-
Get UpdatesGet notified on education trends, resources, and growth opportunities! -
-
Click it and Unblock the Notifications
IndiaAI Compute GPU Slots Open: How Startups Can Secure Subsidised Access in 2026
IndiaAI Compute has opened fresh subsidised GPU slots this week. The portal shows an update dated August 25. Requests under 5,000 GPU‑hours continue to receive auto‑approval, letting teams start within hours. Bengaluru and Hyderabad startups are accelerating GenAI hiring as access improves for pilots and finetuning.
Live pricing shows practical entry points. Nvidia L40S lists at ₹67.5 an hour on‑demand, with 12‑month reservations at ₹45. H100 PCIe shows ₹252 on‑demand and ₹185 reserved. Intel Gaudi 3 begins at ₹153. AMD MI300X lists ₹168. A subsidy up to 40 percent applies post‑approval.
Who can book IndiaAI Compute subsidised GPUs now
Eligibility spans researchers, IndiaAI fellows, students, startups, micro, small and medium enterprises, and government entities. Researchers with an h‑index of at least five or 150 citations qualify for streamlined access. Students need APAAR identification. Startups need DPIIT recognition; MSMEs require Udyam registration. Verification happens within the portal workflow.
Auto‑approval threshold and subsidy workflow
If you request 5,000 GPU‑hours or fewer, seek no subsidy, and need up to 50 GPUs, allocation is automated after profile checks. Larger asks or any subsidy go to the Program Monitoring and Evaluation Committee. Requests are batched monthly and decisions published on the tenth.
Indicative hourly bands and best‑fit instances
Use these live bands to shape budgets and hiring plans. Entry L40S suits inference and light finetuning. H100 PCIe and MI300X handle heavier jobs. Gaudi 3 offers solid value. Committing for twelve months lowers rates and stabilises cash burn, which helps offers and onboarding.
| Instance | On‑demand (₹/GPU‑h) | 12‑month reserved (₹/GPU‑h) | Typical use |
|---|---|---|---|
| Nvidia L40S (1x) | 67.5 | 45.0 | Medium LLM inference, light finetuning |
| Nvidia H100 PCIe (1x) | 252.0 | 185.0 | Heavier finetuning, ASR training |
| Intel Gaudi 3 (1x) | 153.0 | 117.0 | Enterprise RAG, LLM serving |
| AMD MI300X (1x) | 168.2 | 148.0 | Memory‑intensive inference, agents |
How to apply now and get quick access
Sign up with your organisation email, verify by OTP, and complete your profile. Upload APAAR, DPIIT, Udyam or fellowship proof as applicable. Use the Price Calculator to build your bill of materials. Submit the project form and track status. Auto‑approved requests can launch immediately on available capacity.
Startups in Bengaluru and Hyderabad are speeding up offers for GenAI, MLOps and data engineering. Expect keywords like MLflow, Kubeflow, Kubernetes, Databricks, SageMaker, LangChain and LangGraph. Arrive with reproducible finetuning logs, RAG evaluation, containerised pipelines, and a compute budget plan aligned to IndiaAI rates and reservations.
- Block for 8 hours
- Block for 12 hours
- Block for 24 hours
- Don't block
Continue with Google- Male
- Female
- Others
- Under 18
- 18 to 25
- 26 to 35
- 36 to 45
- 45 to 55
- 55+