Data Center Watch: Token Spend Growth Accelerates in October Amid Mixed GPU Rental and Memory Pricing

In the Data Center Watch report, we outline key data points relevant to AI infrastructure demand, including LLM token volumes and prices, GPU rental prices, and memory costs. Taken together, the latest data on LLM token spend growth suggest AI infrastructure demand remains strong, but pricing signals from GPU rentals and memory spot prices were more mixed. LLM token spend growth accelerated to +40% m/m in October from +28% m/m in September, driven by strong volume growth (+25% m/m, 33x y/y) and a rebound in volume-weighted pricing (+12% m/m, -38% y/y), primarily reflecting a mix shift toward higher-priced models, particularly Claude Opus 5.5, which added ~7.9 tn tokens at a premium price, as well as same-model repricing. Meanwhile, sequentially, GPU rental prices declined further for A100s (-3.5% m/m to $1.53/GPU-hour), while H100 pricing rebounded (+6.5% m/m to $2.83) and B200 price growth accelerated (+4.7% m/m to $6.00), with the B200-to-H100 ratio at 2.12x and the H100-to-A100 rati