How much does GPU cloud cost in India? (2026)
Prices vary more than the GPUs do: the same card can cost nearly twice as much from one Indian provider as from another, and two to three times as much on the big clouds. This guide sets out what each GPU costs in India per hour and per month, why prices differ, and how to choose the card before you compare prices.
Updated · 6-minute read
From Indian providers, one NVIDIA L4 costs about ₹26,000–₹37,000 a month, an RTX PRO 6000 about ₹1.09–₹1.34 lakh and an H200 about ₹1.7–₹2.95 lakh, before GST. The same GPUs on AWS in Mumbai cost two to three times as much. At Chandnas: ₹37,499/month, ₹1,35,099/month and ₹2,79,399/month.
| GPU | Memory | Indian providers, a month | AWS Mumbai | Chandnas, a month |
|---|---|---|---|---|
| NVIDIA L4 | 24 GB | ₹26,000–₹37,000 | $0.97 an hour, about $705 a month | ₹37,499 |
| NVIDIA RTX PRO 6000 | 96 GB | ₹1.09–₹1.34 lakh | $5.49 an hour, about $4,010 a month | ₹1,35,099 |
| NVIDIA H100 | 80–94 GB | ₹1.4–₹2.6 lakh | $8.26 an hour, about $6,030 a month | Not offered |
| NVIDIA H200 | 141 GB | ₹1.7–₹2.95 lakh | 8-GPU servers only, about $9.49 per GPU-hour | ₹2,79,399 |
One GPU. Indian providers' published prices, 29 September 2026, before GST where stated; some are "starting from" or promotional, and the processor and memory around the GPU differ. AWS on-demand list prices in US dollars, before GST; a month = 730 hours.
What does GPU cloud cost at Chandnas?
₹37,499 a month for one L4, ₹1,35,099 for an RTX PRO 6000 and ₹2,79,399 for an H200, exclusive of GST — about ₹51, ₹185 and ₹383 an hour if the GPU runs all month.
Our prices sit at the upper end of Indian providers' list prices, or just above it, and well below the big clouds. The RTX PRO 6000 and H200 plans come with 64 vCPU and 192–256 GB of memory around the GPU, more than most listings include. The price also covers engineers in India who check your model fits before you pay, driver and CUDA set-up on request, and a GST invoice from a company you can call.
| Plan | GPU | vCPU | RAM | A month | Running all month |
|---|---|---|---|---|---|
| L4 ×1 · 8 vCPU | 1× NVIDIA L4 24 GB | 8 | 32 GB | ₹37,499 | ₹51 an hour |
| L4 ×2 · 16 vCPU | 2× NVIDIA L4 48 GB | 16 | 64 GB | ₹69,999 | ₹96 an hour |
| L4 ×4 · 32 vCPU | 4× NVIDIA L4 96 GB | 32 | 128 GB | ₹1,38,499 | ₹190 an hour |
| RTX PRO 6000 | 1× NVIDIA RTX PRO 6000 96 GB | 64 | 192 GB | ₹1,35,099 | ₹185 an hour |
| H200 | 1× NVIDIA H200 141 GB | 64 | 256 GB | ₹2,79,399 | ₹383 an hour |
Mumbai, exclusive of 18% GST. Every plan includes 100 GB of NVMe storage, 10 TB of transfer and one IPv4 address. Hourly billing: on request. All GPU plans
How much does a GPU cost per hour in India?
From Indian providers, on demand: about ₹36–₹63 an hour for an L4, ₹149–₹215 for an RTX PRO 6000, ₹192–₹356 for an H100 and ₹231–₹476 for an H200. AWS in Mumbai charges about $0.97 an hour for an L4 and $8.26 for an H100.
Hourly billing suits short jobs: a fine-tuning run, a batch of rendering, a test. A GPU that's busy most of the day costs less on a monthly plan — divide the monthly price by the hourly rate to see how many hours a month it takes for monthly to win.
Why do prices for the same GPU differ so much?
Because the listing isn't only the GPU. The processor and memory around it, the card's version, how you commit and what's included all change the price.
- The server around the GPU. One provider's RTX PRO 6000 comes with 16 vCPU and 128 GB; ours with 64 vCPU and 192 GB. Data loading and pre-processing need that CPU and memory.
- The card's version. H100 and H200 come as SXM modules and as PCIe "NVL" cards, with different power and interconnect. Check which one you're being quoted.
- On demand or committed. Monthly and yearly commitments cost less per hour; some listings show the committed price.
- "Starting from" and promotional prices. A few listings are introductory, or quote the smallest configuration.
- GST. Some providers show prices before GST, some don't say. Add 18% to compare like with like.
- What's included. Storage, transfer, driver set-up and someone to call when it breaks.
Are AWS, Azure and Google cheaper for GPUs?
No. For the same GPU in India, the big clouds list at two to three times Indian providers' prices. You pay for their other services and global reach, not for a cheaper GPU.
- AWS Mumbai offers the L4, RTX PRO 6000, H100 and H200 — the H200 only in 8-GPU servers.
- Azure offers only the H100 in India, at about ₹934 an hour on Microsoft's own India price list.
- Google Cloud offers the L4 in Mumbai; its H200 is available there only by reservation, not on demand.
Which GPU should you pay for?
Choose by memory first, then compare prices. A model that doesn't fit in the GPU's memory won't run well at any price, and a bigger card than you need is money wasted.
| GPU | Memory | Fits | Chandnas, a month |
|---|---|---|---|
| NVIDIA L4 | 24 GB | An open model of about 8 billion parameters at 16-bit; speech, OCR, video | ₹37,499 |
| NVIDIA RTX PRO 6000 | 96 GB | A 70B model at 8-bit; rendering and 3D | ₹1,35,099 |
| NVIDIA H200 | 141 GB | The largest open models, long context, fine-tuning | ₹2,79,399 |
Rule of thumb: about 2 GB of GPU memory per billion parameters at 16-bit, roughly half that at 8-bit, plus room for context.
What costs extra on top of the GPU?
GST, and anything beyond the base server: more storage, more data transfer, and — on hourly billing — the hours the GPU sits idle while someone forgets to switch it off.
- GST at 18%, on top of most listed prices.
- Storage. Model files are large: a 70B model is about 140 GB at 16-bit. Our plans include 100 GB of NVMe; more is price on request.
- Data transfer. 10 TB a month is included with us; AWS charges $0.1093 a GB beyond the first 100 GB.
- Idle time. On hourly billing, a GPU left running over a weekend is billed for the weekend.
Which is right for you?
| If this is you | Do this |
|---|---|
| Inference on a model up to about 8B, running all day | NVIDIA L4 |
| A 70B model at 8-bit, or rendering and 3D | RTX PRO 6000 |
| The largest models, long context or fine-tuning | NVIDIA H200 |
| A few hours of GPU time a week | Hourly billing — ask us about it (on request) |
| Training across many GPUs at once | A multi-GPU cluster — tell us the job and we'll say plainly whether we can help |
Questions people ask
How much does an H200 cost per hour in India?
About ₹231–₹476 an hour from Indian providers, on demand. Ours is ₹2,79,399/month, about ₹383 an hour if it runs all month. AWS sells H200s in Mumbai only in 8-GPU servers, at about $9.49 per GPU-hour.
How much does an H100 cost in India?
About ₹192–₹356 an hour, or ₹1.4–₹2.6 lakh a month, from Indian providers. AWS lists one H100 in Mumbai at $8.26 an hour, and Azure at about ₹934 an hour in Central India. We offer the H200, which has more memory, as standard.
What's the cheapest GPU cloud in India?
For a single GPU, an L4 from an Indian provider, from about ₹36 an hour. Cheapest isn't always best value: check the processor and memory you get, whether the price is promotional, and whether GST is included.
Are GPU cloud prices in India quoted with GST?
Mostly not. Most providers, and AWS, list prices before GST; add 18%. Ours are exclusive of GST too, with a GST invoice so a registered business can claim the input tax credit.
L4 or H200: which is better value?
Whichever fits your model. An H200 costs about 7.5 times as much as an L4 with us, and holds models an L4 can't. If your model fits in 24 GB, the L4 is far better value; if it doesn't, the L4 is no value at all.
Do you offer hourly GPU billing?
on request. Our standard plans are monthly, which is cheaper for a GPU that runs most of the day.
See GPU plans and prices.
NVIDIA L4, RTX PRO 6000 and H200 servers in Mumbai, monthly in rupees — with engineers who check your model fits before you pay.