Elon Musk talks about plans to build catapult to launch from moon

Now that SpaceX has bought the AI company xAI, it seems that SpaceX is leading them into the often-talked about solution to the enormous power problems with AI - data centers in space.

Last week, SpaceX founder Elon Musk advised workers at the newly acquired company xAI that he wants to set up a factory on the moon to build artificial intelligence (AI) satellites. And he called for a colossal catapult on the lunar surface to fling them into space.

"My estimate is that, within two to three years, the lowest-cost way to generate AI compute will be in space," Elon Musk wrote in a Feb. 2 update that announced SpaceX's acquisition of xAI.

While launching satellites with existing technologies like the Falcon family or the Starship, Musk envisions something launched by the planned human outposts on the moon. Factories on the moon could take advantage of lunar resources to produce the satellites, as well as to build mass drivers on the moon to hurl the satellites into their orbits. 

"By using an electromagnetic mass driver and lunar manufacturing," he wrote, "it is possible to put 500 to 1000 TW/year [terawatts per year] of AI satellites into deep space, meaningfully ascend the Kardashev scale and harness a non-trivial percentage of the sun’s power." 
...
Musk isn't the first person to propose the use of mass drivers — which are basically railguns — on the moon. He's following in the footsteps of space visionary Gerard O'Neill, who floated the idea back in 1974.

Railguns may not be as familiar as coil guns, which are numbers of coils aligned along a straight path with the power to each coil pulsed in time to add energy into a payload traveling along the controlled path as the payload reaches each coil. O'Neill worked on the design of mass drivers at the Massachusetts Institute of Technology (MIT), along with colleague Henry Kolm and a group of student volunteers to construct their first mass driver prototype. They eventually concluded that a mass driver only 520 feet long could boost material off the lunar surface.

Artist's illustration of an electromagnetic mass driver launching a payload from the surface of the moon. (Image credit: General Atomics Electromagnetic Systems)

You may have heard of the existing electromagnetic aircraft launch system, now operating reliably on the U.S. Navy's Gerald R. Ford nuclear aircraft carrier. It's the same basic technology as a mass driver.

How far out in the future is a scenario like this? It's going to require a largely self-sufficient colony on the moon that operates at all times - not shutting down for the (two week) night. I'd guess such a colony would be underground for a number of reasons. Starship is the only vehicle being talked about that has capacity to deliver the kinds of numbers of tons of payloads that it would require to build and support such a colony, with the ability to deliver 100 metric tons at a time to the lunar surface. While I like the way Musk refers to "... within two to three years, the lowest-cost way to generate AI compute will be in space," I'd love to see that, but it doesn't look like anything is moving that fast. I think predicting 2035 is going out on a limb.

I stand by my opinion that AI is the biggest hype episode in world history. Everyone acts like there will be one winner and they'll be the one.