Persistent supply constraints for NVIDIA's Blackwell B200 GPUs, combined with robust demand for large-scale AI training and inference workloads, continue to support elevated rental rates into late 2026. On-demand medians currently range from roughly $7 to $8.60 per GPU-hour across providers, with spot instances available lower and cheapest listings near $4–5.50, reflecting a 16–23% rise over recent 30- to 90-day periods in several indices. NVIDIA has signaled ongoing bottlenecks through at least early 2028, while neocloud operators and marketplaces compete on availability against higher-priced hyperscaler options. Key near-term catalysts include potential February 2027 list price adjustments and the gradual ramp of next-generation Rubin silicon, which could ease pressure only after meaningful volume arrives.