NASA has been trying to build a working nuclear reactor in space since the Apollo era. On Monday, it finally said enough.
At the agency's Ignition event in Washington, Administrator Jared Isaacman announced SR-1 Freedom, a nuclear electric propulsion mission to Mars launching at the end of 2028. The reactor will produce 20 kilowatts of electrical power using high-assay low-enriched uranium, or HALEU. A truss structure will separate it from the spacecraft's electronics, with radiators handling heat rejection. The Power and Propulsion Element, or PPE — hardware built by Lanteris Space Systems for the now-cancelled lunar Gateway — will handle electric propulsion, drawing power directly from the reactor. Solar arrays will provide supplemental power, particularly in the period between launch and reactor activation.
None of this is new technology. The problem has never been the physics. NASA and its predecessors have spent a combined $20 billion over decades on space nuclear concepts that never left the drawing board. Steve Sinacore, program executive for NASA's Fission Surface Power program, laid out the failure modes plainly at Ignition: no sustained mission pull, projects too ambitious for their budgets, timelines untethered from reality, and leadership fragmented across agencies. An Idaho National Laboratory assessment published last summer reached the same conclusion. SR-1 Freedom is designed to break every one of those patterns — by actually having a mission that demands the hardware, keeping the reactor small and grounded in demonstrated capabilities, locking to a launch window that won't wait, and putting NASA in the captain's seat.
The in-house approach is deliberate. We're going to bring this in-house, with NASA as the prime. We're going to leverage the expertise of the Department of Energy to design and assemble the reactor, Sinacore said. The agency will share the reactor design with industry — no proprietary rights, to the benefit of all reactor companies. It's a playbook borrowed from the early days of nuclear power on Earth: build the reference design, let others iterate.
What makes 2028 at all plausible is PPE. The Power and Propulsion Element was built for the lunar Gateway, a station NASA no longer plans to build. It exists as flight hardware. Adapting it for SR-1 Freedom — using power from a reactor instead of solar panels — is the single biggest schedule saver in a program that has no margin for slip. PPE gives us a huge leg up. That's the only thing that makes this achievable, Sinacore said. That's a very capable spacecraft bus that is going to be adaptable.
Once SR-1 Freedom arrives at Mars roughly a year after launch, it will deploy SkyFall: three helicopter scouts derived from the Ingenuity design that flew on Perseverance. They will survey a potential human landing site, including searching for subsurface water ice — data any future crewed mission will need before touching down. They're going to be very, very similar in capabilities to Ingenuity, said Nicky Fox, NASA associate administrator for science. They're going to have cameras on them to be able to take images, but they're not going to be full-up heavy science birds. It's reconnaissance, not a rover.
What happens after SkyFall deployment remains an open question. NASA has not decided whether SR-1 Freedom enters Mars orbit, flies by the planet en route to another destination, or pushes the nuclear electric propulsion system further out. We do want to push the bounds with this demonstrator, Sinacore said.
The skepticism about 2028 is fair. This is an agency that has announced nuclear space missions before. The reactor will need to be built, tested, and certified. PPE will need to be modified and integrated. The launch infrastructure for a nuclear reactor is not trivial. Three years is aggressive by any measure. But the architecture is simpler than previous attempts, the timeline is fixed by orbital mechanics rather than budget cycles, and the program has a clear owner. The lack of an operational space nuclear reactor is not a technology problem. It's an execution problem, Sinacore said. We'll find out if he is right.
