GPU compute on NVIDIA DGX H200, by the hour
Rent a share of the Centre of Excellence's DGX H200 for AI training, deep learning, generative AI, scientific computing and other heavy workloads. Open to students, faculty, researchers, startups and companies — pick the size you need and pay only for the hours you book.
8
H200 GPUs
141 GB
HBM3e per GPU
1,128 GB
Total GPU memory
Available sizes
NVIDIA H200 partitions
500 GB storage included with every size. Subject to availability and approval.
Save your data before your session ends
Copy your results, checkpoints and datasets to your own storage before your allocation ends. Once the session closes, the workspace is wiped so the hardware can be handed to the next user — nothing can be recovered afterwards.
Plan ahead
Please submit your request at least 3 days before you intend to start. Requests go through requirement review, payment verification, technical coordination and resource allocation. Same-day access is not guaranteed.
Choose your size
Every size is a partition of the same DGX H200. Larger partitions hold larger models.
What it costs
Nothing is booked here — this only works out the price.
First 500 GB is included. Beyond that, ₹2 per GB per month, charged only for the days you book.
Total
₹0
3 × 12 hours × ₹0 per hour
Who it's for
Ideal for
Flexible and fair-use allocation
Compute is provided on a fair-use basis. The exact allocation is determined according to availability, approved requirements and operational capacity.
Create an account, complete your details, and submit a request. Nothing is charged until your requirement is reviewed.