electric thrusters Medium AI leverageScale lithium MPD, Hall and similar thrusters from 120 kW to 500 kW-1 MW per thruster with thousands of hours of life; a crewed Mars mission might need 2-4 MW.
- First flight of a space reactor Low AI leverageFly a reactor-powered spacecraft to Mars (SR-1 Freedom, planned for December 2028) to prove start-up, heat rejection and thruster operation in space.
- Safe ground tests of nuclear engines Low AI leverageDevelop test methods and facilities that verify engine performance without releasing radioactive exhaust; DRACO's flight demonstration was cancelled in 2025.
Where AI could help
Medium AI leverage. AI speeds reactor, fuel and plasma-thruster design through fast surrogates, but reactors, fuels and thrusters must still be built and tested.
- Optimise reactor core geometry and fuel layout with fast emulators
- Build surrogate control models for start-up and shutdown
- Model plasma flow and electrode erosion in electric thrusters
- Screen materials for fuels and coatings that survive hot hydrogen
Shown so far
- In October 2025, Idaho National Laboratory and Abilene Christian University published a surrogate model of a nuclear thermal propulsion reactor start-up sequence that was accurate to within 5% and ran in minutes instead of days. source
- In December 2021, an Oak Ridge-led team reported an AI-designed reactor core with a threefold better temperature peaking factor, using a machine-learning emulator to screen thousands of geometries on the Summit supercomputer (a general reactor, not a rocket reactor). source
Prerequisites
- Tsiolkovsky's Rocket Equation1903The rocket equation sets the limit of chemical propulsion
- Liquid-fuel rocket (V-2)1942
- Nuclear Power1954
- Deep-space probes (Voyager)1977
- Ion Thruster1998Ion drives are the first working alternative to chemical rockets
- Reusable rockets2015