Sovereignty in the Age of Intelligence: The Indian AI Infrastructure Roadmap

Source Attribution: This summary is based on the Government of India’s white paper, “Democratising Access to AI Infrastructure,” released by the Office of the Principal Scientific Adviser in December 2025 and discussed at the World Economic Forum 2026, as reported by *The Hindu*:
Infrastructure over Algorithms: The paper argues that the future of AI will be determined by physical and digital infrastructure—compute power and datasets—rather than just software or chatbots.
Strategic Autonomy: Control over AI infrastructure is framed as a matter of digital sovereignty, ensuring that India remains an innovator rather than a mere consumer of foreign technology.

Source

The Public Good Model: AI infrastructure is envisioned as a foundational utility, comparable to electricity grids or national highways, enabling innovation across all sectors.
Physical Layer Components: This layer includes high-performance computing clusters, GPU clouds, specialized data centers, and advanced energy systems.
Digital Layer Components: This includes curated national datasets, foundational model repositories, interoperable governance frameworks, and secure access protocols.

Global Disproportion: India generates nearly **20% of global data** but currently hosts only about **3% of global data center capacity**, creating a massive strategic gap.
Foreign Dependency: This asymmetry forces Indian researchers and startups to rely on foreign platforms, potentially exposing sensitive data and increasing operational costs.
Sovereign Hosting: The policy aims to localize compute resources to ensure data security, regulatory alignment, and national control over critical AI upgrades.

Massive GPU Deployment: The mission has successfully deployed over **38,000 GPUs** (including H100s), a significant jump from the initial target of 10,000, to support domestic AI training.
Subsidized Compute Access: High-end compute resources are provided to startups and researchers at a highly subsidized rate of approximately **₹65 per hour**.
Foundational Models: The government is backing 12 Indian entities to develop “India-first” Large Multimodal Models optimized for 22 official languages and local contexts.

AI Kosh (Dataset Platform): A centralized repository hosting over **7,000 datasets** and 264 AI models across 20 sectors, allowing developers to build on existing blocks.
Bhashini (Language Technology): An AI-powered platform designed to break language barriers, providing voice-first multilingual tools for all Indian languages.
Interoperability Standards: By treating AI building blocks as Digital Public Goods (DPGs), the government ensures that innovation is not limited to a handful of large corporations.

Market Entry Barriers: The concentration of advanced chips and frontier models in a few global firms creates high entry barriers for smaller players and developing nations.
Bargaining Power: Dependence on external AI infrastructure weakens India’s ability to negotiate terms and exposes critical sectors to external vulnerabilities.
Avoiding Isolationism: The white paper advocates for “shared access pathways” that allow global competition while retaining control over domestic critical systems.

Environmental Planning: Scaling AI data centers is projected to require an additional **45–50 million square feet** of real estate by 2030, necessitating careful environmental planning.
Rising Power Demands: Data center energy consumption is expected to rise from **0.5% to 3% of India’s total electricity** by 2030, highlighting the need for renewable integration.
Advanced Cooling Systems: The policy calls for the adoption of liquid cooling and energy-efficient architectures to manage the high thermal output of AI-ready facilities.

Leveraging Private Efficiency: The scale of infrastructure required cannot be met by the State alone; PPPs are identified as the primary lever for expanding GPU clouds and regional hubs.
Sovereign AI Clouds: Collaborative projects, such as the Digital Sangam with IIT Madras and Sarvam AI, are creating at-scale compute hubs for state-level adoption.
State-Level Adoption: States like Odisha and Tamil Nadu are partnering with private firms to build dedicated 50MW AI-optimized facilities for public utilities.

Moving Beyond IT: While finance and e-commerce have adopted AI rapidly, sectors like **agriculture and healthcare** have lagged due to high infrastructure costs.
Precision Agriculture: Democratized access enables tools like the “Vivasāya Nanban” assistant, providing 24/7 digital advisory to nearly 8 million farm households.
Inclusive Healthcare: Localized compute and data platforms allow for advanced diagnostics and language-accessible health tools in vernacular contexts.

Third Path for India: India is charting a path that is neither a state monopoly nor a laissez-faire corporate concentration, but a trust-based, public-good ecosystem.
Global Template: The DPI-led approach to AI provides a potential model for other nations in the Global South seeking technological autonomy.
Final Choice: The success of AI in India will ultimately be decided by infrastructure investment, ensuring it remains a shared capability for inclusive growth.

Swastika Shiksha