India's $1.25 Billion AI Bet: Inside the IndiaAI Mission, and Why the World Is Watching
India is spending 1.25 billion dollars on sovereign AI compute, models and startups under the IndiaAI Mission. Inside Modi's bet, and how it compares globally.
While the United States and China dominate the artificial intelligence headlines, a third contender is spending heavily to avoid being left behind. India has committed 10,371.92 crore rupees, about 1.25 billion dollars, to the IndiaAI Mission, a national programme to build what its government calls sovereign AI: home-grown compute, datasets, models and startups that do not depend on foreign infrastructure.
What the money buys
The headline pillar is raw computing power. Through the mission, India has made 34,000 graphics processing units (GPUs) available to developers at 115 to 150 rupees per GPU-hour, roughly 42 percent below market rates, and a shared national facility has already surpassed 38,000 GPUs. The government has committed to scaling this to 100,000 GPUs by the end of 2026. The compute component alone is funded with 4,563.36 crore rupees over five years, and projects of national importance can receive up to a 40 percent cost reduction.
Beyond hardware, the mission funds an open Indian dataset platform, application-development grants, and direct support for startups building on domestic infrastructure. It is designed to roll out in four phases between 2025 and 2035.
Home-grown models arrive
The strategy is already producing results. On 18 February 2026, the startup Sarvam released two open-source models, Sarvam-30B and the larger Sarvam-105B. Separately, BharatGen, India's first government-funded multimodal large language model, was launched with support for 22 Indian languages and training on domestic datasets meant to reflect the country's cultural diversity.
Why it matters beyond India
| Pillar | What it provides |
|---|---|
| Compute | 34,000 GPUs now, scaling to 100,000 by end 2026, at subsidised rates |
| Models | Open-source sovereign models (Sarvam) and a 22-language LLM (BharatGen) |
| Data & startups | Open dataset platform, grants, and startup support on domestic infrastructure |
| Timeline | Four phases, 2025 to 2035 |
For the rest of the world, India's push matters for two reasons. First, India is one of the largest sources of AI engineering talent and IT services, so the tools and norms it builds will ripple into products used globally. Second, India is testing whether a middle power can build a credible AI base without simply renting it from American or Chinese cloud giants, a question every government outside those two countries is now asking. The security dimension is built in too: sovereign compute and data are as much about control and trust as they are about cost, and India has paired the mission with work on safe and trusted AI to address risks like deepfakes.
The gaps are real. India still lacks a frontier foundation model to rival the largest American and Chinese systems, and 100,000 GPUs, while significant, is modest next to the millions deployed by the biggest US labs. But as a statement of intent, the IndiaAI Mission marks India's clear decision to be a builder of AI, not just a consumer of it.