Son vs Musk: Who's Right About Data Centers in Space?

Let's be clear: we're not in the 1950s anymore. Son looked straight at the camera and said Musk's idea of parking data centers in space has zero practical merit. Why? Simple math: launching and maintaining critical infrastructure in orbit is complicated, expensive, and what actually wins you the AI race is having tons of computing power that's accessible, reliable, and — here's the thing — still on Earth where you can actually manage it.

Son's take is more grounded: whoever wins the AI battle will be the one who stacks the most raw computing resources down here on our planet. I get it, "servers in space" sounds cooler than "massive Earth-based infrastructure," but boring tends to turn into actual money.

Now, Son isn't saying space is useless — satellites and telecom deserve to exist. What he's saying is that for the AI race specifically, orbital data centers aren't the answer. It's a fascinating idea that probably costs ten times more and delivers ten times less than solid ground-based infrastructure.

What's really interesting here is that this clash shows two totally different visions: Musk keeps looking higher (literally), while Son stays laser-focused on what works now and scales tomorrow without exploding in the process. In Silicon Valley they call this "vision versus pragmatism," but really it's the same old fight between "wouldn't that be cool" and "but how much does it actually cost."

That said, Son's not right 100% of the time forever. Innovation often comes from weird places. But when you're building the actual infrastructure that's going to run AI for the next decade, you're betting on a ton of processors on solid ground, not bizarre orbital experiments.