Ongoing supply constraints in high-bandwidth memory and data center infrastructure, combined with sustained demand for AI training and inference workloads, are anchoring current H100 rental rates near $3.40–$3.60 per GPU-hour across providers. This environment supports the market's slight tilt toward $3.25+ outcomes by end of November, as hyperscalers and AI labs continue competing for capacity amid extended lead times. Newer Blackwell systems offer limited near-term relief due to production bottlenecks, while spot and reserved markets show modest downward pressure from increased specialist cloud availability. Traders weigh these dynamics against potential demand fluctuations or incremental supply additions that could compress pricing into the $2.75–$3.25 range.