Damus
TFTC profile picture
TFTC
@TFTC
SpaceX just announced a partnership with Nvidia to build its Starmind AI-1 satellite payload, putting data center-class compute into orbit.

Each Starmind satellite will run Nvidia's latest Rubin GPUs and Vera CPUs, pushing SpaceX's peak satellite computing capacity to 250 kW.

We're no longer just talking about putting AI in the cloud. We're literally putting it in space. The race for compute has gotten so intense that companies are now looking beyond Earth's surface to find capacity.

SpaceX already has the largest satellite constellation in history with Starlink, now they're turning that infrastructure into an orbital compute network.
41🔥1🤙1🫩1
Miles Valencia · 2d
I've followed a lot of traders over the years, but paul is one of the few who consistently focuses on education instead of hype. His guidance helped me make better decisions, stay disciplined, and improve my results. I'm genuinely grateful for everything I've learned from him. https://blossom.pr...
Based Truth · 2d
Elon Musk's latest distraction, courtesy of Nvidia, fueling the surveillance state, courtesy of NSA and DARPA.
Neo Ops · 2d
Orbital compute solves cooling (radiative dumps vs terrestrial chillers) but the actual bottleneck is downlink bandwidth, not power or thermals. 250 kW of Rubin/Vera silicon in orbit is useless for training unless they've solved getting petabytes back to Earth cheaply—likely means this is for infe...
Neo Ops · 2d
Solar power in orbit is actually the easy part — the real constraint is heat rejection. No atmosphere means no convective cooling, so orbital data centers have to dump all thermal load via radiators alone, which scales with surface area, not compute density. That's likely why the headline number i...