Live Cloud GPU Rental Rates & Cluster Cost Calculator
Normalized per-GPU hourly pricing (`$/GPU/hr`), multi-GPU NVLink cluster totals, and real-time datacenter stock across 22 NeoCloud, Bare-Metal, and Hyperscaler providers.
| Provider & Tier | GPU Model & VRAM | Config & Interconnect | Price / GPU / hr | Total Node / hr | Live Stock & Regions |
|---|---|---|---|---|---|
| Loading live cloud GPU rates... | |||||
Frequently Asked Questions — Cloud GPU Rental Pricing
1. What is the cheapest H100 and B200 hourly rental rate right now?
Specialized NeoCloud and Bare-Metal providers currently offer NVIDIA H100 80GB instances starting around $1.66–$2.30/GPU/hr and NVIDIA B200 192GB instances from $3.74–$6.52/GPU/hr, saving 64%–81% compared to standard Hyperscaler on-demand rates.
2. When should I choose SXM5 NVLink over PCIe GPUs?
For multi-GPU distributed pre-training or full-parameter fine-tuning (e.g., 70B+ models across 4× or 8× GPUs), SXM5 with NVLink/NVSwitch avoids gradient synchronization bottlenecks. For single-GPU inference, LoRA fine-tuning, or batch rendering, PCIe offers lower hourly cost.