The post says data-center capacity and U.S. electricity—not GPU production—are the main constraints, while GPU availability is expected to increase over the next one to two years as NVDA ramps production. It also names GOOGL, AMZN, MSFT, and SPCX in discussing cloud providers and spare capacity entering the market.