Lunar mass driver

What we know so far

• On 2 February 2026, in the update announcing SpaceX's acquisition of xAI, Musk wrote that an electromagnetic mass driver and lunar manufacturing could put 500 to 1,000 terawatts a year of AI satellites into deep space.^[4]
• At an xAI all-hands meeting in February 2026, whose video the company posted on X, Musk said the company should build a factory on the Moon to make AI satellites and a large catapult to launch them.^[3][5]
• Musk told the meeting We're actually going to have a mass driver on the Moon, while acknowledging that such a thing was for now in the realm of science fiction.^[10]
• As of 16 February 2026, Musk had given no schedule for his lunar plans, and no source records a design, contract, or hardware for the launcher.^[2]
• Independent researcher Keith Sadlocha estimated, in a Fast Company report of 20 February 2026, that such a system would need 8.7 to 20 megawatts of continuous power and could take 5 to 15 years to deploy after serious investment.^[6]
• The only mass driver hardware on record is Mass Driver 1, a small prototype that Gerard K. O'Neill and Henry Kolm built at MIT in 1976 and 1977.^[6][9]

What it is

A proposed electromagnetic launcher for sending material from the Moon into space^[1]

Musk's proposal

Launching AI-related satellites from a lunar factory^[2][3]

First public description

SpaceX's 2 February 2026 update on acquiring xAI, then an xAI all-hands meeting^[4][5]

Schedule

None announced for Musk's proposal^[2]

Historic prototype

Mass Driver 1^[9]

Readiness rating

NASA Technology Readiness Level 5, as rated by independent researcher Keith Sadlocha^[6]

What Musk and SpaceX have said

At an xAI all-hands meeting in February 2026, Elon Musk said a lunar mass driver would launch satellites from a factory, describing it as a maglev train that sends payloads into the solar system.^[2][7] He also acknowledged that the technology remained within the realm of science fiction.^[10]

Musk had first written about the idea in SpaceX's 2 February 2026 update announcing its acquisition of xAI, where he said an electromagnetic mass driver and lunar manufacturing could put 500 to 1,000 terawatts a year of AI satellites into deep space.^[4] Space.com dates the all-hands meeting to 11 February 2026, the day xAI posted its video; Engadget reported on it that morning from audio heard by The New York Times.^[3][5]

The lunar mass driver is a proposal, not an operating project.^[2] Musk gave no schedule for any of these lunar ambitions.^[2]

On 8 February 2026, Musk said on X that SpaceX could complete a self-growing lunar city in less than 10 years.^[3][2] That claim concerned the wider city concept and set no construction or launch schedule for the mass driver.^[2]

How a mass driver works

A mass driver accelerates a payload with a sequence of magnetic fields instead of chemical propulsion.^[1] In the coilgun form described for the lunar proposal, energized coils push a carrier along the track without touching it.^[1][8]

Space.com likened mass drivers to railguns, but other technical descriptions class them as coilguns and favor a maglev or coilgun design that avoids a railgun's physical contact.^[5][1][8] The two are related electromagnetic approaches, but the terms are not technically interchangeable.^[1]

In May 2026, Casey Handmer calculated that a system modeled at 10 million metric tons of raw rock a year would need about 450 megawatts of kinetic power, at a stated driver efficiency of 90 percent.^[8] Launching material at 1.6 kilometers per second, he estimated, would take roughly the velocity change needed to reach low lunar orbit, with solar-powered tugs handling later transfers.^[8]

History of the idea

Gerard K. O'Neill worked on mass drivers in the 1970s and proposed using them to move material from the Moon into space.^[5] With Henry Kolm and students at MIT, he built Mass Driver 1 in 1976 and 1977.^[9][1] The model used about 20 drive coils, carried a small bucket, and reached an acceleration near 30 times Earth's gravity on a budget of about $2,000.^[9]

1:34 Mass Driver 1 The Space Studies Institute's archive film of Mass Driver 1, the prototype O'Neill's group built at MIT in the 1970s.

O'Neill, Kolm, and student researchers proposed a lunar driver several kilometers long that could move 600,000 metric tons a year toward an Earth-Moon Lagrange point.^[5] Prototypes backed by the Space Studies Institute also demonstrated that a mass driver 160 meters long could accelerate material off a lunar surface.^[5]

An illustration shown at the February 2026 xAI meeting resembled the lunar mass driver concepts O'Neill had proposed about 50 years earlier.^[2] Musk's 2026 proposal therefore revisited an older approach to lunar launch rather than introducing a new type of launcher.^[2][5]

Feasibility

In estimates reported by Fast Company in February 2026, independent researcher Keith Sadlocha calculated that a driver for Musk's purposes would need 8.7 to 20 megawatts of continuous power, 400,000 to 634,000 square feet of solar arrays, and about 362 metric tons of hardware.^[6] He put the track's length at 1,620 to 5,350 feet, with payloads pushed at 30 to 400 times Earth's gravity.^[6]

Sadlocha rated lunar mass driver technology at NASA Technology Readiness Level 5 and said deployment could take 5 to 15 years after serious investment, possibly extending into the 2040s.^[6] Lluc Palerm of Analysys Mason told PC Magazine that the lunar server plan posed a challenge comparable to a Mars mission.^[6]

Writing on his blog in May 2026, Casey Handmer said pulsed-power infrastructure could dominate the mass and cost of a system with a peak output of 16 gigawatts.^[8] A catcher handling about 56,000 launches a day, he added, would require demanding vehicle coordination, and braking the empty sleds could imply a reactor of 700 to 800 megawatts.^[8]

Robert Peterkin of General Atomics Electromagnetic Systems said in February 2026 that electromagnetic launch could draw on abundant lunar solar energy instead of imported chemical rocket fuel.^[5] His 2023 report for the Air Force Office of Scientific Research said that delivering 100 metric tons to the lunar surface by Starship could enable a lunar electromagnetic launch system.^[5]

What exists today

Engadget noted in February 2026 that no one had landed on the Moon in more than 50 years and that no lunar colony or factory had ever been built, both steps that would have to come before any lunar launcher.^[3] Musk had given no schedule for his lunar plans.^[2]

The documented hardware on the ground is Mass Driver 1, the student-built prototype constructed at MIT in 1976 and 1977.^[1][9] A later account describes Mass Driver Two, with a nominal design acceleration of 5,000 meters per second squared and a final velocity of 112 meters per second, but it never became a lunar installation.^[1]

Timeline

1. 1974 Gerard K. O'Neill proposes lunar mass drivers, according to Space.com.^[5]
2. 1976 Construction of Mass Driver 1 began at MIT under Gerard K. O'Neill and Henry Kolm.^[9][1]
3. 1977 Mass Driver 1 demonstrated acceleration near 30 g.^[9][1]
4. 1982 An IEEE paper described Mass Driver Two's nominal acceleration and final velocity.^[1]
5. 2023 Robert Peterkin filed an Air Force Office of Scientific Research report on lunar electromagnetic launch.^[5]
6. 2026-02-02 SpaceX's update announcing its acquisition of xAI names an electromagnetic mass driver and lunar manufacturing.^[4]
7. 2026-02-08 Musk said on X that SpaceX could complete a self-growing lunar city in less than 10 years.^[3][2]
8. 2026-02-11 Space.com dates the xAI all-hands meeting at which Musk described a lunar factory and mass driver to this day.^[2][3]
9. 2026-02-20 Keith Sadlocha presented estimates for the power, hardware, and deployment time of a lunar mass driver.^[6]
10. 2026-05-08 Casey Handmer publishes calculations and engineering concerns for a lunar mass driver on his blog.^[8]