Recent developments show A100 rental rates clustered around $1.00–$1.80 per GPU-hour on marketplaces like Vast.ai, RunPod, and Thunder Compute, with hyperscalers charging premiums above $3. Trader consensus reflects a tight balance between expanding supply—new Blackwell B200 capacity and enterprise upgrades freeing older Ampere cards—and sustained demand for cost-effective large language model inference and fine-tuning. Competitive dynamics hinge on decentralized platforms undercutting traditional clouds through spot and peer-to-peer listings, while energy costs, regional availability, and potential utilization spikes near year-end could push prices either way. The closely matched $0.75–$1.00 and $1.00–$1.25 bands capture uncertainty over whether incremental hardware inflows will outweigh ongoing AI workload growth by late October.