India's AI Strategy Is Not the US-China Race. It's a Third Way.

The global AI race is described as a two-horse contest — Silicon Valley's model layer against Beijing's state-coordinated compute build. India is running a third race entirely, and it is further along than most outside observers have noticed.

India's AI strategy in 2026 is not an overnight arrival. It is the compounding of a decade of quieter bets. In 2017, the Ministry of Commerce established an AI Task Force. In 2018, NITI Aayog published the National Strategy for Artificial Intelligence under the framework "AI for All." At a time when the global AI race had not yet become geopolitical, India was already framing AI as a tool for development rather than dominance. That framing has turned out to be a strategic advantage.

The infrastructure numbers are now substantial. India's AIRAWAT platform — a public AI compute infrastructure — has deployed over 18,000 GPUs with capacity crossing 38,000 following 2026 announcements. Subsidised access at approximately ₹60 per hour has lowered barriers for startups and researchers in a way that no private cloud provider has matched. Google announced a $15 billion AI data centre hub in Visakhapatnam in April 2026 — one of the largest single foreign AI infrastructure investments in any country (Spherical Insights, 2026).

The language infrastructure is the least-discussed but most significant piece. BHASHINI, India's AI language platform, now processes 100 million inferences monthly across 22 Indian languages. AI that cannot operate in Hindi, Tamil, Telugu, Odia, and Bhojpuri cannot serve 1.4 billion people. India built the infrastructure to do exactly that. Stanford's AI Index ranked India first globally in AI skill penetration — talent growing over 250% since 2016.

The model India is building draws directly from its own playbook. India Stack — the digital public infrastructure that democratised payments and identity verification for a billion people — is now the template for AI. The same principle applies: the data of 1.4 billion citizens must be processed and analysed within India's borders, using models trained on its linguistic and cultural diversity. Sovereignty is not an ideology here. It is an engineering requirement.

At the India AI Impact Summit in February 2026, 91 countries signed the New Delhi Declaration. The Brookings Institution noted the summit signalled a shift in the global AI agenda — from "safety" as the organising frame toward diffusion, adoption, and development outcomes (Brookings, 2026). That reframe serves the Global South. It also happens to serve India's geopolitical interests perfectly.

If India's AI sovereignty model exports to Africa, Southeast Asia, and Latin America — where most of the world's future AI users live — India will have shaped global AI more profoundly than any individual foundation model.