Here’s how Elon Musk’s giant moon cannon would actually work

“I really want to see a mass driver on the moon that is shooting AI satellites into deep space,” Elon Musk said last week when he announced his plan to go to the moon. “It’s going to be incredibly exciting to see it happen.” He’s right. I want to see it too, although probably we will both be dead before his vision is realized.

The lunar mass driver—essentially a cannon that uses magnetic power to accelerate an object—is a key component to launch the million satellites Musk wants to put in orbit around the Earth. But Musk wasn’t the first person to come up with the idea. Smarter people than him thought about this in the 1970s as the solution to a key problem for human exploration.

Launching spacecraft from Earth is extremely expensive. Every pound lifted from Cape Canaveral to low Earth orbit costs thousands of dollars in fuel, hardware, and operational complexity. The farther you want to go in space, the more massive and complex the rocket has to be, increasing costs. Chemical rockets must carry their own oxidizer and propellant, which means most of the vehicle’s mass is just fuel to lift more fuel. This tyranny of the rocket equation has strangled space development for seven decades, only slightly eased by the economics of reusable rockets like the Falcon 9.

A mass driver could break that stranglehold by using electricity instead of explosives, turning launches into a utility-scale operation rather than a high-wire act. On the Moon, where gravity is one-sixth of Earth’s and there’s no atmosphere to create drag, this technology could launch payloads at a fraction of the cost—a few dollars per pound in electricity. Compare that to the $1,200 per pound it currently costs to launch a payload on a reusable Falcon 9 rocket.

An elegant design

American physicist Gerard O’Neill and MIT physicist Henry Kolm built the first prototype of a mass driver in 1976 with a $2,000 budget. The Mass Drive 1 could fire objects at 131 feet per second while experiencing 33 times Earth’s gravity. Their next version achieved 10 times greater acceleration with a comparable funding increase. University of Texas researchers subsequently priced a serious version at $47 million for a device capable of launching a 22-pound payload at 13,400 miles per hour.

A mass driver is basically a very long track stretching across the lunar surface, angled gently skyward at its far end. The track is lined from end to end with hundreds of electromagnetic coils, which are simply loops of wire that snap into powerful magnets the instant electricity runs through them. A payload sits inside a magnetizable carrier called a bucket.

To move the bucket, the coils fire in a precise sequence, one after another, each energizing at exactly the right moment as the bucket reaches it, grabbing it forward, then cutting off the instant it passes. The result is a cascade of invisible magnetic hands, each passing the bucket to the next. The bucket never makes mechanical contact with any surface: It is held aloft and guided entirely by the interplay of magnetic fields, which is why these systems have a theoretical operational lifespan of up to millions of launches with negligible wear.

Musk describes it as a large maglev train, the same levitation technology that holds high-speed trains above their rails in Japan or China. But the mass driver reaches much faster speeds than any train on Earth: about 1.5 miles per second, enough to escape the gravity pull of the Moon. To achieve that speed, the mass driver uses two distinct engineered stages. In the first, the coils sit at equal intervals and their electrical timing locks to the bucket’s exact position—each successive push arrives at precisely the right instant, so the force compounds as velocity builds.

In the second stage, the interval between coils progressively widens, which paces the pushes further apart in distance and holds the rate of acceleration constant rather than letting it keep climbing so the increase in acceleration doesn’t destroy the bucket or its cargo. At the terminal end of the track, the bucket releases its payload—a xAI satellite according to Musk’s vision—into space at a minimum speed of 5,300 mph, enough escape the Moon’s gravity. The trajectory that the load follows depends on the position of the Moon at the moment of the launch, following the orientation of the mass driver relative to the space.

Then the bucket gets caught by a braking system, recovered, and sent back to the beginning for the next launch. No combustion. No exhaust. No rocket equation. No problems. 

It’s a beautiful solution. It’s also doable, as O’Neill and Kolm demonstrated practically. According to independent researcher and author Keith Sadlocha, a working lunar mass driver would require a track between 1,620 and 5,350 feet long, operating at accelerations between 30 and 400 times Earth’s gravity in standard operation.

At those forces, only rugged, non-human cargo survives the ride—which is exactly what Musk is planning. Musk has his sights set on manufacturing AI computing satellites on the lunar surface. The system can fire one payload every 10 to 11 seconds. Scaled to Musk’s stated target of one million satellites in orbit, that cadence, sustained continuously, is what makes the economics viable in a way no rocket ever could.

But to accomplish this, you will need a lot of electricity. For Musk’s purposes, system requires 8.7 to 20 megawatts of continuous power, enough to run a small town. Delivering that on the lunar surface requires between 400,000 and 634,000 square feet of solar arrays—somewhere between seven and 11 NFL football fields’ worth of panels according to Sadlocha’s calculations and NASA’s estimates. That’s using solar panel’s with an efficiency of roughly 30%, the figure NASA uses for cells especially designed for space use.

Since the Moon endures two weeks of total darkness every month, this means the mass driver either sits idle for half of every lunar cycle or relies on supplemental power to keep firing through the night. NASA and the U.S. Department of Energy are developing a solution: The Fission Surface Power (FSP) project, which builds on the earlier Kilopower research program to produce compact nuclear fission reactors targeting 10 to 40 kilowatts of continuous output each, capable of running for a decade without refueling.

Each FSP reactor will produce enough electricity to power just a few homes. Bridging the full gap between those modest reactors and a mass driver that demands the output of a small power plant would require deploying them not one or two at a time, but in the hundreds. That is not a technology problem so much as a logistics one—every reactor has to be launched from Earth, landed softly on the Moon, and connected to the grid before the mass driver fires its first payload. The program, however, is still in development and a lunar deployment is not expected before the late 2020s at the earliest. And that’s extremely optimistic, given the constant delays of those nuclear projects and the Moon return plans.

Unrealistic timeline

Scaling from the $47 million 22-pound-launch prototype that University of Texas researchers projected to a working lunar installation capable of launching huge satellites is where you begin to feel just how vast the distance is between a compelling idea and a functioning machine.

Sadlocha estimates that a full lunar mass driver system requires approximately 362 metric tons of hardware. That’s 24 heavy-lift rocket launches worth of components that must be manufactured on Earth, survive a 239,000-mile journey through the void, and be assembled by people wearing pressurized suits on a surface that—bathed by radiation—is extremely hostile to humans.

That surface is coated in lunar dust too, ultra-fine abrasive shaped by billions of years of micrometeoroid impacts into particles with microscopic cutting edges that never dulled, because there has never been wind or rain or any erosive force on the Moon to round them off. It clings electrostatically to visors, suits, seals, and coil windings alike. The payload carriers themselves face thermal melting at extreme velocities, demanding materials that do not yet exist in proven lunar-rated form. You can argue that maybe Musk’s Optimus robots can avoid this, but his robots can barely function on Earth. 

Musk’s stated plan is to mine lunar silicon and oxygen and manufacture the server hardware on the surface—a bootstrapping strategy that, if it works, would reduce Earth-launch dependency over time toward what he called a “self-growing city” capable of rapid expansion from local resources.

The Moon does contain silicon, oxygen, helium-3, and water ice at the poles. But the superconducting coils at the heart of the mass driver require precisely manufactured materials that the lunar industry will not be able to produce in decades. Every critical component rides to the Moon on Starship until that changes. We know that Starship is so behind schedule that it has pushed the first mission back to the Moon from 2027’s NASA projected time to 2028.

Sadlocha rates the technology at readiness level 5 on NASA’s 1-to-9 scale — components validated in laboratories, not yet tested in space. Realistic deployment, his study concludes, will take between 5 and 15 years from the moment serious investment begins. That can take the project into the 2040s, easily. That’s why Lluc Palerm, satellite research director at Analysys Mason, said to PC Magazine that Musk’s lunar server plan carries a magnitude of challenge equivalent to a Mars mission. 

But like we already pointed out, Musk’s timeline is fantasy, or “aspirational” as he qualifies his predictions. The gap between his ambitious renderings and actual functioning hardware remains a dream measured in decades, not the 10 years he’s promising investors before his planned June 2026 initial public offering targeting a $1.5 trillion valuation. Musk is no JFK, and building factories and a mass driver on the Moon is orders of magnitude more complex than just putting boots on the Moon like the Apollo program did. It’s doable, yes. We’ll get the mass driver, eventually. Just not on Musk time.