Rent the NVIDIA L40S 48GB in India from ₹55,000 a card a month with host included, one-month minimum. Ready now.
48GB per card for inference, fine-tuning and rendering.
A short paid trial, with price and duration agreed before setup.
Businesses we’ve hosted and supported since 2009
Inference · fine-tuning · rendering
₹55,000a card a month + taxes
Trial node on L40S, then ₹55,000 a card a month if you keep it.
Trial on L40S Chat on WhatsAppReselling GPUs to your own customers? Partner pricing on request.
Hosted in Mumbai. Each card keeps its own memory; two cards don’t pool.
Host included: CPU, RAM and storage, on one monthly INR invoice. Your quote lists the configuration and any extras.
| GPU | Memory | Per card a month | Choose it when |
|---|---|---|---|
| NVIDIA L4On request | 24GB | From ₹32,000 | Light inference and video that fit in 24GB. |
| NVIDIA L40SReady now | 48GB | From ₹55,000 | Inference, fine-tuning, image generation or rendering up to 48GB. |
| NVIDIA RTX PRO 6000Ready now | 96GB | From ₹1,10,000 | 48GB is the limit, or the work mixes AI with rendering and media. |
Also ready now: H200 141GB from ₹2,20,000. All GPUs on the GPU cloud page.
One variant: a dual-slot, passively cooled PCIe card. Two cards stay two separate 48GB devices.
| NVIDIA L40S | Value |
|---|---|
| Memory | 48GB GDDR6 with ECC |
| Memory bandwidth | 864 GB/s |
| Max power | 350W |
| Interface | PCIe Gen4 x16, 64 GB/s |
| NVLink and MIG | Not supported |
| BF16 Tensor, dense | 362 TFLOPS |
| FP8 Tensor, dense | 733 TFLOPS |
| Cooling | Passive, dual slot |
Source: NVIDIA L40S specifications, NVIDIA.
Weights only: parameters times bytes, plus 20% headroom for the runtime. Context and users add KV cache on top.
| Model | FP16 | 8-bit | 4-bit | One L40S runs it at |
|---|---|---|---|---|
| 8B | 16GB | 8GB | 4GB | FP16 |
| 13B | 26GB | 13GB | 6.5GB | FP16 |
| 34B | 68GB | 34GB | 17GB | 8-bit |
| 70B | 140GB | 70GB | 35GB | 4-bit |
Two cards don’t pool memory; your software splits the model. Full table: GPU inference sizing.
We rent cards; we don’t sell them. Buying means the card, a host built for it, duty, GST, power and your own staff.
| Topic | Buy it | Rent it from us |
|---|---|---|
| Up front | Card plus host, duty and GST. | First month, ₹55,000 plus taxes. Host included. |
| Try it first | You own it before you know it fits. | A short paid trial node. Keep it and it becomes production. |
| A year | Depends on the quote and how long you keep the card. | ₹6,60,000 for 12 months, before taxes (our maths). |
| If plans change | You resell the card. | One-month minimum. Move to RTX PRO 6000 or H200 when 48GB is the limit. |
More GPU prices: L4 24GB rental price · RTX PRO 6000 96GB rental price · A100 80GB rental price · H200 141GB rental price · GPU inference hosting · All GPU cloud prices.
An NVIDIA L40S 48GB is ₹55,000 per GPU card per month with a baseline host. Your quote lists its CPU, RAM, storage and network, and the card count. The minimum term is one month.
We rent by the month, not the hour. ₹55,000 a month is about ₹75 an hour if the card runs all month (our maths, 730 hours). For a few GPU hours, use AWS, which we also manage.
Yes, on a short paid trial node with the NVIDIA driver installed. You run your own test, then decide. If you keep it, the trial node becomes your production node.
Rent L40S when 24GB is too small, or the work mixes inference with fine-tuning, image generation or rendering. L4 at ₹32,000 a month suits light inference and video that fit in 24GB.
No. NVIDIA lists both as unsupported on L40S. Two cards stay two separate 48GB devices, so your software splits the work. For NVLink, ask for an NVLink-connected H100 or H200 node.
Yes. Prices exclude taxes, and Indian invoices carry applicable GST, on one monthly rupee invoice. Whether you can claim input credit depends on your tax status and the law.
Need GPUs by the hour? Run them on AWS, managed by us: managed AWS hosting.
