Elon Musk Is Making a Moon Catapult: What the Market Is Missing

SpaceX filed it. Not a joke. Not a sci-fi sketch. A magnetic catapult on the moon sat inside the company's own S-1 thinking before it went public: a machine that throws satellites and cargo off the lunar surface without lighting an engine and without burning a drop of fuel.

People hear "moon catapult" and laugh. Then the numbers show up, and the laugh dies.

Leaving Earth is the expensive part

To escape Earth for good, you need about 11.2 km/s — roughly 25,000 mph. That number is why rockets are mostly propellant. A Falcon 9 on the pad is something like 90% fuel by weight. You are launching a flying gas tank with a tiny payload on top.

The moon is a different game. Lunar escape speed is about 2.38 km/s — roughly a fifth of Earth's. Energy scales with speed squared. Cut the speed by a factor of five and the energy per kilogram falls by about a factor of 22.

Throwing one kilogram off the moon takes about 0.79 kWh. That is less electricity than running a microwave for an hour. Escaping an entire world's gravity for the price of reheating leftovers. Hold that number. Everything else in this story hangs on it.

Why the moon, and why not Earth

Three facts make the moon the natural home for this machine.

First: that 22x energy advantage. Second: surface gravity is about one-sixth of Earth's, so the track and structure fight a much gentler pull. Third — the decider — the moon has no atmosphere.

On Earth, flinging something off the ground at thousands of miles per hour means the air hits like a wall. One serious Earth-based proposal put nose heating around 30 kW per square centimeter: a blowtorch the size of your fist, thousands of times hotter, slamming the tip of your craft the instant it leaves the launcher. On the moon you fire straight into vacuum. Nothing in the way. You just go.

Shallow gravity well. One-sixth gravity. No air. The moon was built to be a launchpad. We've treated it like a destination. It is also infrastructure.

The machine is a maglev that lets go

If you know maglev trains, you already know the core idea. Magnets float a carrier above the rails and push it with a magnetic wave. No wheels. No onboard engine.

A mass driver is that track turned into a launcher. The carrier — engineers call it a bucket — holds the payload, rides the magnetic wave, hits release speed, drops the cargo, then peels off, loops back, and reloads. The payload coasts into space. The bucket stays for the next shot. No fuel burned. Just electricity and magnets. In principle the bucket never touches the rail, so friction wear is almost zero. Fire it a million times. That is the dream: a launch system with no rocket and almost nothing to wear out.

Built in 1977 out of scrap and car batteries

This is not a new idea. In May 1977, a small MIT lab bolted together about $2,000 of scrap parts, roughly 20 copper coils, and a bank of ordinary car batteries. They flipped the switch. The coils fired in sequence. A little bucket shot down the line at about 40 m/s. They called it Mass Driver One.

The mind behind it was Gerard K. O'Neill, a Princeton physicist who had already invented the particle storage ring that became fundamental to accelerators. In the early 1970s he asked his students a blunt question: is the surface of a planet really the right place for an expanding technological civilization? He thought no. Build giant habitats in space. Mine the moon. Fling the material out with an electromagnetic catapult. Catch it. Build with it. He laid that out in a 1974 paper and a 1976 book, The High Frontier.

His team did not stop at one machine. Mass Driver One pulled about 33 Gs. A few years later at Princeton, Mass Driver Two hit around 500 Gs. Mass Driver Three, around 1980, pushed past 1,800 Gs. Working prototypes. Almost 50 years ago. Scrap metal and car batteries.

Track length is a trade with how hard you push

Harder acceleration means a shorter track. Gentler acceleration means a longer one. Distance needed is speed squared over twice the acceleration.

Slam the payload at 1,000 Gs and the track to lunar escape speed is only about 289 meters — roughly two football fields. At 1,800 Gs, the number O'Neill's third machine actually hit, you are looking at about 160 meters. Less than a city block for a launch system that can throw cargo off an entire world.

Catch: you can only push that hard on stuff that will not break. Rock does not care about 1,000 Gs. A delicate satellite might take tens of Gs. A human can handle maybe two or three before the body fails. Soften the ride to a few Gs and the track jumps from 160 meters to hundreds of kilometers. Same physics. Machine roughly 5,000 times bigger. The length of your catapult is set by the most fragile thing you ever want to throw.

For dumb mass — regolith, oxygen, construction material — the catapult stays tiny. That is where the industrial case lives.

Average power is cheap. Peak power is the monster.

Scale the microwave number. Throw 10,000 metric tons off the moon every year and the average power sits around 1 to 2 MW. A single modern wind turbine puts out a few megawatts. The freight bill looks almost absurdly small.

So why doesn't this exist? Because average power is not the hard part. Peak power is. A mass driver does not sip electricity. It fires in violent pulses, dumping a huge jolt in a fraction of a second, again and again.

One 2026 engineering study by Casey Handmer put a large lunar driver at roughly 450 MW average with peak demand around 16 GW — more than a dozen large nuclear plants, needed for an instant. Handmer compared the pulsed-power problem to building ITER. The track is the easy part. The capacitor and flywheel plant that trickle-charges and then dumps everything in a flash is the monster.

The moon stops being a rock and becomes a commodity

Mechanically, a mass driver throws things. Economically, it rewrites what the moon is worth.

Right now, getting mass off the moon means rockets and propellant. A mass driver deletes the rocket. Cost of throwing a ton stops being a fuel bill and becomes an electricity bill — kilowatt-hours from a reactor or a solar array, already 22 times cheaper than Earth, with no atmosphere in the way.

The moon stops being an expensive destination and becomes the cheapest place in the solar system to get mass moving. You walk into a space-launch story and walk out into a commodity story: who controls the cheapest mass in the inner system.

Lunar dirt is roughly 40 to 45% oxygen by mass. Oxygen is the heavy majority of rocket propellant. Pull it out of regolith, fling it into orbit with a catapult, and you have built a gas station in space. Spacecraft stop getting topped up with oxidizer hauled from Earth and start getting topped up with lunar oxygen priced like electricity.

Water ice sits in polar craters that never see sunlight. NASA confirmed it in 2009. Estimates range from roughly 600 million to about 2 billion metric tons — treat that as a wide band, not a hard figure. Even the low end is staggering. Split the water and you get hydrogen, oxygen, fuel, and breathable air.

Dust, night, and a chicken-and-egg trap

Lunar dust is not Earth dust. The grains are jagged, like microscopic broken glass, and electrostatically charged so they cling to everything. Apollo gear started degrading after hours. Seals bogged down. Astronauts described it as an abrasive grinding away at whatever it touched. A mass driver is kilometers of close-tolerance magnets and precision surfaces that have to run for years in an environment that chewed through Apollo hardware in a workday.

Temperature swings from about 121°C in the day to -133°C at night. Lunar night lasts about 14.5 Earth days, so pure solar vanishes for two straight weeks. Raw radiation. Micrometeorites. This is the most hostile factory floor humans have ever tried to design.

Then the chicken-and-egg problem. A catapult only pays off if there is already a mountain of mined, refined, packaged material waiting to throw. Cheap transport is part of what creates demand for that material. You almost certainly cannot lead with the catapult. It is one of the last machines in a mature lunar industry, not the first. Early cycles lean on machinery hauled up from Earth the expensive way.

Power makes the gap concrete. In summer 2025 NASA pushed toward a 100 kW fission reactor on the moon, targeting a launch around 2030, with a formal DOE agreement in January 2026. China is chasing its own lunar reactor in the early-to-mid 2030s. Important. Real. And still kilowatt-scale. A serious mass driver wants hundreds of megawatts continuous and gigawatts of peak — three to four orders of magnitude above anything funded today. Up to 10,000 times more power.

The order this actually has to happen

If the catapult becomes real, the sequence is not optional.

First, Starship has to land cargo and crews on the moon routinely. SpaceX already holds NASA's primary crewed lunar lander contract, worth around $4 billion. Second, surface power has to scale from that 100 kW reactor toward hundreds of megawatts. Third, industrial mining and refining of regolith. Fourth, manufacture track segments and magnets from local metal — shipping a 160-meter precision machine from Earth defeats the point. Fifth, robots that can build and align the machine to sub-millimeter tolerance. Only then do you build the catapult.

The day there is a real lunar industry with mines, refineries, and power, a mass driver becomes almost inevitable because the economics are too obvious to ignore. The open question is when. It might be decades. It might never happen. Nothing in the physics blocks it. The bottleneck is whether civilization decides to do the brutal, decades-long work of building an industrial base on another world.

The dream of flinging cargo off the moon without a rocket has sat fully formed on a shelf for half a century. Elon Musk put a version of it in SpaceX's roadmap, backed by the most capable rocket company on Earth and fused with an AI and robotics stack that can eventually build what humans cannot. For the first time, the forcing function is not a paper. It is a company that ships hardware.

Check the video here.