L40S
48GB 350W 5 providers 262 offersAn Ada Lovelace universal GPU with 48 GB GDDR6, balancing AI inference, fine-tuning and graphics.
Cheapest
$0.430/hr
AWS
Median
$3.960/hr
across 262 offers
P90
$14.000/hr
90th percentile
Memory
48 GB
TDP
350 W
Bandwidth
864 GB/s
FP16
366 TFLOPS
Brand
NVIDIA
Series
NVIDIA
Year
2023
Arch
Ada Lovelace
Price by provider
cheapest rate · $/GPU/hr · lower is betterAWS
0 regions
$0.430/hr
Verda
0 regions
$0.479/hr
Nebius
0 regions
$0.749/hr
Vultr
0 regions
$1.671/hr
OCI
0 regions
$14.000/hr
5 providers
| Provider | Price Range | Median | Type | CPU / RAM | Action |
|---|---|---|---|---|---|
AW AWS Best price | $0.430$43.699 | $3.252 | Spot | 4-192 32-1536GB | Browse |
VE Verda | $0.479$10.960 | $2.329 | Spot | 20-160 60-480GB | Browse |
NE Nebius | $0.749$11.006 | $2.044 | Spot | 8-192 32-1152GB | Browse |
VU Vultr | $1.671$3.342 | $2.506 | Mixed | 16-32 180-375GB | Browse |
OC OCI | $14.000$14.000 | $14.000 | Mixed | 224-224 1024-1024GB | Browse |
Best Deal Right Now
AW
AWS
$0.430/hr
per GPU / hour
Configurationx1 · L40S
System RAM32-1536GB
Region—
Estimated Running Cost
At cheapest rate · per GPU
1 hour$0.43
24 hours$10.31
1 week$72.17
1 month$7526.59
Explore NVIDIA L40S
Related comparisons, provider pages, raw specs, and model-fit tools.
NVIDIA L40S vs NVIDIA H100Compare live rental prices, VRAM, FP16 throughput, and provider availability.NVIDIA L40S vs NVIDIA H200Compare live rental prices, VRAM, FP16 throughput, and provider availability.NVIDIA L40S vs NVIDIA A100Compare live rental prices, VRAM, FP16 throughput, and provider availability.NVIDIA L40S on AWSSee SKUs, regions, spot status, and current prices from $0.430/hr.NVIDIA L40S on VerdaSee SKUs, regions, spot status, and current prices from $0.479/hr.NVIDIA L40S on NebiusSee SKUs, regions, spot status, and current prices from $0.749/hr.NVIDIA L40S full specificationsOpen the raw GPU Ark spec detail page for clocks, memory, cores, and API fields.Find models that fit this class of GPUMatch LLM size and precision to GPUs with enough VRAM.