NVIDIA L40 48GB cloud GPU
L40: 48 GB of GDDR6 memory per GPU, Ada Lovelace architecture. Rent it by the hour on the EverXCloud spot exchange, at the live price set by sellers and buyers.
- Best live price
- $0.69
- per GPU-hour, excl. tax
- GPU memory
- 48 GB GDDR6
- Bandwidth
- 864 GB/s
- FP16 tensor
- 181 TFLOPS
- Architecture
- Ada Lovelace
- Max power
- 300 W
L40 on the spot exchange
As of October 2026, the L40 rents from $0.69 per GPU-hour on EverXCloud, excluding tax.
No L40 product is listed on the spot exchange right now. Tell us the quantity and the dates you need: we look for this capacity across the EverXCloud Network.
Request this capacityL40 specifications
| GPU memory | 48 GB |
|---|---|
| Memory type | GDDR6 |
| Memory bandwidth | 864 GB/s |
| FP16 tensor | 181 TFLOPS |
| BF16 tensor | 181 TFLOPS |
| FP8 tensor | 362 TFLOPS |
| Max power | 300 W |
| Architecture | Ada Lovelace |
| Interconnect | PCIe 4.0 |
Manufacturer specifications, per GPU. Tensor throughput is shown without sparsity. Values the manufacturer does not publish for this exact model are left out.
Recommended uses
- Model training
Full training or fine-tuning of models up to about 2 billion parameters in mixed precision.
- Fine-tuning
Parameter-efficient fine-tuning (LoRA, QLoRA) of open models that fit in memory with 4-bit quantization.
- LLM inference
Language models up to about 15 billion parameters in 16-bit precision, or about 60 billion with 4-bit quantization.
- Image generation
Image and video diffusion models; larger models and higher resolutions need more GPU memory.
- 3D rendering
Ray-traced rendering, visual effects and simulation, with dedicated ray-tracing cores.
Cautious estimates for this product (48 GB of GPU memory in total), counting 80% of memory for model weights. Real limits depend on context length, batch size and framework.
Workloads for the L40
- GPUs for 3D rendering and visual effects
Ray-traced rendering, visual effects and real-time engines.
- GPUs for training large language models
Pre-training and full fine-tuning on data center GPUs with HBM memory.
- GPUs for fine-tuning open models
LoRA and QLoRA adapters on a single GPU, from 24 GB up.
- GPUs for LLM inference
Serve open models: what fits in memory, and how fast tokens come.
- GPUs for image and video generation
Diffusion models for images and video, from 24 GB cards to 80 GB GPUs.
GPUs similar to the L40
| GPU | From/h |
|---|---|
| H100 SXM80 GB HBM3 · Hopper | $2.88 |
| A100 SXM40 GB HBM2 · Ampere | $0.34 |
| A100 PCIe40 GB HBM2 · Ampere | $0.36 |
| A4048 GB GDDR6 · Ampere | $0.35 |
| L40S48 GB GDDR6 · Ada Lovelace | $0.79 |
| Tesla V10032 GB HBM2 · Volta | $0.10 |
L40: frequently asked questions
How much does it cost to rent the L40?
Right now, from $0.69 per GPU-hour excluding tax: the best live offer on the spot exchange. The price moves with the offers and requests of sellers and buyers.
How much memory does L40 have?
48 GB of GDDR6 memory per GPU. Memory bandwidth is 864 GB/s.
Which AI models can I run on L40?
As a cautious estimate, this product (48 GB of GPU memory in total) runs language models up to about 15 billion parameters in 16-bit precision, or about 60 billion with 4-bit quantization. Context length, batch size and framework change the real limit.
Can I rent several L40 GPUs together?
Servers with several GPUs are listed as separate products on the spot exchange when sellers offer them, for example 2, 4 or 8 GPUs. For a larger quantity or fixed dates, request this capacity.
How is L40 rental billed?
By the hour. You buy rental hours on the spot exchange, up to 720 hours per order, and pay them from your prepaid balance. Prices are shown excluding tax, and EverXCloud is the merchant of record for every rental.
Which currencies and payment methods can I use?
Euros, US dollars or pounds, each with its own balance, never converted. Top up by card; SEPA bank transfer may be offered to business accounts on request. A receipt is issued for every top-up and an invoice, with VAT or sales tax where it applies, for every rental.
Is buying GPU hours on the spot exchange a financial product?
No. You buy hours of GPU rental to use them: no derivatives, no leverage, no transferable credits.
Plan your GPU rental
- GPU rental: how to rent a cloud GPU by the hour
- How to rent your first GPU
- How to choose a GPU for LLM inference
- How to estimate a GPU rental session
- H100 price per hour: what sets the rate
- How a GPU spot exchange works
- H200 price per hour: cloud pricing explained
- B200 price per hour: B200, B300 and GB200
- A100 price per hour: rent an NVIDIA A100
- RTX 4090 price per hour: rental guide
- RTX 5090 price per hour: rental guide
- L40S price per hour: rental guide
- MI300X price per hour: rental guide
- How much VRAM do you need for an LLM?
- Best GPU for Stable Diffusion and Flux
- Cheapest way to rent a GPU for AI
- Best GPU for scientific computing (FP64)
Check the workload, complete configuration and billing conditions before choosing a live offer.
Be ready on launch day
Create your account now: you can top up your balance and launch a GPU as soon as the market opens.
Have GPU servers?
Businesses can join the EverXCloud Network and sell hours of their GPU servers on the spot exchange.