NVIDIA H100 On request
Training · fine-tuning · inference
₹1,80,000a month + taxes
Get an H100 quoteHopper with FP8 for training, fine-tuning and inference. ₹1,80,000 a GPU a month, host included.
On request. We confirm the variant and availability before you commit.
Businesses we’ve hosted and supported since 2009
Training · fine-tuning · inference
₹1,80,000a month + taxes
Get an H100 quoteHosted in Mumbai. Each card keeps its own memory; two cards don’t pool.
| GPU | Memory | Per card a month | Choose it when |
|---|---|---|---|
| NVIDIA A100On request | 80GB | From ₹99,000 | Your stack runs on Ampere and 80GB is enough. |
| NVIDIA H100On request | 80GB | From ₹1,80,000 | You need FP8 and the Transformer Engine. |
| NVIDIA H200Ready now | 141GB | From ₹2,20,000 | 80GB is the limit: long context, bigger models. |
Also ready now: L40S 48GB from ₹55,000 and RTX PRO 6000 96GB from ₹1,10,000. All GPUs on the GPU cloud page.
An NVIDIA H100 80GB is ₹1,80,000 per GPU card per month with a baseline host. Your quote lists PCIe or SXM, card count, CPU, RAM, storage and network before you commit. The minimum term is one month.
Yes for large transformer training or serving: H100 adds FP8, the Transformer Engine and up to 3.35TB/s of memory bandwidth on SXM. If your workload already fits an A100 comfortably, A100 at ₹99,000 is the better value.
When 80GB is the limit. H200 has 141GB of HBM3e at 4.8TB/s for larger models, longer context and bigger KV caches. It costs ₹40,000 more a card a month, and it is ready now.
PCIe has 80GB HBM2e at 2TB/s, up to 350W. SXM has 80GB HBM3 at 3.35TB/s, up to 700W, with NVLink for multi-GPU training. We confirm the variant before setup.
H100 is sourced on request. Tell us the card count and when you need it; we confirm availability and send a written quote within one business day.