Request access
Private Preview

L40SAda Lovelace

NVIDIA L40S

Inference, small-model fine-tuning, rendering and video on PCIe servers.

On-demand
$1.86 /GPU·hr
Reserved
Request a quote
Request this GPU

GDDR6 with ECC per GPU

48 GB

NVIDIA reference figure

§01 Specifications Reviewed 2026-10-10
DWG 01L40S Source · NVIDIA datasheet
NVIDIA L40S, dimetric shop drawing
  1. 1. PCIe x16 edge connector · PCIe Gen4
  2. 2. Passive heatsink · no fan on the card; the chassis moves the air
  3. 3. Cover · drawn cut away to show the fins
  4. 4. GPU, 48 GB GDDR6 with ECC · under the heatsink
  5. 5. Bracket · dual-slot, full height
  6. 6. 16-pin power connector · one per card

Point at a part or a row. The other one lights.

L40S specifications.

TAB 01L40S specifications Source · NVIDIA Reviewed 2026-10-10
L40S specifications
Parameter Value Scope Source
Memory
Memory 48 GB GDDR6 with ECC Per GPU [1]
Memory bandwidth 864 GB/s Per GPU [1]
Compute, peak · dense / sparse
FP4 Tensor Core Not supported on Ada Lovelace Per GPU
FP8 Tensor Core 733 / 1,466 TFLOPS Per GPU [1]
BF16 Tensor Core 362 / 733 TFLOPS Per GPU [1]
Interconnect
NVLink None
Host interface PCIe Gen4 x16 · 64 GB/s bidirectional Per GPU [1]
Power and form factor
Power 350 W Per GPU [1]
Form factor PCIe card, dual-slot, full height, full length, passive cooling [2]
  1. [1] NVIDIA L40S
  2. [2] NVIDIA Enterprise Reference Architecture, Appendix A
Reviewed 2026-10-10

Specifications are NVIDIA reference figures.

Best for

  • Inference
  • Fine-tuning smaller models
  • 3D graphics, rendering and video (3 NVENC, 3 NVDEC, AV1)

Look elsewhere for

  • Multi-node training that needs InfiniBand

Notes on the figures

  • FP4: not supported on Ada Lovelace.
§02 Price and terms USD per GPU-hour

On-demand or reserved.

Metered by the second, billed hourly. Reserved rates are quoted per term.

  • On-demand

    $1.86 /GPU·hr

    Per GPU-second
    $0.000517
    Per GPU-month, 730 h
    $1,357.80

    List price for the base shape: 1 GPU, 8 vCPU, 32 GiB. Larger shapes are quoted.

  • Spot

    Not offered

    Spot is not offered on this listing.

  • Reserved

    Request a quote

    Capacity you need on a date, guaranteed. 1 month, 6 months or 12+ months.

    Reservations

Ways to pay Prepaid credit, pay-as-you-go and Enterprise. Each one covers on-demand, spot and reserved capacity. How billing works

EQ 01L40S terms Source · Fantasti price list Reviewed 2026-10-10
Terms for NVIDIA B300, USD per GPU-hour. The on-demand rate is the preview list price of $10.45. Reserved rates without a published figure are quoted.
TermCommitmentCapacityRate
On-demandNone · Not heldNoneNot held$10.45 /GPU·hr
1 monthTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
6 monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
12+ monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
Terms for NVIDIA B200, USD per GPU-hour. The on-demand rate is the preview list price of $9.35. Reserved rates without a published figure are quoted.
TermCommitmentCapacityRate
On-demandNone · Not heldNoneNot held$9.35 /GPU·hr
1 monthTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
6 monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
12+ monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
Terms for NVIDIA H200 SXM, USD per GPU-hour. The on-demand rate is the preview list price of $5.94. Reserved rates without a published figure are quoted.
TermCommitmentCapacityRate
On-demandNone · Not heldNoneNot held$5.94 /GPU·hr
1 monthTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
6 monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
12+ monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
Terms for NVIDIA H100 SXM5, USD per GPU-hour. The on-demand rate is the preview list price of $5.40. Reserved rates without a published figure are quoted.
TermCommitmentCapacityRate
On-demandNone · Not heldNoneNot held$5.40 /GPU·hr
1 monthTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
6 monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
12+ monthsTake-or-pay · Held for the term, on one fabricTake-or-payHeld for the term, on one fabricRequest a quote
Terms for NVIDIA RTX PRO 6000 Blackwell Server Edition, USD per GPU-hour. The on-demand rate is the preview list price of $2.16. Reserved rates without a published figure are quoted.
TermCommitmentCapacityRate
On-demandNone · Not heldNoneNot held$2.16 /GPU·hr
1 monthTake-or-pay · Held for the termTake-or-payHeld for the termRequest a quote
6 monthsTake-or-pay · Held for the termTake-or-payHeld for the termRequest a quote
12+ monthsTake-or-pay · Held for the termTake-or-payHeld for the termRequest a quote
Terms for NVIDIA L40S, USD per GPU-hour. The on-demand rate is the preview list price of $1.86. Reserved rates without a published figure are quoted.
TermCommitmentCapacityRate
On-demandNone · Not heldNoneNot held$1.86 /GPU·hr
1 monthTake-or-pay · Held for the termTake-or-payHeld for the termRequest a quote
6 monthsTake-or-pay · Held for the termTake-or-payHeld for the termRequest a quote
12+ monthsTake-or-pay · Held for the termTake-or-payHeld for the termRequest a quote

List price for the base shape: 1 GPU, 8 vCPU, 32 GiB. Larger shapes are quoted.

Reserved rates are quoted per term. Reservations are tied to a GPU type, a region and one InfiniBand fabric. Take-or-pay under a signed order form. No early cancellation without penalty (Terms §3).

§03 Node shape Platform configuration

Runs as single instances.

PCIe servers with 1 to 4 L40S cards per machine. No NVLink and no InfiniBand.

L40S machine shapes. Source · Fantasti platform configuration.

L40S is a PCIe card passed through whole to one machine: 1 GPU (8 to 40 vCPU · 32 to 160 GiB), or up to 4 gpus (16 to 192 vCPU · 96 to 1,152 GiB). No NVLink and no InfiniBand: these machines run as single instances. Connections: L40S to 1 GPU; L40S to Up to 4 GPUs; 1 GPU to InfiniBand fabric; Up to 4 GPUs to NVLink.

  • L40SPCIe card
    L40S, line drawing
    memory
    48 GB GDDR6 with ECC
    host
    PCIe Gen4 x16
  • 1 GPUInstance
    vCPU
    8 to 40
    RAM
    32 to 160 GiB
    1 GPU
  • Up to 4 GPUsInstance
    vCPU
    16 to 192
    RAM
    96 to 1,152 GiB
    Up to 4 GPUs
  • NVLinkNone

    Cards in one machine are not linkedto each other.

  • InfiniBand fabricNone

    These machines do not join amulti-node cluster.

  • L40SPCIe card
    L40S, line drawing
    memory
    48 GB GDDR6 with ECC
    host
    PCIe Gen4 x16
  • 1 GPUInstance
    vCPU
    8 to 40
    RAM
    32 to 160 GiB
    1 GPU
  • Up to 4 GPUsInstance
    vCPU
    16 to 192
    RAM
    96 to 1,152 GiB
    Up to 4 GPUs
  • NVLinkNone

    Cards in one machine are not linked toeach other.

  • InfiniBand fabricNone

    These machines do not join a multi-nodecluster.

  • L40SPCIe card
    L40S, line drawing
    memory
    48 GB GDDR6 with ECC
    host
    PCIe Gen4 x16
  • 1 GPUInstance
    vCPU
    8 to 40
    RAM
    32 to 160 GiB
    1 GPU
  • Up to 4 GPUsInstance
    vCPU
    16 to 192
    RAM
    96 to 1,152 GiB
    Up to 4 GPUs
  • NVLinkNone

    Cards in one machine are not linked to eachother.

  • InfiniBand fabricNone

    These machines do not join a multi-nodecluster.

L40S listing.

CPU, RAM and local NVMe vary by pool and are reported at placement. GPUs are passed through whole to one machine.

Status
Private Preview
Runs as
Single instances
Capacity
On-demand · Reserved
Metering
Metered by the second, billed hourly

GPU Instances

TAB 02L40S node shape Source · Fantasti platform configuration Reviewed 2026-10-10
L40S node shape
ParameterValueScope
1 GPU8 to 40 vCPU · 32 to 160 GiBPer machine
Up to 4 GPUs16 to 192 vCPU · 96 to 1,152 GiBPer machine
InfiniBandNone

NVIDIA L40S on Fantasti: fantasti.ai/gpus/l40s · Request capacity: fantasti.ai/contact

Sheet
01 / 01
Title
L40S datasheet
Figures
NVIDIA reference
Reviewed
2026-10-10
§04 Datasheets 8 listings
§05 Request Private Preview

Request L40S capacity.

Tell us the count, the term and where it should run. We reply to a reviewed request, usually within one business day. A reserved term is quoted.

Request this GPU GPU catalog

Next