They claim 12.6 watts/kg. A fully reusable Starship is expected to orbit 100 to 150 tons of cargo, so ~1 to 2 megawatts of compute per launch.
A 1MW datacenter in the US would cost around $10M to build, and consume around $100k/month in electricity, free to the satellite.
A Starship launch is projected to cost around $10M, which doesn't include the satellites. So it looks more expensive than a conventional data center even with optimistic assumptions. It may depend on those becoming more expensive to be practical.
Battery backup for 24/7 operation would roughly double the CapEx for power. Maybe that's still cheaper than space? Land scarcity alone, not to mention social resistance or ecological costs, will tend to favor scaling in space for the longer term.
What will the data-center operations team look like and who is "go for launch" to replace a failed component? Will DCOps be rocket robots and just hang out in space? But seriously what is the real purpose of this and don't say AI. If so those would have to be some incredibly massive never seen before solar panels.
The purpose is to build hype for the SpaceX IPO. And to engineer a convincing way for Musk to fold xAI into SpaceX and use it to release capital from the AI bubble.
https://x.com/seti_park/status/2015114363531866448
They claim 12.6 watts/kg. A fully reusable Starship is expected to orbit 100 to 150 tons of cargo, so ~1 to 2 megawatts of compute per launch.
A 1MW datacenter in the US would cost around $10M to build, and consume around $100k/month in electricity, free to the satellite.
A Starship launch is projected to cost around $10M, which doesn't include the satellites. So it looks more expensive than a conventional data center even with optimistic assumptions. It may depend on those becoming more expensive to be practical.
reply