Rent the NVIDIA L40

48 GiB VRAM · ada architecture · released 2022. Billed by the millisecond, a 10% platform fee included in the price, USDC payouts on Base.

Browse live NVIDIA L40 offers Host your own

Reference rate (est.)$0.60 to $1.40/hReference estimate, not a live quote.

VRAM48 GiBada · nvidia

Live offers0Reservable on AmpleRun now.

Live NVIDIA L40 offers

No NVIDIA L40 is free right now.

New capacity appears as hosts publish it. Got one? List yours.

See all GPUs

Apps that fit a NVIDIA L40

Qwen3 1.7B Q8_0 (llama.cpp)llamacpp · needs 7.5 GiB Qwen2.5 7B Instruct (TGI)oci · needs 22 GiB Llama 3.3 70B Q4_K_M (llama.cpp, any GPU)llamacpp · needs 44 GiB gpt-oss-20b (llama.cpp)llamacpp · needs 15 GiB BGE-M3 Q8_0 (llama.cpp)llamacpp · needs 3.5 GiB Llama 3.1 8B Q4_K_M (llama.cpp)llamacpp · needs 7.5 GiB

Models that fit

NVIDIA L40 rental FAQ

How much does it cost to rent a NVIDIA L40?

Each host sets its own rate, so you always see the live price before reserving. Our reference estimate for a NVIDIA L40 is $0.60–$1.40/h (not a live quote). Billing is by the millisecond, with no hourly minimum and the 10% platform fee already inside the rate. See the live offers on this page.

What can I run on a NVIDIA L40?

With 48 GiB of VRAM (ada architecture, 2022 generation), a NVIDIA L40 handles LLM inference, fine-tuning, image generation and training workloads. Start from a published template or bring your own digest-pinned OCI image.

Is renting a NVIDIA L40 on AmpleRun cheaper than the cloud?

We don't publish a cloud price comparison. Each host sets its own rate, AmpleRun's 10% platform fee is already included in it, and billing is by the millisecond, so the quote is the full price to compare with other providers for the same card. Platform fee models are in our sourced fee comparison.

Can I host my own NVIDIA L40 and earn?

Yes, on Linux. Run the one-command installer, set your price, and earn from every rental, paid in USDC on Base. The earn guide has an illustrative estimate.