Kimi demand surge shows cheaper AI can boost compute demand

vaibhavbetter · x · 2026-07-20

Wall Street’s “cheap Chinese models kill NVIDIA” narrative is flipped by Kimi’s latest situation: subscriptions were paused within 48 hours, GPUs are maxed out, and the team is adding capacity as fast as possible.

The post frames this as a Jevons paradox moment: cheaper intelligence may not lower compute demand — it can actually explode it, translating lower model prices into a bigger CUDA bill.

Original post →

More from Infra

Infra channel →