NVIDIA A40 (48GB GDDR6)
- 48GB GDDR6
- Memory
- 696GB/s
- Memory bandwidth
- Q4 2020
- Released
Price history
What it costs to rent a NVIDIA A40
The NVIDIA A40 is an Ampere-architecture inference NVIDIA GPU with 48GB of GDDR6 memory and 696 GB/s of memory bandwidth, released Q4 2020. It is built for inference, fine-tuning and mixed production workloads.
Renting one currently costs between $0.27 and $2.05 per GPU-hour across the 7 cloud providers we track, with a median of $0.40. The cheapest live rate is $0.27/hr from Nosana (Community), and hosting one pays about $0.32/hr on Nosana while it is running. All rates below are normalised to a single GPU per hour so providers with multi-GPU nodes compare like-for-like.
Where to rent the A40
How we rank
The median is one provider, one vote: each provider contributes the median of its own on-demand listings, then the headline is the median across providers — so no single marketplace or hyperscaler sets the price alone. Below two providers we show the range instead.
Default order is relevance, weighed the way a buyer would: price per GPU first, lightly blended with total node cost; on-demand preferred over spot and reserved when prices are close; established clouds edged above marketplaces; in-stock, recently-checked offers above stale ones; and no single provider is allowed to crowd the top. Click a column to sort by the raw numbers. No paid placements — ranking is never sold.
| Provider | Tier | Config | Total VRAM | Billing | $ / GPU / hr | Total / hr | Availability | Go to provider |
|---|---|---|---|---|---|---|---|---|
 Vast.ai | GPU cloud | 2× A40 | 96GB | On-demand | $0.28 | $0.56 | In stock | Rent |
 Nosana | GPU cloud | 1× A40 | 48GB | Community | $0.27 | $0.27 | Unknown | Rent |
 GPU.ai | Marketplace | 1× A40 | 48GB | Community | $0.29 | $0.29 | In stock | Rent |
| Vast.ai | GPU cloud | 4× A40 | 192GB | On-demand | $0.28 | $1.12 | In stock | Rent |
| Vast.ai | GPU cloud | 1× A40 | 48GB | On-demand | $0.34 | $0.34 | In stock | Rent |
| Nosana | GPU cloud | 1× A40 | 48GB | On-demand | $0.36 | $0.36 | Unknown | Rent |
 RunPod | GPU cloud | 1× A40 | 48GB | On-demand | $0.44 | $0.44 | In stock | Rent |
| Vast.ai | GPU cloud | 1× A40 | 48GB | Spot | $0.32 | $0.32 | In stock | Rent |
| GPU.ai | Marketplace | 1× A40 | 48GB | On-demand | $0.44 | $0.44 | In stock | Rent |
| RunPod | GPU cloud | 2× A40 | 96GB | On-demand | $0.44 | $0.88 | In stock | Rent |
| EExoscale | GPU cloud | 1× A40 | 48GB | On-demand | $1.05 | $1.05 | Unknown | Rent |
 Hyperstack | GPU cloud | 1× A40 | 48GB | On-demand | $1.16 | $1.16 | Unknown | Rent |
| EExoscale | GPU cloud | 2× A40 | 96GB | On-demand | $1.05 | $2.09 | Unknown | Rent |
| EExoscale | GPU cloud | 4× A40 | 192GB | On-demand | $1.05 | $4.18 | Unknown | Rent |
| EExoscale | GPU cloud | 8× A40 | 384GB | On-demand | $1.05 | $8.36 | Unknown | Rent |
 Runcrate | Marketplace | from A40 | — | On-demand | $2.05 | — | Unknown | Rent |
Earn by hosting the A40
How this is worked out
“Renters pay” is the median live rate for this GPU on that marketplace — the median, not the floor, because the cheapest listing is one host undercutting everyone. “You earn” is that figure less the platform’s cut.
“Per day” is a ceiling: 24 hours, every hour rented, which no host achieves — scale it by the utilisation you expect. A card sitting idle earns nothing.Fees marked as published come from the platform’s terms; the rest are our estimates, labelled on each row.
Technical specifications
- Memory
- 48 GB
- TDP
- 300 W
- Released
- 2020
| Architecture | Ampere |
|---|---|
| Memory | 48 GB GDDR6 ECC |
| Bandwidth | 696 GB/s |
| Tensor compute | 149.7 TFLOPS FP16 (sparse) |
| Form factor | passive dual-slot PCIe |
| Power | 300 W |
- Class
- Inference
- Cloud providers
- 7 tracked
- Price range
- $0.27 – $2.05 / GPU / hr
Other inference cards
NVIDIA A10SNAPSHOT 2026-08-23T22:42:15.889486+00:00 · BACK TO THE FULL INDEX