India AI tracker
India’s AI rules, public compute and home-grown models in one place. Every entry links to the official source, with the date we last checked it.
At a glance
IndiaAI Mission GPUs, by stage
Each bar is a different stage, reported at a different date. They show progress, not one capacity number.
View data
| Stage | GPUs | Date | Source |
|---|---|---|---|
| Original target | 10,000+ | Mar 2024 | Official source |
| Onboarded | 38,231 | Mar 2026 | Official source |
| Shared capacity | 45,000+ | Jun 2026 | Official source |
India AI timeline
Rules, compute milestones and model releases, oldest first.
- IndiaAI Mission approvedApproved
- Draft amendment on synthetic information (2025)Superseded draft
- IT Rules amendment on synthetically generated informationIn force
- GPUs onboardedReported
- Shared compute capacity and usageReported
- C-DAC tender for up to 10,000 AI compute unitsBidding
- Broader AI governance consultation paperAnnounced, not published
- EKA-1 (Soket)Released; early access
- Dhīr 2B and Dhīr 5B (ZenteiQ)Downloadable (Apache 2.0)
Rules and regulation
What is legally in force for AI in India today, and what has only been proposed or announced.
| Item | Status | Date | Details | Source |
|---|---|---|---|---|
| IT Rules amendment on synthetically generated information | In force | Notified on 10 February 2026 (G.S.R. 120(E)) and in force from 20 February. Adds due-diligence duties for intermediaries on specified synthetic audio, visual and audiovisual content. Explainer → | Gazette of India, G.S.R. 120(E) | |
| Draft amendment on synthetic information (2025) | Superseded draft | MeitY’s October 2025 draft, which proposed a fixed 10% marking formula. The final February 2026 rules use different wording. Explainer → | MeitY draft amendment rules, 2025 | |
| Broader AI governance consultation paper | Announced, not published | The IT minister said a consultation paper would be floated within a month. As of our 31 October check it was not on official sources; a consultation paper would not itself be binding law. Explainer → | News on AIR, 8 October 2026 |
IndiaAI Mission and public compute
Official figures for India’s shared AI compute programme. They describe different stages—target, onboarded, shared, used—and do not add up to a single capacity number.
| Item | Status | Date | Details | Source |
|---|---|---|---|---|
| IndiaAI Mission approved | Approved | Five-year outlay of ₹10,371.92 crore, with a common compute target of more than 10,000 GPUs. Explainer → | PIB, Cabinet approves IndiaAI Mission | |
| GPUs onboarded | Reported | 38,231 GPUs onboarded from 14 providers, according to the government. Onboarded capacity is not the same as capacity in active use. Explainer → | PIB, 25 March 2026 | |
| Shared compute capacity and usage | Reported | Shared capacity of more than 45,000 GPUs as of June 2026; 237 projects and 93.18 lakh GPU-hours reported by August. The factsheet does not define whether hours are allocated or metered. Explainer → | PIB factsheet, Semiconductor and AI Revolution | |
| C-DAC tender for up to 10,000 AI compute units | Bidding | Request for empanelment of AI compute service providers in two stages. A 27 October corrigendum set bid submission for 13 November 2026, 3:30 p.m. IST. A tender requirement, not deployed capacity. | C-DAC RFE C-DAC/GPU/01/2026 |
Indian foundation models
Models released under the IndiaAI Mission, and whether you can actually download them.
| Item | Status | Date | Details | Source |
|---|---|---|---|---|
| EKA-1 (Soket) | Released; early access | 26.3 billion parameters, trained on about 7.5 trillion tokens, covering all 22 scheduled Indian languages plus English, according to the government. Soket offered early access rather than a public download as of 9 November. | PIB, 3 November 2026 | |
| Dhīr 2B and Dhīr 5B (ZenteiQ) | Downloadable (Apache 2.0) | Small models for English and ten Indian languages that ZenteiQ says run fully offline. Dhīr 5B is a mixture-of-experts model with 5.09 billion parameters, 1.79 billion active per token. | ZenteiQ, Dhīr model page |
How to read this tracker
Official AI figures in India describe different stages: a target, GPUs onboarded by providers, shared capacity, and hours used. They are listed separately here and should not be added together. Rules are marked as in force only when notified in the Gazette and past their start date.