Data Center NVIDIA

L40S

Q: How much does L40S cost?

The L40S has a market price of approximately $9.0k (MSRP: $8.0k). Cloud rental starts at $0.910/hr. Prices may vary based on retailer, region, and availability.

Q: Is L40S good for AI and machine learning?

Yes, the L40S with 48GB VRAM is suitable for many AI/ML workloads. It's particularly recommended for AI Inference. For large language models, you may need multiple GPUs or consider higher-VRAM options like A100 or H100.

Q: Should I buy L40S or rent from cloud?

The breakeven point is approximately 9,890 hours of usage. Buy if you'll use it more than this; rent for shorter projects or variable workloads. Cloud rental from Datacrunch (Verda) starts at $0.910/hr.

Q: What can I run on L40S?

With 48GB VRAM and 183.0 FP16 TFLOPS, the L40S can run: Large language models (7B-13B), Stable Diffusion XL, video AI, and professional 3D rendering.

Q: L40S vs alternatives - which is better?

The L40S offers 48GB VRAM and 183.0 FP16 performance at $9.0k. Compare with similar GPUs using our comparison tool above. Key factors: VRAM for model size, TFLOPS for speed, and price for budget.

Ada Lovelace Architecture · 48GB GDDR6 · PCIe

VRAM

48GB

FP16

183.0

TDP

350W

Hardware Price

$9.0k

MSRP: $8.0k

Cloud from

$0.910/hr

4 providers

Cheapest at Datacrunch (Verda) →

Quick Insights

Performance/Dollar

20.33 TFLOPS/$k

FP16 performance per $1000

VRAM/Dollar

5.3 GB/$k

VRAM per $1000

vs Data Center Average

+2% perf

FP16 TFLOPS comparison

Cloud Availability

4 providers

from $0.910/hr

Specifications

VRAM	48GB GDDR6
Memory Bandwidth	864 GB/s
FP16 TFLOPS	183.0
Tensor TFLOPS	733.0
FP32 TFLOPS	91.6

TDP	350W
Form Factor	PCIe
Architecture	Ada Lovelace
NVLink	No
Release Date	2023-08

Buy vs Rent Analysis

Buy Hardware

$9.0k

One-time cost, unlimited usage
Full control over hardware
Electricity & cooling costs extra
Depreciation over 2-3 years

Best if using >9890 hours total

Rent Cloud GPU

$0.910/hr

Pay only for what you use
No upfront investment
Scale up/down instantly
No maintenance required

Best for <9890 hours or variable usage

Breakeven Point

9,890

hours of usage

At $0.910/hr cloud pricing, buying the hardware pays off after 9,890 hours (~412 days or 13.7 months of 24/7 usage).

Usage	Monthly Cloud Cost	Months to Breakeven
100 hrs/month	$91.00	99 months
200 hrs/month	$182.00	50 months
500 hrs/month	$455.00	20 months

Cloud GPU Pricing

Rent L40S from 4 cloud providers. Prices shown per GPU per hour.

Provider	Type	Instance	GPUs	On-Demand	Per GPU	Spot	Availability
Datacrunch (Verda)	gpu-cloud	datacrunch-l40s	1x	$0.910/hr	$0.910/hr Cheapest	-	-
RunPod	gpu-cloud	NVIDIA L40S	1x	$0.990/hr	$0.990/hr	$0.990/hr	-
Fluidstack	gpu-cloud	fluidstack-l40s	1x	$1.30/hr	$1.30/hr	-	-
CoreWeave	gpu-cloud	coreweave-l40s	1x	$2.25/hr	$2.25/hr	-	-

Best Spot Deal: RunPod offers spot pricing at $0.990/hr (0% off on-demand).

L40S vs Alternatives

Compare L40S with similar GPUs from other brands.

GPU	VRAM	FP16 TFLOPS	Bandwidth	Hardware Price	Cloud Price
L40S Current	48GB	183.0	864 GB/s	$9.0k	-	-
AMD Instinct MI210 AMD	64GB (+33%)	181.0 (-1%)	1.6 TB/s	-	-	Compare
AMD Instinct MI100 AMD	32GB (-33%)	184.6 (+1%)	1.2 TB/s	-	-	Compare
AMD Radeon RX 7900 XTX AMD	24GB (-50%)	122.0 (-33%)	960 GB/s	-	-	Compare

Best Use Cases for L40S

⚡

AI Inference

5/5

Top cloud inference choice

🎬

Video Generation

4/5

Video generation

Compare L40S

Popular Comparisons

A100 vs L40S – Training vs Inference GPU

Other NVIDIA GPUs

B200 192GB · $45k
H200 141GB · $38k
H100 80GB · $32k
H100 PCIe 80GB · $28k
H100 NVL 93GB · $35k

Alternatives

Add to Comparison Tool

Frequently Asked Questions about L40S