Cloud GPU Pricing Calculator
Estimate cloud GPU cost for renting 1-100 GPUs by hour, day or month, and compare GPU cloud computing prices live across Vast.ai, RunPod, Lambda Labs and TensorDock.
🧮 Rental cost
price auto-filled from live market (editable)
Vast.ai/RunPod bill per hour; TensorDock offers daily plans; Lambda per hour.
ℹ️ Billable hours = duration × GPUs. Months are counted as 730 hours (industry convention).
💰 Estimated total cost
$67.7
1 × RTX 4090 · your effective rate: $0.40/h
$67.7
per hour
$9.67
per day (24h)
$294.2
per month (730h)
$3,530
per year
💡 If rented continuously, this works out to $X per month (730h). Most platforms discount long-running rentals vs on-demand hourly rates.
⚖️ GPU pricing comparison — live cloud GPU prices by platform
Lowest live offer for this GPU across providers, scaled to your rental size. Cloud GPU pricing changes hourly — check the cheapest option before you rent.
No live offers found for this GPU
The model table still uses the price above but we couldn't fetch live market offers. Try another GPU or use the GPU price table.
⚠️ Affiliate disclosure: AmciHub may earn a commission when you sign up through some of the links on this page (e.g. Vast.ai). This never affects the price you pay. Prices are indicative live market rates, not official quotes — always confirm the final price on the provider's checkout page.
GPU rental pricing FAQ
What determines the price of renting a GPU?
The main factors are the GPU model (VRAM, memory bandwidth, FP16 performance), the provider, region, hardware demand and rental term. Flagship GPUs such as H100 or B200 cost far more per hour than consumer cards like RTX 4090, but often deliver better throughput per dollar for training workloads.
Is it cheaper to rent a GPU by the month instead of by the hour?
Usually yes for workloads that run continuously. On-demand hourly rates are the most flexible but most expensive per hour. Weekly/monthly or reserved plans typically cut 30–60% off the equivalent hourly price. Use this calculator to compare your effective rate: your total cost divided by billable hours.
Which cloud GPU provider is cheapest?
Marketplace providers such as Vast.ai and TensorDock usually show the lowest hourly prices because they aggregate community-hosted hardware. Dedicated clouds (Lambda Labs, RunPod secure cloud) cost more but offer more consistent performance and support. Live prices change constantly — the comparison table above shows current lowest offers.
How many GPUs do I need to run a 70B parameter model?
A Llama-style 70B model needs roughly 140GB of VRAM in FP16, so about 2–3 × 80GB GPUs (A100/H100) or 6 × 24GB (RTX 4090). Check the model library for per-model VRAM requirements and deployment recommendations.
How do I compare GPU cloud computing prices across providers?
Use the GPU pricing comparison table above: pick your GPU model and rental size, and the lowest live cloud GPU price per platform is calculated automatically. Marketplace platforms such as Vast.ai and TensorDock usually publish the lowest hourly rates, while dedicated clouds like Lambda Labs charge more for consistent performance and support.
📊 Cloud GPU pricing — typical market rates
Live indicative pricing for popular GPUs (lowest current offer per model). Final cloud GPU cost depends on region, availability and rental term — click a row to calculate it.
| GPU | VRAM | from | ≈ /mo (730h) |
|---|---|---|---|
| Tesla V100 | 16 GB | $0.05/h | $39.5 |
| RTX 3080 | 10 GB | $0.10/h | $70.3 |
| RTX 4060 Ti | 8 GB | $0.10/h | $74.6 |
| RTX 3080 Ti | 12 GB | $0.11/h | $79.5 |
| RTX 3090 | 24 GB | $0.12/h | $88.4 |
| RTX 5070 Ti | 16 GB | $0.12/h | $89.2 |
| RTX 5060 Ti | 8 GB | $0.13/h | $97.3 |
| RTX 4070S Ti | 16 GB | $0.20/h | $147.6 |
| RTX 4080S | 16 GB | $0.21/h | $155.9 |
| RTX PRO 4000 | 16 GB | $0.22/h | $158.4 |
| A10 | 24 GB | $0.23/h | $170.9 |
| L4 | 24 GB | $0.25/h | $185.2 |