Every major AI system you have heard of — ChatGPT, Gemini, Claude — was built in the United States. They work well in English. They work reasonably in Hindi. They work poorly in Bhojpuri, Maithili, Santali, Dogri, and dozens of other languages spoken by hundreds of millions of Indians.

This is not only a convenience problem. When AI understands only dominant languages, it creates a second-order disadvantage: people who speak other languages cannot access its benefits. Their health questions go unanswered. Their crop problems go undiagnosed. Their legal rights go unexplained. The AI revolution, left to the market, would widen India's existing knowledge gap rather than close it.

This is the argument behind India's decision to build its own AI — not to compete with OpenAI, but to ensure the technology works for every Indian, in their language, on their problems.

IndiaAI Mission: The Plan

On 7 March 2024, the Cabinet Committee on Economic Affairs approved the IndiaAI Mission with a total outlay of ₹10,371.92 crore (approximately $1.25 billion) over five years. It is the largest single investment in AI infrastructure India has made.

The mission has seven pillars:

1. IndiaAI Compute Capacity: Building shared AI computing infrastructure — GPU clusters — that Indian startups, research institutions, and government can access at subsidised rates. The goal is to have 10,000+ GPUs available for Indian AI builders by 2025-26. Without access to computing, Indian developers cannot train competitive AI models.

2. IndiaAI Innovation Centre (IAIC): Developing indigenous foundational AI models — large language models trained on Indian data, in Indian languages, for Indian contexts. This is the core research engine.

3. IndiaAI Datasets Platform: A national data platform that aggregates non-personal, non-sensitive government datasets — agricultural data, health records, satellite imagery, language corpora — so AI models can be trained on Indian reality, not American or European reality.

4. IndiaAI Application Development Initiative: Funding AI applications in priority sectors — healthcare (diagnosis support for rural health workers), agriculture (crop advisory), education (personalised tutoring), and smart cities.

5. IndiaAI FutureSkills: Training 5 lakh+ people in AI over five years through programmes in IITs, state universities, and online platforms. Includes creating 2,000 PhD-level researchers in AI.

6. IndiaAI Startup Financing: A ₹2,000+ crore fund providing deep-tech AI startups access to capital, computing, and data — reducing dependence on foreign venture funding.

7. Safe and Trusted AI: Developing India's AI safety and governance framework, so AI systems deployed in India meet ethical, privacy, and safety standards.

₹10,372 Cr
Total IndiaAI Mission budget approved by Cabinet, March 2024
Source: CCEA, Ministry of Electronics & IT, 2024
22
Scheduled Indian languages supported on Bhashini platform
Source: Bhashini, Digital India Corporation, 2025
10,000+
GPUs being provisioned for shared Indian AI computing infrastructure
Source: IndiaAI Mission operational guidelines, 2024
5 Lakh
People to be trained in AI through IndiaAI FutureSkills over 5 years
Source: Ministry of Electronics & IT, IndiaAI Mission

Bhashini: The AI That Speaks Your Language

Bhashini (bhashini.gov.in) — named after the Sanskrit word for "speaker" — is a government-built AI platform for Indian language translation and speech processing. It is not a chatbot. It is a translation and voice infrastructure layer that other applications can use.

What Bhashini does:


  • Text translation between any two of 22 scheduled Indian languages (and English)

  • Speech-to-text in Indian languages — you speak, it transcribes

  • Text-to-speech — AI reads text aloud in Indian languages

  • Voice-to-voice translation — real-time spoken translation between Indian languages

It is built by the Digital India Corporation under MeitY, using AI models developed by IITs, IISc, and other institutions — trained specifically on Indian language data.

Where Bhashini Is Already Being Used

  • DigiLocker: Bhashini provides Hindi and regional language support in DigiLocker
  • UMANG app: Regional language navigation powered in part by Bhashini models
  • Common Service Centres (CSCs): Staff at CSCs in non-Hindi states can use Bhashini for translation with villagers
  • iGOT Karmayogi: Government training platform uses Bhashini for multilingual content
  • MyGov portal: News and government communications translated to regional languages through Bhashini

How to try Bhashini: Visit bhashini.gov.in and click "Try Now" — you can type text and translate it to any Indian language, or test the voice translation features directly in your browser.

Private Indian AI: What the Ecosystem Is Building

Sarvam AI

A Bengaluru startup founded by IIT and IISc alumni, Sarvam AI is building Sarvam-1 — a large language model trained specifically on Indian languages, particularly Hindi, Tamil, Bengali, Telugu, and Marathi. It is designed for Indian business and government applications. Sarvam AI partnered with the IndiaAI Mission as one of the first compute recipients.

Shuka 1.0 — Sarvam's Hindi voice AI — can converse fluently in Hindi, understand Indian accents, and is optimised for voice-first interfaces suited to rural India where typing is a barrier.

Krutrim (Ola)

Krutrim (Sanskrit: "artificial") is Bhavish Aggarwal's (Ola founder) AI company. Krutrim Si Designs runs Krutrim, an AI model pre-trained on 2 trillion tokens of Indian-language data across 20+ languages. In early 2024, it became India's first AI unicorn. Available at krutrim.com.

BharatGPT / Hanooman

A consortium of IIT Bombay, IIT Madras, and Reliance Jio developed Hanooman — a multilingual AI series trained on 40+ Indian languages and dialects. Designed for healthcare, governance, education, and finance use cases. Uses data from government sources, Wikipedia, and Indian media.

CoRover (BharatBot)

The company behind BharatBot — an AI assistant deployed on IRCTC (train ticketing), HDFC Bank, and several state government portals. BharatBot handles queries in 12 Indian languages and has processed billions of conversations. It is the most deployed Indian-origin AI in terms of real user interactions.

What This Means for Ordinary Indians

The point of a national AI programme is not prestige — it is whether it changes the experience of an ordinary person accessing a government service, understanding a medical result, or learning a new skill.

The practical impact over the next 2-3 years:

  • Government portals in your language: Bhashini integration means PM-KISAN, Ayushman Bharat, and other scheme portals will increasingly offer voice-based and regional language interfaces — you speak your query, the system responds in your language.
  • Voice-first AI for low-literacy users: The AI systems being built for India prioritise voice over text — recognising that typing (especially in regional scripts) is a barrier that speech bypasses.
  • AI health workers: Experiments are already underway in states like Andhra Pradesh and Kerala where AI is being used to guide ASHA workers through patient interactions, identifying risk factors and suggesting referrals.
  • Agricultural AI in local languages: Crop advisory, market price information, and weather alerts delivered in the farmer's language through voice AI — the IFFCO-Kisan initiative already uses AI voice calls to reach 7+ crore farmers.

What You Can Do

  • Try Bhashini at bhashini.gov.in — translate a paragraph to your state language, or test the speech-to-text in Hindi. It is a government service, already live.
  • Follow IndiaAI.gov.in for updates on the mission — datasets being released, compute access for researchers, and AI application pilots in your sector.
  • If you are a student or developer: The IndiaAI Mission is offering subsidised GPU compute access for research and startup projects. Apply at indiaai.gov.in.
  • If you work in health, agriculture, or education: AI tools specifically designed for Indian languages and Indian conditions are becoming available. Watch for your state government's digital initiatives — many are now AI-backed.

India's bet is that AI built for India will do what foreign AI cannot: speak Bhojpuri, understand the problems of a smallholder farmer in Vidarbha, navigate the complexity of 22 official languages and hundreds of dialects. Whether that bet pays off will depend on sustained investment, capable execution, and whether the tools actually reach the people they were designed for. The infrastructure is being built. The question is the last mile.

Sources