ALTMAN SAYS HOME AI INFERENCE WASTES SCARCE DRAM AND THE CLOUD IS THE BETTER USE OF IT

Summary from the TLDR Hardware newsletter

Sam Altman says that with silicon capacity constrained, it's poor global allocation for individuals to each hold a slice of the world's limited DRAM to run local models, and that utilization is highest in the cloud.