**Trader consensus on H200 rental prices at end-November 2026 reflects balanced uncertainty between persistent AI demand and expanding supply.** Current on-demand medians hover near $4.50–$4.60 per GPU-hour across marketplaces and neoclouds, with spot rates notably lower, amid a modest 3% rise over the past 90 days and mixed September trends. Sustained large-model training and inference workloads continue to support pricing, yet growing Blackwell (B200/B300) availability, additional provider capacity, and HBM memory allocation dynamics introduce downward pressure. Export restrictions, selective China shipments, and hyperscaler buildouts add volatility without clear resolution by late November. The tight spread across bins underscores how near-term supply ramps or demand spikes could shift the outcome.