GPU rental income depends less on the card and more on utilization and price. Here are realistic expectations without inflated promises.
What drives income
The key factors are GPU type, uptime stability, internet channel and your pricing strategy. The same card can earn wildly different amounts.
- GPU class and VRAM size
- Average monthly utilization
- Uptime and network stability
- Pricing strategy
What it means in money
Market rates move, so look at the median for your model rather than the top offer. Below is a snapshot of the vast.ai catalogue from August 2026 (single cards available to rent). On the left is the price the client sees, on the right what the host receives: the platform markup is about 33% on top.
| GPU | Client $/hr (median) | Host $/hr (median) | Host range |
|---|---|---|---|
| RTX 5090 | 0.45 | 0.34 | 0.30 – 0.42 |
| RTX 4090 | 0.40 | 0.30 | 0.27 – 0.45 |
| RTX 5080 | 0.17 | 0.13 | 0.13 – 0.15 |
| RTX 3090 | 0.15 | 0.115 | 0.10 – 0.17 |
Monthly income = host rate × 730 × utilization. For a single RTX 4090 at the $0.30 median that is ≈ $66 at 30% utilization, ≈ $110 at 50% and ≈ $153 at 70%. Disk and traffic come on top. Note: traffic is only billed on docker rentals — vast does not bill traffic for virtual-machine (VM) rentals at all.