Rent NVIDIA L40S in the Cloud

Expect NVIDIA L40S cloud price to begins at $0.55 per hour on-demand. According to our July 14, 2026 snapshot, the market's lowest-priced on-demand option is Vast.ai, while RunPod offers the lowest spot rate at $0.79 per hour — a up to 60% saving for interruption-tolerant jobs. Reserved contracts add another ~up to 40% discount for multi-month bookings.

Typical one-month on-demand bills run into four figures at full utilization, which is why production inference services usually migrate to reserved and experimental work stays on-demand or spot.

Top cloud providers offering NVIDIA L40S: Vast.ai (from $0.55/hr), Novita AI, and Latitude.sh. Start comparing live rates.

At a glance: NVIDIA L40S = Ada Lovelace architecture, 48 GB GDDR6 memory, 864 GB/s bandwidth, 366 FP16 TFLOPS, 91.6 FP32 TFLOPS, 350W, 2023.

Those specs tell most of the story for machine learning: VRAM sets the model ceiling, bandwidth throttles attention-heavy production inference, and TFLOPS set pre-training throughput. NVIDIA L40S sits firmly in the class of accelerators targeted at modern transformer workloads — the bandwidth/TFLOPS balance is tuned for large-batch pre-training and production production inference rather than gaming.

Top cloud providers offering NVIDIA L40S: Vast.ai (from $0.55/hr), Novita AI, and Latitude.sh. Start comparing live rates.

366 FP16 TFLOPS and 864 GB/s of memory bandwidth put NVIDIA L40S squarely in the class of accelerators targeted at modern transformer workloads. FP32 caps at 91.6 TFLOPS, which still handles most non-AI scientific compute comfortably.

For training from scratch, token throughput roughly tracks FP16 TFLOPS. For production inference on foundation models, throughput tracks bandwidth. Real-world numbers will depend heavily on the framework stack (PyTorch, TensorRT-LLM, vLLM), and can vary 30-50% depending on how aggressively you quantise.

The cheapest NVIDIA L40S right now is on Vast.ai at $0.55/hr; Novita AI and Latitude.sh also list it — pick the region and billing that fits your workload.

NVIDIA L40S is listed by 7 cloud GPU providers: DigitalOcean, Latitude.sh, Massed Compute, Novita AI, RunPod, Vast.ai, and others. Entry on-demand pricing starts at $0.55 per hour via Vast.ai, while spot capacity starts at $0.79 per hour via RunPod.

That's a relatively crowded market, which keeps rates competitive but also means provider-level differences in support, uptime, and egress policies are the real decision criteria. Serverless GPU offerings (where available) further reduce cost by billing only for the inference seconds you actually consume.

Ready to deploy? Vast.ai has NVIDIA L40S from $0.55/hr. Novita AI and Latitude.sh also carry it.

The sweet spot for NVIDIA L40S: Inference, fine-tuning, video processing, AI-accelerated graphics. Those workloads exploit its 48 GB VRAM and Ada Lovelace architecture well and don't get bottlenecked by its compute envelope.

If you're still unsure whether NVIDIA L40S is right, benchmark your actual workload on it — at $0.55 per hour you can spin up a test instance, run your pipeline for an hour, and make an informed decision with real data.

Spin up NVIDIA L40S on Vast.ai for $0.55/hr, with Novita AI and Latitude.sh as strong alternatives — compare and deploy in minutes.