Elon Musk has a bold new plan for space-based infrastructure.

During a weekend presentation at Giga Texas, the tech billionaire unveiled Project TERAFAB, a plan to relocate the world’s most power-based AI infrastructure to the lunar surface. 

The goal is to shatter the “terawatt ceiling” of Earth-bound computing and usher in a petawatt-scale era of intelligence.

Musk’s latest vision involves overcoming Earth’s energy constraints by building electromagnetic mass drivers on the moon to facilitate this massive-scale AI infrastructure. 

It would use the moon’s vacuum environment, low gravity, and solar power. 

Benefits of moon

Earth has less room and power for large AI projects. To build a massive AI network, experts plan to tap the moon’s unlimited space and energy.

Solar panels in space are roughly five times more efficient than on Earth, where clouds and a thick atmosphere reduce efficiency. 

Instead of using more rockets, the plan relies on a giant electromagnetic track to fling satellites into space using the Sun’s energy.

These tracks, potentially stretching dozens of kilometers, would use magnetic fields to accelerate AI-packed satellites to lunar escape velocity.

Notably, the strategy eliminates the need for expensive, volatile chemical propellants and the discarded booster stages that clutter Earth’s oceans.

Furthermore, this plan could use SpaceX’s Starship to transport initial hardware, eventually establishing a distributed orbital network that provides 1,000 times the power of current systems.

If successful, the deployment of AI infrastructure in space would reduce competition between data centers and cities for Earth-based power, enabling AI operations to scale within the Moon’s vast, unconstrained energy environment.

It could take decades

Although the concept of a magnetic space gun has been around since Edward Fitch Northrup first theorized it in 1937, it remains purely theoretical, having yet to leave the drawing board and reach the lunar surface.

Musk’s proposal for a lunar mass driver is a key component of a broader strategic shift announced in February 2026, where SpaceX prioritized a “self-growing” lunar city over immediate Mars colonization.

While Mars remains humanity’s long-term life insurance policy, the moon is now humanity’s immediate industrial priority. 

Logistically, the moon offers launch windows every 10 days, compared to the 26-month wait for Mars.

Skeptics, however, are already pointing to the staggering math. Critics suggest that launching over a million tons of material needed to reach a petawatt of power would require roughly 135 Starship launches a day.

“We are going to push the limits of physics in compute and do some wild and crazy things,” Musk said.

“I want to live long enough to see the mass driver on the moon,” he added.

Musk’s other ambitious vision is to make a lunar factory that uses robotics to manufacture solar-powered AI satellites, which are then launched into a distributed orbital network via electromagnetic mass drivers.

Two primary designs are under consideration: railguns, which use a single powerful burst of force, and coilguns, which employ a sequence of timed magnets to provide steady, controlled acceleration. The latter is said to be the preferred choice for protecting sensitive AI cargo.