The suspicious compute slowdown: is workload shifting from training to inference?

pdamodaran · x · 2026-09-13

A market commentator finds the industry's collective "slowdown" agreement suspicious, offering two readings: it may be a tacit admission that compute workload is shifting from unnecessary training to inference, and an admission that the current stack is inefficient and needs improvement from multiple angles — with implications for $QQQ.

Original post →

More from AGI Musings

AGI Musings channel →