NVIDIA Ada Lovelace

NVIDIA L40S

L40S increases clocks and FP8 support over L40, boosting inference throughput in Ada data centers.

Launch year
2023
Memory
48 GB GDDR6
Memory bandwidth
864 GB/s
Peak FP16 / FP32
181 TFLOPS · 91 TFLOPS

Market snapshot

$0.18 /hr

Range $0.18 – $43.70

Catalog coverage

260 live offerings

Across 5 providers · 60 regions

  • 4th-gen Tensor cores with FP8
  • NVENC/NVDEC
  • PCIe Gen4
NVIDIA L40S render

Last refreshed Oct 20, 2025, 1:51 AM

Performance snapshot

Normalized versus NVIDIA A100 (=1.0). Values use public reference benchmarks for training and inference workloads.

  • AI Training (FP16)×0.65
  • AI Inference (FP16)×0.75
  • Memory Bandwidth×0.43

Provider availability

Price bands per provider pulled from the live catalog.

  • Datacrunch logoDatacrunch32 offers
    $0.18 – $7.31 /hr
  • RunPod logoRunPod56 offers
    $0.26 – $6.02 /hr
  • Amazon Web Services logoAmazon Web Services120 offers
    $0.55 – $43.70 /hr
  • Nebius logoNebius12 offers
    $1.55 – $11.01 /hr
  • Oracle Cloud logoOracle Cloud40 offers
    $14.00 – $14.00 /hr

Popular regions

  • us-east-216 offers
  • eu-north-116 offers
  • us-west-216 offers
  • us-east-116 offers
  • eu-central-116 offers
  • ap-northeast-216 offers
  • ap-northeast-116 offers
  • OC-AU-114 offers

Weighted average price $6.30 /hr · median $3.00