Cloud GPU Hourly Price Comparison (H100, A100, RTX 4090)
Compare on-demand GPU instances and claim free developer cloud credits for self-hosting LLMs, Stable Diffusion, and AI agents.
On-Demand Cloud GPU Pricing Matrix
| GPU Instance | VRAM | On-Demand Rate | Est. Speed | Best Workload | Deploy |
|---|---|---|---|---|---|
| NVIDIA RTX 4090 (24GB VRAM) | 24 GB | $0.39/hr ($281/mo) | ~85 tok/s | 7B to 14B models & Q4 quantized 32B models | Rent GPUs on Vast.ai (Save up to 70%) |
| NVIDIA A100 (80GB VRAM) | 80 GB | $1.39/hr ($1,001/mo) | ~65 tok/s | 32B to 70B models (Llama 70B, Qwen 72B, Mixtral) | Deploy on RunPod (From $0.20/hr) |
| NVIDIA H100 SXM (80GB VRAM) | 80 GB | $2.69/hr ($1,937/mo) | ~140 tok/s | High-concurrency production LLM inference & 70B+ models | Deploy on RunPod (From $0.20/hr) |
| 2x NVIDIA H100 (160GB VRAM) | 160 GB | $5.38/hr ($3,874/mo) | ~110 tok/s | 100B+ & MoE frontier open-weight models (DeepSeek R1/V3, Llama 405B) | Deploy on RunPod (From $0.20/hr) |
Verified Cloud GPU & VPS Partner Offers
Best for Serverless & Dedicated AI GPUs
RunPod Cloud GPUs
Instant PyTorch, vLLM, and Ollama GPU pods plus auto-scaling Serverless GPU endpoints billed by the second.
Deploy on RunPod (From $0.20/hr) →
Cheapest RTX 4090 & H100 Rentals
Vast.ai GPU Marketplace
Global GPU rental marketplace offering the lowest hourly spot and on-demand rates on RTX 4090, A100, and H100 rigs.
Rent GPUs on Vast.ai (Save up to 70%) →
$300 Free Credit for New Users
Vultr Cloud GPU & VPS
Enterprise-grade NVIDIA H100, A100, and L40S cloud instances across 32 global data centers with $300 in new-user test credits.
Claim Free Vultr GPU Credit →
$200 Free Credit for 60 Days
DigitalOcean GPU Droplets
Developer-friendly NVIDIA H100 GPU Droplets, managed databases, and AI app hosting backed by a 60-day $200 free credit.
Get $200 Free Credit on DigitalOcean →
B2B Sponsorship Slot
📣 Sponsor a Featured GPU Tier →
Run an AI Cloud, GPU Marketplace, or Inference API?
Feature your H100 / A100 / serverless GPU pricing across 1,550+ AI cost & VRAM calculator pages and our local tokenbench CLI.