India's Sovereign AI Strategy Gains Momentum Across Government Departments
From the IndiaAI Mission's GPU bank to homegrown models like BharatGen now running in state and central offices, India's push for AI self-reliance is moving from plan to deployment.
By Naina, 24th June 2026
India's sovereign AI strategy is gaining real momentum as homegrown models and shared computing power move from launch events into the day-to-day work of government departments. Anchored by the ₹10,371 crore IndiaAI Mission and accelerated by its second phase unveiled at the India AI Impact Summit in February 2026, the effort aims to give the country control over the full AI stack, from chips and data to models and applications. Across central ministries and state governments, sovereign models are now being piloted in governance, healthcare, and citizen services, marking a shift from ambition to execution.
The logic is strategic. With most advanced AI controlled by a handful of global firms, New Delhi sees domestic capability as essential to protect sensitive data, serve Indian languages and contexts, and reduce exposure to foreign policy shifts. The government frames sovereign AI not as a vanity project but as a national imperative, with estimates suggesting AI could add well over a trillion dollars to the economy in the coming decade. Yet the build-out also carries real dependencies. Here is where the strategy stands and where it still has gaps.
The IndiaAI Mission Backbone
At the foundation sits the IndiaAI Mission, sanctioned at ₹10,371 crore in 2024 and run by a dedicated division under the IT ministry. It rests on seven pillars spanning compute, foundation models, datasets, applications, safety, startup support, and skills. Its most concrete achievement is computing power: by mid-2026 the mission had onboarded around 34,000 to 38,000 GPUs across Indian data centres, offered to startups, researchers, and government agencies at roughly ₹65 per GPU-hour, a fraction of commercial cloud rates. The government is adding 20,000 more, with a target of 100,000 public GPUs by December 2026.
The Shift to Mission 2.0
At the India AI Impact Summit in New Delhi, IT Minister Ashwini Vaishnaw unveiled IndiaAI Mission 2.0, a strategic transition from building infrastructure to deep research, indigenous development, and inclusive adoption across the economy. The new phase widens the definition of sovereign AI beyond models to include domestic chip development, control systems, and scalable applications, what officials call full-stack sovereignty. The goal is strategic autonomy: the ability to scale AI without relying on foreign gatekeepers for approvals or upgrades, backed by projected investment of around $200 billion into the ecosystem over two years.
The Sovereign Models
India now has homegrown models to deploy. BharatGen, described as a government-funded multimodal effort, launched Param2, a 17-billion-parameter model spanning 22 Indian languages and built on India-centric data. Sarvam AI has open-sourced 30-billion and 105-billion-parameter models, also tuned for Indian languages, and was among four startups selected in the mission's first foundation-model round. Underpinning these is AIKosh, a curated repository of Indian datasets meant to give sovereign models high-quality, locally relevant training data rather than reliance on Western corpora.
The Rollout Across Departments
The clearest sign of momentum is deployment inside government. Sovereign models are powering working solutions rather than demos. In Maharashtra, a governance assistant has been built with the state government to support urban-development and revenue workflows. Central departments are piloting AI-enabled citizen access for water and sanitation services, while Goa is pursuing a broader digital transformation through its state electronics arm. In healthcare, a BharatGen-powered application helps structure and summarise information between doctors and patients. The common thread is practical use across citizen services, finance, health, and education.
The UPI-for-AI Play
One of the boldest ideas is to make AI as easy to adopt as digital payments. The government is developing a common platform, modelled on the Unified Payments Interface, that would host a bouquet of ready-to-use AI tools for small and medium enterprises. The aim is to let a small business plug into AI services as seamlessly as it sends money over UPI, lowering the cost and complexity that keep smaller firms out. If it works, the approach could spread sovereign AI well beyond large corporations and into the wider economy.
The Skills and Safety Pillars
Talent and trust round out the strategy. The mission is funding hundreds of PhD fellows and thousands of postgraduate and undergraduate students, and is setting up data and AI labs in Tier-2 and Tier-3 cities to widen access beyond the metros. India already ranks among the world's leaders in AI skill penetration and talent concentration. On safety, an IndiaAI Safety Institute is backing projects in bias mitigation, privacy-preserving machine learning, explainability, and auditing, an attempt to build responsible-AI capacity alongside raw capability.
The Governance Questions
Policy is evolving in parallel. A government committee has recommended a statutory licensing regime that would require AI developers to pay news publishers for using copyrighted content, a move that could make India one of the first countries to set government-determined royalty rates for AI training. Officials are also emphasising data residency, keeping sensitive information in finance, defence, and healthcare within national borders. These debates reflect a broader effort to align AI development with Indian law, values, and security priorities rather than importing rules set elsewhere.
The Dependence Problem
For all the progress, real constraints remain. India's sovereign compute still runs largely on imported chips, leaving the strategy exposed to the same export controls that have restricted other countries, even if a recent India-US trade framework includes language protecting chip access. Domestic semiconductor capacity is years away from meeting demand. A persistent brain drain sends many of India's best AI researchers to firms abroad, and the data ecosystem remains fragmented. Analysts describe the current position as a foundation rather than an arrival, genuine but still structurally dependent.
The Road Ahead
India's sovereign AI strategy has clearly moved from intent to execution, with models live, compute scaling, and deployments spreading across government departments. The next two years will decide whether that foundation becomes a genuinely self-reliant AI sector or remains an ambitious infrastructure project leaning on foreign chips and talent. Closing the gaps in semiconductors, data, and research depth is the harder, slower work ahead. If India manages it, sovereign AI could become both an economic engine and a model for other emerging economies seeking control over their digital future.
Frequently Asked Questions
What is India's sovereign AI strategy?
It is the effort to design, build, and govern AI using domestic infrastructure, Indian data, and local talent. Anchored by the IndiaAI Mission, it aims to give India control over the full AI stack, from chips and data to models and applications.
What is the IndiaAI Mission?
A government programme sanctioned at ₹10,371 crore in 2024, run under the IT ministry, with seven pillars covering compute, foundation models, datasets, applications, safety, startups, and skills. It provides subsidised GPU access at around ₹65 per hour.
What sovereign AI models has India built?
Homegrown models include BharatGen's Param2, a 17-billion-parameter multilingual model across 22 Indian languages, and Sarvam AI's open-sourced 30-billion and 105-billion-parameter models, supported by the AIKosh dataset repository.
How is sovereign AI being used in government?
Through deployments such as a governance assistant in Maharashtra, AI-enabled citizen services for water and sanitation, digital transformation in Goa, and a healthcare application that summarises doctor-patient information.
What are the main challenges?
Heavy reliance on imported chips and exposure to export controls, limited domestic semiconductor capacity, a brain drain of AI talent abroad, and a fragmented data ecosystem.