AI as the Next Digital Public Infrastructure (DPI)

UPSC Syllabus– GS Paper II: Government policies, DPI & e-Governance. GS Paper III: AI, Digital Infrastructure, Science & Technology.

Introduction

India’s digital transformation over the last decade has been driven by the JAM trinity and its extensions-Aadhaar (identity), UPI (payments), and DEPA/Account Aggregator (data)-which together form one of the world’s most advanced Digital Public Infrastructure (DPI) ecosystems. By making these services widely accessible at minimal cost, India turned digital infrastructure into a public utility that millions rely on every day. As artificial intelligence (AI) begins to reshape the global digital economy, an important question arises: can India replicate this model by making AI the next public utility through the same approach of building shared digital infrastructure and bringing down the cost of access? AI as the Next Digital Public Infrastructure (DPI).

1. India’s DPI Success: Lessons for AI

India has repeatedly demonstrated that state-enabled competition, rather than subsidies alone, can make essential digital services affordable and accessible.

●       Affordable data: Between 2016 and 2019, mobile data prices fell sharply due to infrastructure expansion and market competition, bringing nearly 500 million people online within a few years.

●       Affordable identity: Aadhaar enrolled over a billion residents, replacing costly paper-based identity verification with a low-cost, API-driven digital system.

●       Affordable payments: UPI transformed digital payments into a near-free public utility by enabling seamless interoperability across banks and payment applications.

In each case, the government created the digital infrastructure, standards, and regulatory framework, while competition among private players reduced costs and expanded access.

2. India’s Comparative Advantage in Digital Public Infrastructure

Unlike many countries that have developed digital systems in isolated sectors, India has integrated identity, payments, and data-sharing into a single interoperable public digital ecosystem.

●       Estonia has an advanced digital identity system but lacks an equivalent public payments infrastructure.

●       Brazil’s Pix provides an efficient instant-payment platform but functions largely as a standalone system.

●       Singapore’s Singpass and SGFinDex, along with Europe’s open banking framework, offer data portability but only in limited domains.

India’s strength lies not in any one platform, but in the seamless integration of Aadhaar, UPI, and the Account Aggregator framework as interoperable digital public goods. This unique ecosystem provides a strong foundation for extending the DPI approach to artificial intelligence.

3. Why India Needs an AI-DPI

Despite its success in digital public infrastructure, India occupies a relatively weak position in the global AI value chain.

●       Indian engineers develop and fine-tune frontier AI models that are largely created abroad.

●       India’s digitised public life-including languages, transactions, and documents—serves as valuable training data for foreign AI models.

●       Indian universities and the Indian diaspora contribute significantly to global AI research.

Yet, Indian startups often have to buy back AI capabilities through expensive, dollar-denominated API services controlled by foreign companies and hosted outside India’s jurisdiction.

This resembles India’s historical experience during the colonial textile trade, where raw cotton was exported cheaply while finished cloth was imported at a premium. Similarly, India today exports data, talent, and AI usage, while importing finished AI capabilities on terms set by others. This underscores the need for an AI ecosystem built on India’s own digital public infrastructure.

4. Building an Indian AI-DPI Ecosystem

(a) Democratising Compute

Through the IndiaAI Mission, backed by significant government investment, India has adopted a public-private partnership model for AI compute. Instead of creating a state-owned data-centre monopoly, the government has empanelled private cloud providers and aggregated national demand.

●       Tens of thousands of GPUs have already been onboarded, with plans for further expansion.

●       This enables startups and researchers to access computing resources at a fraction of global market costs.

A complementary proposal is to integrate AI compute requirements into national electricity and grid planning, treating power supply for data centres as strategic infrastructure, much like coal linkages were once prioritised for steel production.

(b) Promoting Open-Weight Foundation Models

Dependence on proprietary foreign AI models exposes Indian startups to pricing uncertainty and policy changes beyond their control.

The proposed approach includes:

●       Aggregating anonymised public datasets-such as legal judgments, agricultural records, and educational content in India’s official languages-for developing open-source AI models.

●       Requiring that AI models built using publicly subsidised compute or public datasets be released under an open-weights licence.

This reflects the philosophy behind UPI, where the government builds and standardises the underlying infrastructure while private firms compete by creating innovative applications and services.

(c) Unified Intelligence Interface (UII): “UPI for AI”

Just as UPI’s greatest strength lies in interoperability rather than free payments alone, the proposed Unified Intelligence Interface (UII) aims to provide a standardised API gateway through which applications can seamlessly access multiple AI models—both sovereign and private.

The UII would establish common standards for:

●       Identity verification

●       User consent

●       Billing

●       Safety and governance

This would enable startups, government departments, and educational institutions to access translation, vision, and text-generation services at minimal cost.

5. Financing Inclusive AI Access

To ensure broad and affordable access, proposals include introducing an Aadhaar-linked freemium token model, under which:

●       Verified students, researchers, and startups would receive subsidised AI API tokens.

●       Commercial enterprises would gradually shift to paid access, helping cross-subsidise free users.

Another proposal is to reallocate a portion of existing subsidy expenditure—for example, from fertiliser subsidies—to support AI access for schools, universities, and research institutions.

6. Transformative Potential of AI-DPI

If AI becomes affordable and widely accessible, it could significantly improve service delivery across multiple sectors.

●       Healthcare: India-trained AI models could strengthen healthcare delivery in underserved rural areas.

●       Education: Personalised multilingual tutoring could become accessible to hundreds of millions of learners.

●       Agriculture and governance: Voice-enabled services in Indian languages could simplify access to crop insurance, public services, and grievance redressal.

This would not only reduce the cost of accessing intelligence but also make AI a powerful tool for inclusive development across the country.

Challenges and Concerns

• Data privacy and consent: Large-scale use of public datasets raises concerns about anonymisation, surveillance, data misuse, and informed consent.
• Quality of open-weight Indic models: Building globally competitive AI models requires sustained investment in compute, skilled talent, and high-quality multilingual datasets.
• Energy and infrastructure constraints: Scaling AI infrastructure demands reliable electricity, robust data centres, and integration with long-term national energy planning.
• Digital divide: Affordable AI alone cannot ensure inclusion without widespread access to devices, internet connectivity, and digital literacy.
• Global competitiveness: Since frontier AI training is highly capital-intensive, India may achieve better outcomes by prioritising cost-effective inference capabilities alongside targeted model development.
• Governance of open-weight models: Open AI models require strong regulatory safeguards to address concerns related to misuse, safety, accountability, and responsible deployment.

Way Forward

• Integrate AI into India’s DPI framework: Position the IndiaAI Mission and the proposed Unified Intelligence Interface within the broader Digital India architecture to ensure interoperability and common standards.
• Treat AI compute as strategic infrastructure: Integrate data-centre power requirements into the National Electricity Plan with dedicated renewable and nuclear energy support.
• Promote open-weight innovation: Encourage open licensing for AI models developed using public compute or datasets while incentivising private-sector innovation.
• Strengthen data governance: Build robust safeguards for anonymisation, consent, transparency, and independent oversight in the use of public datasets.
• Bridge the digital divide: Complement affordable AI access with continued investments in rural connectivity, affordable devices, and digital literacy.
• Shape global AI governance: Play an active role in international forums to shape norms on open-source AI, cross-border data flows, and AI infrastructure, positioning India as a rule-shaper rather than a rule-taker.

Conclusion

India’s success with Aadhaar, UPI, and data infrastructure shows that open, interoperable digital public infrastructure can make essential services widely accessible. Applying this model to AI through affordable compute, open-weight models, and a Unified Intelligence Interface can help India become a global AI leader while ensuring that AI serves as an inclusive public good rather than a privilege for a few.

Source- TH