Analytical digest for EPW July 2026, built from verified data and public debate.
The launch of IndiaAI Mission 2.0 at the India AI Impact Summit 2026 placed a phrase at the centre of policy debate: sovereign AI. EPW’s July commentary treated it not as a technical slogan but as a political-economy claim, that India needs national control over the artificial-intelligence stack to support its long-standing commitment to strategic autonomy. The argument is intuitive. It is also expensive, and EPW’s analytical value lies in asking whether the sovereignty framing clarifies India’s choices or obscures its trade-offs.
What Sovereign AI Means
Sovereign AI is the doctrine that a nation should control the full stack on which its critical AI systems run: compute (hardware), data, and models. The rationale is that dependence on foreign platforms for any layer creates a chokepoint, whether through export controls on chips, opaque model behaviour, or the offshore processing of citizens’ data. In this reading, AI capability becomes an extension of strategic autonomy, the doctrine that has guided Indian foreign policy since non-alignment, now applied to technology and supply chains.
The IndiaAI Mission
The IndiaAI Mission, approved in March 2024 with an outlay of about Rs 10,000 crore, is the institutional vehicle. It rests on several pillars: a subsidised common compute facility, support for indigenous foundational models, curated datasets, and applications in Indian languages.
| Pillar | Purpose | Verified Detail |
|---|---|---|
| Common compute | Affordable GPU access for startups and researchers | GPU access offered at roughly Rs 65 per hour, a fraction of global commercial rates |
| Foundational models | Reduce dependence on foreign LLMs | BharatGen positioned as a government-backed multimodal model effort |
| Language and data | Serve Indian linguistic diversity | Bhashini’s translation stack moved onto Indian cloud and GPU infrastructure |
| Safety and governance | Trustworthy deployment | Emphasis on data residency and jurisdictional control |
The subsidised compute price is the mission’s most concrete instrument. By offering GPU time far below commercial cost, the state lowers the entry barrier for domestic AI builders, a classic infant-industry intervention.
The Case For
- Data sovereignty: keeping citizens’ data, especially in languages, health and governance, within Indian jurisdiction reduces exposure to foreign surveillance and regulatory reach.
- Supply-chain resilience: the global scramble for advanced GPUs, and the possibility of export restrictions, makes domestic compute capacity a hedge against denial.
- Developmental relevance: foreign models under-serve Indian languages and contexts; a sovereign stack can prioritise Bharat-specific use cases.
- Strategic signalling: AI capability is increasingly a marker of great-power status, and India seeks to be an “AI builder,” not merely a consumer.
The Case For Caution
EPW’s analytical scepticism centred on cost and opportunity. Building sovereign compute and frontier models is capital-intensive, and the hardware itself, advanced GPUs, is still overwhelmingly imported, which limits how “sovereign” the stack really is at the silicon layer. There is a risk that the sovereignty frame justifies prestige spending on frontier models while more grounded priorities, digital public infrastructure, AI skilling, and safety regulation, go under-funded. A pragmatic reading distinguishes strategic sovereignty (control over critical, security-sensitive systems and data) from blanket autarky (attempting to reproduce the entire global stack domestically), and argues that only the former is worth the fiscal cost.
Strategic Autonomy in the AI Age
Applied to technology, strategic autonomy does not require self-sufficiency in every layer. It requires avoiding single points of dependence and preserving the ability to choose partners. India’s approach, subsidised domestic compute plus selective international collaboration on infrastructure, is consistent with a hedging posture rather than isolation. The policy question EPW posed is where to draw the line: which layers are genuinely security-critical and merit sovereign control, and which can be sourced globally at lower cost without compromising autonomy.
Way Forward
- Prioritise the security-critical stack: data residency for sensitive datasets, sovereign control of models used in defence, finance and public administration, rather than duplicating all commercial AI domestically.
- Invest in the enablers of capability: talent, research funding and safety institutions, which compound over time, rather than one-off model launches.
- Deepen the compute base through incentives for domestic chip assembly and long-term GPU procurement agreements, while acknowledging silicon dependence honestly.
- Legislate a data-protection and AI-governance framework so that “sovereignty” is backed by enforceable rights, not just infrastructure ownership.
UPSC Angle
- GS3 (Science and Technology): artificial intelligence, compute infrastructure, indigenous foundational models, data centres.
- GS2 (International Relations): strategic autonomy in the technology domain; hedging between global powers on semiconductors and AI.
- GS3 (Economy): industrial policy, infant-industry support, and the fiscal cost of technological self-reliance.
Likely Mains question: “Sovereign AI is presented as an extension of strategic autonomy. Examine the case for and against, and suggest where India should concentrate its resources.” (GS3, 15 marks)
Facts Corner
- IndiaAI Mission: approved March 2024; outlay about Rs 10,000 crore; common compute facility offers GPU access at roughly Rs 65 per hour.
- IndiaAI Mission 2.0: launched at the India AI Impact Summit 2026, marking India’s shift from AI consumer to AI builder.
- Sovereign AI: national control over compute, data and models to protect data sovereignty and strategic autonomy.
- Bhashini: India’s national language-translation platform, migrated onto Indian cloud and GPU infrastructure for data residency.
- Strategic autonomy: India’s doctrine of independent decision-making, now extended from foreign policy to technology and supply chains.
