Editorial Summary
Writing in the Indian Express (May 6, 2026), former RBI Governor Raghuram Rajan and economist Rohit Lamba issue a strategic warning: India is at risk of becoming a “tenant” of foreign AI systems — using and paying for AI infrastructure built by OpenAI, Google, Microsoft, and Meta — without owning the underlying capabilities, controlling the data governance, or sharing in the economic value created.
The piece argues that India possesses a unique and unreplicable foundation for building sovereign AI: its Digital Public Infrastructure (DPI) — the India Stack (Aadhaar, UPI, ONDC, DigiLocker, ABHA). No other country has a comparable combination of scale (1.4 billion biometric IDs), transactional depth (18–21 billion UPI transactions monthly), and public-good design principles. The authors argue this DPI can serve as the data and governance backbone for developing India-specific AI applications at scale.
But DPI alone is not enough — the editorial calls for three parallel investments: sovereign compute infrastructure, open multilingual datasets, and open-source large language models for Indian languages.
Key Arguments
The “Tenant” Risk
| Scenario | Implication |
|---|---|
| India uses OpenAI/Google AI with no sovereign stack | Strategic dependency; data sovereignty risk; no domestic economic value capture |
| India’s data trains foreign AI models | Indian users’ data enriches foreign companies; India pays for the output |
| Critical sectors (healthcare, agriculture, governance) run on foreign AI | Geopolitical vulnerability; terms can change; no recourse |
| India lacks compute infrastructure | Cannot train or fine-tune large AI models independently |
The “tenant” metaphor is precise: a tenant pays rent, cannot modify the property, can be evicted, and builds no equity — the opposite of strategic technology ownership.
India’s Unique Advantage — The DPI Foundation
| Component | Scale | AI Relevance |
|---|---|---|
| Aadhaar | 1.4 billion biometric IDs | Identity verification; consent-based data access |
| UPI | 18–21 billion transactions/month | Financial behaviour patterns; economic activity data |
| ONDC | Open Network for Digital Commerce | Commercial transaction data; supply chain |
| DigiLocker | 250+ crore documents | Educational, government records |
| ABHA | Ayushman Bharat Health Account | Healthcare records; disease pattern data |
| CoWIN | Vaccination data | Demographic + health |
This DPI generates structured, consent-based, high-quality data across demographics, sectors, and regions — rare at this scale globally.
What India Must Build
| Component | What Is Needed | Why |
|---|---|---|
| Sovereign compute | GPU clusters for AI training (India AI Mission: 10,000 GPUs initially) | Without compute, India cannot train or fine-tune large models |
| Open multilingual datasets | Curated training data in 22 Scheduled Languages | India’s AI must work for Tamil farmers, not just English speakers |
| Open-source LLMs | Indian-language foundation models (Sarvam AI model is a start) | Prevents monopoly; enables ecosystem |
| Public-private R&D | IIT + IISc + private sector (Sarvam, Krutrim) collaboration | Distributes investment; builds talent retention |
| Data governance framework | Digital Personal Data Protection Act 2023 implementation | Ensures data sovereignty while enabling AI development |
IndiaAI Mission — Current Status
| Parameter | Detail |
|---|---|
| Budget | ₹10,372 crore |
| Compute target | 10,000 GPUs (Phase 1); expanding to 100,000+ (Phase 2) |
| Key component | AI Compute Infrastructure + IndiaAI Datasets Platform + FutureSkills Prime |
| Private players | Sarvam AI (Indian LLM); Krutrim (Ola founder Bhavish Aggarwal’s AI) |
| Space AI example | Pixxel + Sarvam AI “Pathfinder” — orbital AI satellite (Q4 2026) |
UPSC Relevance
| Paper | Angle |
|---|---|
| GS3 — Science & Technology | AI governance, sovereign AI, LLMs, compute infrastructure |
| GS3 — Economy | Digital economy, technology exports, intellectual property |
| GS2 — Governance | IndiaAI Mission, DPI governance, data sovereignty, DPDP Act 2023 |
Mains Keywords: AI sovereignty, Digital Public Infrastructure, India Stack, Raghuram Rajan, Rohit Lamba, IndiaAI Mission, Sarvam AI, Krutrim, DPDP Act 2023, UPI, Aadhaar, open-source LLM, sovereign compute, GPU infrastructure, multilingual AI, technology tenant
Prelims Facts Corner
| Item | Fact |
|---|---|
| Article authors | Raghuram Rajan (former RBI Governor, 2013–16) + Rohit Lamba (Princeton economist) |
| Core argument | India must produce AI, not become a “tenant” of foreign AI platforms |
| DPI foundation | Aadhaar (1.4B IDs), UPI (18–21B transactions/month), ONDC, DigiLocker, ABHA |
| IndiaAI Mission | ₹10,372 crore; 10,000 GPUs (Phase 1) |
| Sarvam AI | Indian open-source LLM company; Pathfinder satellite (with Pixxel) |
| Krutrim | AI startup by Bhavish Aggarwal (Ola founder) |
| DPDP Act 2023 | Digital Personal Data Protection Act — data governance framework |
| Raghuram Rajan’s tenure | RBI Governor: September 2013 – September 2016 |
