2 providers · per GPU-hour
NVIDIA L40 cloud pricing
2 offers from 2 providers, 2 with a published on-demand price. Per GPU-hour, US$. Prices checked 28 Sept 2026.
- Lowest on-demand
- $1.00/GPU-hr
- Lowest spot
- $0.78/GPU-hr
- Memory
- 48 GB GDDR6 ECC
Spec sheet: L40 ↗
| # | Provider | Offer | On-demand | Spot | Committed | As published (original currency) · source | Actions |
|---|---|---|---|---|---|---|---|
| 1 | Hyperstack | L40 x1 1× L40 | $1.00/GPU-hr | — | — | ||
| 2 | CoreWeave | NVIDIA L40 (8x) 8× L40 · 128 vCPU · 1024 GB RAM | $1.25/GPU-hr | $0.78/GPU-hr | — |
What is the L40 used for?
The L40 is a data-centre GPU positioned by its maker for both training and inference, built on the Ada Lovelace architecture. Multi-GPU servers link cards over None listed, which matters when a job is split across several GPUs.
How large a model fits on one L40?
With 48 GB of memory, one card holds the weights of a model up to about 24 billion parameters at 16-bit precision, or about 48 billion at 8-bit. That is weights only: inference also needs room for the KV cache, and training needs several times more for gradients and optimiser state, so larger jobs are split across multiple GPUs.
| Spec (as published) | L40 |
|---|---|
| Architecture | Ada Lovelace |
| Form factor | PCIe dual-slot, passive |
| Memory | 48 GB GDDR6 ECC |
| Memory bandwidth | 864 GB/s |
| Interconnect | None listed |
| Max power (TDP) | 300 W |
| Partitioning | vGPU (vPC/vApps, vWS) |
| Source | maker spec sheet ↗ |
Need L40 capacity for a longer run?
Committed and cluster pricing is often negotiated. Tell us the GPU count, duration and region.