The quoted post says Terafab aims to produce 1 TW of compute annually, roughly 50× current global AI compute output. The image compares that projected Tesla & SpaceX demand with much smaller available new compute supply, suggesting demand would exceed global chip supply.