As 35,000 delegates gather in New Delhi for the fourth Global AI Summit — the first to be hosted in the Global South — India is positioning the event as a milestone in shaping a more inclusive global AI order. At the centre of its pitch lies the idea of Digital Public Infrastructure (DPI): a framework that emphasises public purpose, open platforms, and competitive markets rather than purely proprietary digital ecosystems.
The question now is whether DPI offers a blueprint for financing and governing public AI systems — and if so, how far that analogy can stretch.
From Digital Public Infrastructure to Public AI
India’s DPI architecture — built around identity, payments, and data-sharing layers — has been framed as “digital plumbing”: foundational systems that enable downstream innovation. The approach combines state-backed investment with open protocols that private actors can build upon.
In principle, DPI is not synonymous with state ownership. It emphasises:
- Public interest over private monopoly.
- Open and interoperable platforms.
- Competitive markets built on shared infrastructure.
In practice, however, the state has played a decisive role through policy design, funding, and regulatory backing. Translating this model into AI governance raises a core dilemma: How much should governments invest in AI infrastructure without crowding out private innovation?
AI as Infrastructure: What Fits the DPI Paradigm?
Not all components of AI lend themselves equally to public investment. Governments often gravitate toward two high-profile domains:
- Large frontier language models.
- Domestic data centres and compute capacity.
These investments are frequently justified on grounds of digital sovereignty and strategic autonomy. Yet the AI value chain is complex. Control over compute does not guarantee control over models, training data, or applications. Moreover, demand for AI inference fluctuates, risking both underutilisation and supply bottlenecks.
Similarly, funding domestic frontier models may not yield high public returns. The global ecosystem increasingly offers open-source alternatives whose performance trails cutting-edge proprietary systems by only a short margin. For many public-interest applications, adapting open models to local contexts may be more cost-effective than building new ones from scratch.
High-Leverage Investments: Data and Shared Capabilities
If foundation models and large-scale compute are not ideal public infrastructure, where should governments focus?
One promising domain is high-quality datasets. Governments are uniquely positioned to curate representative, multilingual, and context-sensitive datasets. India’s AI governance guidelines already highlight the need for reliable and inclusive data infrastructure — especially where linked to Digital Public Infrastructure systems.
However, such data reuse must be accompanied by:
- Robust privacy safeguards.
- Clear consent mechanisms.
- Transparent governance frameworks.
Beyond datasets, governments can invest in shared AI capabilities that function as reusable layers. Examples include:
- General-purpose translation modules for multilingual societies.
- Open speech-to-text systems for low-resource languages.
- Context-sensitive evaluation benchmarks aligned with local needs.
Initiatives like Bhashini in India and Masakhane in Africa illustrate how shared linguistic infrastructure can empower downstream innovation without replicating expensive frontier models.
Innovation vs Diffusion: The Real Determinant of Impact
Much AI policy discourse focuses on fostering innovation. Yet long-term societal impact depends equally — if not more — on diffusion. The spread of affordable, usable AI tools across sectors such as agriculture, health, and education may generate larger developmental gains than investment in elite research alone.
Targeted public funding is especially justified in cases of market failure:
- AI tools for smallholder farmers.
- Disease surveillance systems.
- AI-based administrative tools within government.
Private actors may build such tools, but profitability considerations often limit equity, accessibility, and linguistic inclusiveness. Public funding can align AI applications with developmental priorities.
Balancing State Role and Market Dynamism
The DPI analogy underscores a broader policy insight: Infrastructure-based investments generate high leverage because they support multiple downstream uses. Yet infrastructure alone is insufficient.
Governments must avoid two pitfalls:
- Over-investing in capital-intensive projects with limited public return.
- Treating DPI as a rigid blueprint rather than a flexible framework.
In an era of fiscal constraints, AI investment decisions must weigh opportunity costs carefully. The state’s role is not to replicate private innovation, but to fill coordination gaps, correct market failures, and enable inclusive diffusion.
India’s Strategic Moment
By hosting the AI Summit, India is attempting to shape AI governance discourse from a Global South perspective. Its DPI experience offers a compelling narrative of frugal, scalable digital transformation.
Yet the lesson for public AI financing is nuanced: infrastructure first, but not infrastructure only. Governments should prioritise shared data layers, multilingual capabilities, and diffusion-oriented tools while remaining flexible enough to intervene selectively where public-interest goals demand.
What to Note for Prelims?
- Digital Public Infrastructure (DPI) refers to open, interoperable digital systems serving public purposes.
- Digital sovereignty debates concern control over data, compute, and digital platforms.
- AI governance includes privacy safeguards, interoperability, and accountability mechanisms.
What to Note for Mains?
- Examine whether DPI provides a viable model for public investment in AI.
- Discuss the trade-offs between innovation funding and diffusion-oriented policies.
- Analyse the role of the state in balancing AI sovereignty and market dynamism.
- Link to GS Paper II (Governance and policy design) and GS Paper III (Science & Technology, Digital Economy).