Dagogo, I genuinely enjoy your videos — always well-made and interesting.
But you’re totally wrong about SpaceX on both videos.
For orbital data centers: 3-year-old chips are not useless for inference. That’s exactly what they’re doing with the data centers in space. Inference doesn’t require the absolute latest training silicon, and the orbital environment (radiative cooling to vacuum, constant solar potential, no terrestrial power or water limits) changes the economics dramatically. You can’t upgrade chips once they’re up there, but that’s not the killer problem it sounds like — constellations evolve with new launches the same way Starlink does.
This is the same story as Tesla. For years it was never “invest worthy” to people who couldn’t see the full vision. Then it executed, changed the world, and suddenly everyone saw it. SpaceX has a long track record of making the “too ambitious/risky” bets actually work.
The AI integration into the stack isn’t some desperate pivot — it’s a coherent long-term infrastructure play. The risks are real, but writing it off as a long-term loser underestimates them again.