Current A100 rental rates sit near the $1-per-GPU-hour equilibrium because abundant secondary-market supply on platforms such as Vast.ai, RunPod, and Jarvislabs continues to offset steady demand for inference and mid-tier training workloads. On-demand listings range from roughly $0.89 at the low end to medians near $1.80 across dozens of providers, while spot and community capacity frequently trades below $0.70; these figures have remained largely flat over the past 90 days as users migrate toward higher-throughput H100, H200, and Blackwell GPUs. The narrow spread between the $0.75–1.00 and $1.00–1.25 buckets reflects trader uncertainty over which exact benchmark—median on-demand, index-settled transaction price, or provider-weighted average—will determine resolution at month-end. No major capacity shocks or demand surges have materialized in the final week of September to shift the balance.