I have argued myself that CUDA is a moat, but it's a fragile one; a tensegrity moat. It relied entirely on Nvidia hardware being available in China at scale for top 2-3 teams. You *could* have "everyone talented stays in Jensen's shed, CANN stagnates" condition. No longer true.
the smartest pushback i keep getting: chips were the easy part, the real moat is the software stack, fifteen years of cuda and tuned kernels. honestly, correct. but a software moat is not a fab you have to build, it's a grind you have to out engineer, and engineers are the one thing china has in genuine oversupply.
longcat didn't just run on domestic chips, the whole stack under it was domestic end to end. the hard wall already has a door in it. and once the stack stops being american, the flag on the silicon stops mattering at all.
that was my whole point. cheap compute has no geography.
