The IndiaAI Mission is the government’s five-year programme to give Indian startups, researchers, students and public bodies cheaper access to AI computing, shared datasets, homegrown models and training. For most readers the practical questions are narrower: who can actually use the subsidised GPUs, what AIKosh offers, and what has been delivered so far.
The Union Cabinet approved the mission in March 2024 with an outlay of ₹10,371.92 crore over five years. The figures below come from the ministry’s answers in Parliament and its own portals, so each is tied to a date.
The seven pillars
A PIB release of 27 July 2026 lists the seven pillars as IndiaAI Compute, Foundation Models, AIKosh, the IndiaAI Application Development Initiative, FutureSkills, Startup Financing, and Safe & Trusted AI.
They are not equal in size. When the compute portal launched in March 2025, MeitY Secretary S. Krishnan said compute was the mission’s largest component, with nearly 45% of its funding allocated to it.
Compute: what the subsidy covers
The compute pillar works through empanelled AI service providers whose cloud GPUs are offered through the IndiaAI Compute Portal, with a subsidy on top for approved users. Separately, the ministry says a high-performance AI compute system of about 1.1 exaflops is being set up at the NIC data centre in Shastri Park, Delhi.
In a Lok Sabha answer of 1 April 2026, the ministry said more than 38,000 GPUs had been empanelled and that the average rate was about ₹65 per GPU per hour, except for select high-end GPUs. Its answer of 29 July 2026 put the count at 15 empanelled compute providers, 237 end users or projects supported and about 93.18 lakh GPU hours sanctioned.
The same July answer describes the subsidy as “up to 40%”, governed by a published End User Policy. The subsidy reduces the bill; it does not make compute free. The portal’s FAQ says users bear costs beyond the subsidy and pay the service provider according to its billing cycle.
Who can apply for GPUs
The portal’s eligibility page groups applicants into categories, each with its own proof:
| Applicant | What the portal asks for |
|---|---|
| Startups and MSMEs | DPIIT recognition, AI/ML experience, and annual revenue of ₹50 lakh or funding of ₹1 crore |
| Early-stage startups | Four years of AI/ML experience (including research) and a letter from the IndiaAI CEO or an officer of Joint Secretary rank or above |
| Researchers and academia | A research profile in AI/ML, such as publications or citations |
| Students | AI/ML projects and an aligned academic profile |
| IndiaAI fellows | Selection under the IndiaAI Fellowship |
| Government entities | A letter from the Director, MD, GM or an authorised officer |
The End User Policy lists the identifiers used to verify profiles: an APAAR ID for students, DPIIT recognition for startups, an Udyam ID for MSMEs, and a Google Scholar, ORCID or Scopus ID for researchers.
How an application moves
Sign-in uses Meri Pehchaan, through DigiLocker, Jan Parichay or e-Pramaan. After that, the portal describes four steps: complete the end-user registration with documents, wait for verification, submit a compute request with a draft bill based on the rate list (including any subsidy request), and wait for approval.
Two routes follow from the policy. Smaller requests that seek no subsidy, up to 5,000 GPU hours and 50 GPUs, are handled by designated verifiers: faculty endorsers for students and early-stage researchers, and MeitY Startup Hub centres of excellence for startups. Larger requests, or any request for a subsidy, go to a Program Monitoring and Evaluation Committee, which decides whether to allocate capacity and how much subsidy to grant.
Timing matters. The portal’s FAQ says requests should be submitted between the 1st and the 25th of a month, and that the committee publishes the approved list on the 10th of the following month or the next working day. The portal also publishes the allocations it has approved, with the applicant, GPU type, provider and subsidy.
AIKosh: datasets, models and use cases
AIKosh is the mission’s repository for datasets, models, toolkits and documented use cases, with a sandbox for building on them. According to the July 2026 Lok Sabha answer, it hosts more than 13,445 datasets across 22 sectors from over 578 organisations, along with 329 AI models, 35 toolkits and 202 use cases.
Organisations can also contribute. A July 2025 expression of interest invites non-personal, anonymised datasets, models with documentation, and Indian use cases. Contributors must register, supply metadata and licensing details, and pass a review before anything is published.
Before you build on a dataset, read its licence and description. A national platform does not by itself settle whether a dataset suits your purpose or your obligations under data protection law.
Foundation models
Against a call for proposals issued on 30 January 2025, the ministry says it received 506 applications and identified 20 sovereign model proposals for support, 12 large multimodal models and eight small language models. The July 2026 answer names Sarvam AI’s 30-billion and 105-billion-parameter models, Gnani.AI’s speech-to-speech model, BharatGen’s multilingual models and Avataar AI’s video generation model as launched outputs, and says they are available on AIKosh.
Selected teams receive financial assistance to cover actual compute usage, and the April 2026 answer adds that a further 25% of compute expenditure is earmarked for ancillary costs such as datasets and personnel.
Skills, labs and safety
For students and institutions, the July 2026 answer lists 686 fellowships across 178 institutions, 27 India Data and AI Labs already set up, and 188 more identified for ITIs and polytechnics. Under the Safe & Trusted AI pillar, the July PIB release reports 13 approved responsible-AI projects at educational institutions, covering topics such as bias mitigation, deepfake detection and privacy-preserving machine learning.
The governance side, including the AI Safety Institute and the bodies formed under the national guidelines, is covered in our explainer on India’s AI governance guidelines.
What has not been decided
Read announcements about a second phase carefully. In the July 2026 answer, the ministry said the government announced at the India-AI Impact Summit 2026 its intention to build on the mission through a subsequent phase, but that no detailed roadmap, framework or financial outlay for it had been formally approved or notified.
Figures on GPUs, users and datasets also move between answers. The 38,000 GPU figure and the 15-provider count come from different dates, so check the portal and the latest parliamentary reply before quoting them.
Sources
MeitY via PIB: mission overview, 12 October 2025; AI Compute Portal and AIKosh launch, 6 March 2025; Safe & Trusted AI update, 27 July 2026. Lok Sabha unstarred questions 6044 (1 April 2026) and 1688 (29 July 2026). IndiaAI Compute Portal eligibility page, FAQ and End User Policy; IndiaAI AIKosh expression of interest, July 2025. Documents consulted on 10 October 2026; links appear beside the relevant discussion.
