Power and land were never the bottleneck. Compute was. @thejefflutz points to a supply chain idea called postponement. You keep orbital and terrestrial compute hardware as close to identical as possible, and only split them at the end of the build. Nvidia hardware is the most expensive line item. With one design feeding two channels, every GPU can go to whichever channel is ready first, and the ground systems work as a beta test before orbit is ready. Elon expects an orbital compute satellite next year that reuses the V3 design. On Optimus V3, supply chain signals point to a big ramp in engineering builds, and an October update could show production progress. Tesla is also hiring Optimus data collection operators in multiple cities. That looks like groundwork for service center deployment, and possibly interim Robotaxi operations roles. $TSLA $SPCX

Starship made orbit, and SpaceX checked every satellite it put there! @thejefflutz walks through it. There was an orbital insertion burn, then 26 V3 Starlink satellites deployed, with contact made to all 26 and laser links up on 3 shortly after. The bandwidth math is what matters. One launch delivered capacity comparable to Starlink's first year and a half. At about 1 Tbps per V3, that's roughly 26 Tbps from a single flight. He also notes a tense go/no-go call around the 16 to 18 minute mark, and it turned out to be the right one. The heat tile changes are still iterating from flight data. Flight 15 is targeted for around October 19. Elon says hourly Starship flights are 2 to 3 years away. Cadence is the variable now. $SPCX