Colonization.si ← All explainers
Robotics · Explainer

Will robots build the first Moon base?

NASA, China and SpaceX all plan to send machines ahead of people. We look at what those machines would need to do, who is building them, and how much of it has actually been tested.

Colonization.si8 October 202610 min read
Short answer

Robots are planned to do some of the first jobs: every current plan assigns them digging, hauling and site preparation. A base built mostly by robots before people arrive is a different claim, and it's unproven. As of October 2026, no robotic construction machine has operated on the Moon.

Why send machines first

People are expensive to keep alive on the Moon. They need air, water, food, radiation shelter and a ride home, and every hour they spend leveling ground is an hour of life support. A machine needs power and maintenance, and it can keep working for months without anyone else on site.

The jobs that come first are also the most repetitive. A lander arriving near a base kicks up regolith at high speed, so landing pads need to be graded or paved. Habitats need to be covered or surrounded with regolith for radiation shielding. Power cables need to be laid and solar arrays or a reactor set up. Those are earthmoving jobs, and earthmoving is what robots are being designed to do.

That logic shows up in every major plan:

The Moon is close enough to drive from Earth, mostly

Radio signals take about 1.3 seconds to reach the Moon, so a command and its result take about 2.6 seconds round trip. That is slow but workable. The Soviet Union proved it more than 50 years ago: Lunokhod 1 in 1970 and Lunokhod 2 in 1973 were driven by crews on Earth, and Lunokhod 2 covered about 39 km.

Driving a rover slowly is one thing. Running an excavator all day is another. Relays and processing add to the light delay, and NASA's IPEx team has said the real-world delay of several seconds makes continuous joystick control nearly impossible for a long digging job. So the plan is supervised autonomy: people on Earth choose what to dig and where to dump it, and the robot handles each scoop and each turn by itself.

Mars is a different problem. Depending on where the planets are, a signal takes about 3 to 22 minutes each way, so a robot on Mars has to handle much more on its own. That's why the Moon is the natural place to learn robotic construction first. Try the delay yourself:

Try it

Drive a rover with a time delay

There's a rock ahead. Send "forward" commands and stop before you hit it. Each command arrives after the real round-trip light delay.

Ready. The rover moves one step each time a command arrives.

The jobs, and who is working on them

Here is where each piece stands as of October 2026. None of them has yet been demonstrated on the Moon.

JobMain effortStatus
Digging and haulingNASA IPEx, an excavator based on the RASSOR design, built to dig and carry regolith semi-autonomously. A demo goal is moving about 10 tonnes.Ground prototypes. No flight date announced.
Building with regolithICON's Olympus (NASA contract of about $57M, 2022) and China's Chang'e 8 brick-printing experiment.Olympus: ground testing. Chang'e 8: 2028–29.
Power through the nightNASA Fission Surface Power: a 100 kW reactor that NASA wants ready to launch by about the end of 2029.Design phase. Congress funded $250M for 2026, against an estimated $3B total.
DeliveryCommercial landers. Firefly's Blue Ghost landed successfully in March 2025, while Intuitive Machines' Athena tipped over that same month. Blue Origin's large Blue Moon MK1 has passed NASA vacuum testing.MK1 first flight no earlier than early 2027.
General-purpose laborHumanoid robots: Musk has pointed to Tesla's Optimus, and reports suggest China may fly a humanoid on Chang'e 8.No humanoid has worked on the Moon.

What makes it hard

The lunar night. Most of the Moon goes about two weeks without sunlight, and surface temperatures drop below −170 °C. Many landers, including every US commercial lander so far, haven't survived their first night. China's Chang'e 4 has lasted years, partly by using nuclear heaters. A robot that has to sit out half of every month needs either a lot of stored energy or a reactor nearby. That's why fission power keeps coming up in base plans.

Dust. Lunar regolith is sharp, because there's no wind or water to round its grains, and it clings to surfaces through static charge. Apollo crews found that it got into suit joints and seals after just a few days. A digging machine lives in that dust full time, so wear on joints, bearings and seals may limit how long robots last more than software does.

Repairs. On Earth, a broken excavator gets a mechanic. On the Moon, the nearest mechanic is days away and costs millions to send. Robots would need to be simple, redundant, or able to repair each other, and none of those approaches is proven at scale.

Getting there in one piece. Recent landers have a mixed record. Intuitive Machines' first two landers and Japan's SLIM ended up tipped over, and ispace's first two attempts crashed. A robot construction crew is only useful if it lands upright.

Where better AI comes in

This is where the superintelligence angle matters. With more capable AI, robots could need fewer instructions from Earth, recover from problems on their own and handle unfamiliar terrain. On the Moon, that could turn a robot that needs a team of operators into one that a single person supervises. On Mars, it may decide whether robotic construction is possible at all.

Smarter software doesn't fix worn bearings, dust in a seal or a dead battery, though. For the first lunar construction robots, physical durability and power look like bigger limits than intelligence. Better AI could help most once the hardware has proven it can survive.

Verdict

Robots first: very probably, for some jobs. Every serious plan gives robots the early digging, hauling and site preparation, and the Moon is close enough for people on Earth to supervise them.

Robots building a whole base before people arrive: unproven. No construction robot has worked on the Moon yet, power through the lunar night isn't solved, and dust and repairs are open problems. The more likely first base is one built by robots and people working together, with robots doing most of the heavy, repetitive work.

What to watch

Sources and further reading