NVIDIA L40S rental price
DatacenterThe universal inference and rendering card. Widely stocked at every tier of the market, which keeps its rental rate honest.
live · updated 17d ago
VRAM
48 GB GDDR6
Architecture
Ada Lovelace
Segment
Datacenter
Providers quoting
10
best rate
$0.467/GPU-hr at Vast.ai
24h
—
7d
—
30d
—
cheapest 8× node
$11.60/hr
The cheapest NVIDIA L40S you can rent right now is $0.467 per GPU-hour at Vast.ai, against $1.861 at the dearest of the 10 providers quoting it — a 4.0× spread for identical silicon. A full 8-GPU node starts at $11.60/hr at Crusoe. Rates below refresh automatically.
Rent vs. buy
buy outright
$8,095
rent
$0.467/hr
breakeven
23.7 months
Buying a NVIDIA L40S at today’s street price of $8,095 costs the same as 17,333 hours (23.7 months) of renting one at $0.467 an hour. Run it less than that and renting is simply cheaper; run it more and the card starts paying itself back — but this is a floor, not a total cost of ownership. It counts only the purchase price against the rental rate, and ignores power, cooling, rack space, networking, depreciation and the cost of tying up the capital, all of which push the real breakeven further out.
NVIDIA L40S purchase prices and history →L40S hourly rate history
Daily low $/GPU-hour per provider, cheapest provider first · showing 9 of 10 providers — hover the chart for detail.
Where to rent the NVIDIA L40S
Vast.ai
Marketplace
$0.467/GPU-hr
4× L40S · $1.868/hr per node · Japan, JP
- 4× L40S x4$1.868/hr
- 2× L40S x2$1.069/hr
- 1× L40S x1$0.670/hr
Boostrun
GPU cloud
$0.690/GPU-hr
8× L40S · $5.520/hr per node · pcie · US, Dallas, TX
no capacity right now
Latitude.sh
GPU cloud
$0.740/GPU-hr
1× L40S · $0.740/hr per node · pcie · DE, Frankfurt
no capacity right now
- 8× g3-l40s-large · no capacity$6.970/hr
Massed Compute
GPU cloud
$0.880/GPU-hr
1× L40S · $0.880/hr per node · pcie · US, Des Moines, IA
- 4× gpu_4x_l40s$3.520/hr
- 8× gpu_8x_l40s · no capacity$7.040/hr
- 2× gpu_2x_l40s$1.760/hr
Verda
GPU cloud
$1.430/GPU-hr
1× L40S · $1.430/hr per node · pcie · FI, Central
no capacity right now
- 8× 8L40S.160V · no capacity$11.50/hr
Crusoe
GPU cloud
$1.450/GPU-hr
2× L40S · $2.900/hr per node · pcie · US, Houston, TX
- 1× l40s-48gb.1x · no capacity$1.450/hr
- 2× l40s-48gb.2x$2.900/hr
- 8× l40s-48gb.8x$11.60/hr
DigitalOcean
GPU cloud
$1.570/GPU-hr
1× L40S · $1.570/hr per node · pcie · CA, Toronto
no capacity right now
Scaleway
GPU cloud
$1.700/GPU-hr
1× L40S · $1.700/hr per node · pcie · PL, Warsaw
- 8× L40S-8-48G · no capacity$13.60/hr
- 2× L40S-2-48G$3.400/hr
- 4× L40S-4-48G · no capacity$6.800/hr
AWS
Hyperscaler
$1.861/GPU-hr
1× L40S · $1.861/hr per node · us-east-1
- 1× g6e.2xlarge$2.242/hr
- 4× g6e.12xlarge$10.49/hr
- 1× g6e.4xlarge$3.004/hr
L40S cluster nodes
Multi-GPU configurations, cheapest per GPU-hour first. For distributed training the interconnect matters as much as the rate — an NVLink or InfiniBand node and the same count of loose PCIe cards are not the same machine.
| Provider | Configuration | Interconnect | $/GPU-hr | Node/hr |
|---|---|---|---|---|
| Vast.ai | 4× L40S x4 · Japan, JP | — | $0.467 | $1.868/hr |
| Vast.ai | 2× L40S x2 · Japan, JP | — | $0.534 | $1.069/hr |
| Boostrunno capacity | 8× L40Sx8 · US, Dallas, TX | pcie | $0.690 | $5.520/hr |
| Latitude.shno capacity | 8× g3-l40s-large · DE, Frankfurt | pcie | $0.871 | $6.970/hr |
| Massed Compute | 4× gpu_4x_l40s · US, Kansas City, KS | pcie | $0.880 | $3.520/hr |
| Massed Computeno capacity | 8× gpu_8x_l40s · US, Des Moines, IA | pcie | $0.880 | $7.040/hr |
| Massed Compute | 2× gpu_2x_l40s · US, Kansas City, KS | pcie | $0.880 | $1.760/hr |
| Verdano capacity | 8× 8L40S.160V · FI, Central | pcie | $1.438 | $11.50/hr |
| Crusoeno capacity | 10× l40s-48gb.10x · US, Culpeper, VA | pcie | $1.450 | $14.50/hr |
| Crusoe | 2× l40s-48gb.2x · US, Houston, TX | pcie | $1.450 | $2.900/hr |
| Crusoe | 8× l40s-48gb.8x · US, Culpeper, VA | pcie | $1.450 | $11.60/hr |
| Crusoe | 4× l40s-48gb.4x · US, Houston, TX | pcie | $1.450 | $5.800/hr |