GPU Cloud Cost Calculator
Stop guessing. Choose your GPU, set your usage, and instantly see how much you'll spend across every major provider — sorted cheapest first.
Select GPU
Hours per Day
Days per Month
Monthly Cost BreakdownNVIDIA H100 · 8h/day · 20 days
| Rank | Provider | Price / Hour | Monthly Cost | vs Cheapest | |
|---|---|---|---|---|---|
Vast.aiBest Price | $2.50/hr | $400.00 | Cheapest | Start Here | |
| 2 | RunPod | $2.69/hr | $430.40 | +$30.40(+6%) | View |
| 3 | Vultr | $2.99/hr | $478.40 | +$78.40(+15%) | View |
| 4 | Cherry Servers | $3.20/hr | $512.00 | +$112.00(+21%) | View |
| 5 | Latitude.sh | $3.37/hr | $539.20 | +$139.20(+26%) | View |
Best deal: Vast.ai
Only $400.00/month for NVIDIA H100 · 160 hours of compute — save $139.20 vs. Latitude.sh
Pro Tip: Use Spot / Interruptible InstancesSave ~40%
Many providers offer spot or interruptible instances at significantly reduced rates. For workloads that can handle interruptions (batch training, inference jobs with checkpointing), you could pay as little as $240.00/month — that's a saving of $160.00 compared to on-demand pricing at Vast.ai.
Always implement checkpointing before relying on spot instances for long training runs.
How to Estimate GPU Cloud Costs
What affects your bill?
GPU cloud is billed per hour of use. Your total monthly cost is simply hourly rate × hours/day × active days. A single H100 running 8 hours a day for 20 days costs between $400–$540 depending on provider.
On-demand vs. Spot instances
Spot (interruptible) instances can cut costs by 30–60%. They're ideal for training runs with checkpointing. On-demand instances are better for serving or latency-sensitive inference.
Which GPU should I choose?
For LLM inference: H100 or A100 80GB. For image generation: RTX 4090 or RTX 3090 are great value. For large-scale training: H100 or H200 for the best throughput per dollar.
Are these prices accurate?
The prices shown are representative estimates based on publicly available on-demand rates. Actual prices vary with region, instance configuration, and promotions. Always verify on the provider's website before committing.