Economy28 October 20265 min read
Using India's Digital Public Infrastructure Playbook to Make AI Work for the Masses
The Stakes and the Structural Risks
AI is projected to add USD 1.7 trillion to the Indian economy by 2035, and the promise is not confined to large corporations. India's cost geography favours cheap, narrow, task-specific AI — the kind that lets an MSME automate bookkeeping or inventory management without enterprise-grade budgets. ASHA workers could use AI-assisted tuberculosis screening tools to extend diagnostic reach into villages. Farmers could receive localised crop advisories in their own dialects. NITI Aayog projects that India could create up to 4 million AI jobs by 2031, spanning roles from AI trainers to ethical AI specialists.
Yet the same technology carries serious exclusionary risks. Task-level automation freezes the entry-level career pipeline. Fresh graduates find routine data processing and basic coding absorbed by algorithms, while a thin layer of experienced professionals commands steep wage premiums. NITI Aayog estimates that over 60% of formal-sector jobs carry some exposure to AI-driven task changes by 2030.
The environmental picture is equally uncomfortable. India's data centre sector already consumes an estimated 150 billion litres of water annually, a figure projected to exceed 358 billion litres by 2030, with over 70% of capacity concentrated in water-stressed cities — Chennai, Bengaluru, Mumbai, and Hyderabad — triggering local protests over threats to drinking water and agricultural irrigation. Underlying all of this is a dependence on Western hyperscalers for advanced chips, large-scale compute, and risk capital, which creates geopolitical vulnerability and risks entrenching monopolies. India's 490 million informal workers, roughly 90% of the workforce, remain largely outside the digital economy that AI would build upon.
The DPI Analogy and What It Demands
India's experience with Digital Public Infrastructure offers a tested counter-model. India Stack demonstrated that technology need not drive inequality. Its logic rests on open and interoperable technology, strong institutional rails, and low-cost market incentives that drive high-volume adoption. UPI provided an open payments protocol; NPCI provided the institutional backbone that kept the system neutral and accessible. Private players then competed on top of that foundation rather than controlling it.
The same architecture can be adapted for AI. Bhashini, the government's multilingual AI platform, plays the role of the open protocol, enabling developers to build regional-language applications without training models from scratch. AIRAWAT, the public computing infrastructure under the IndiaAI Mission, provides the sovereign compute layer, offering subsidised GPU access to domestic startups and academic institutions that cannot afford hyperscaler pricing. AI Kosha, the IndiaAI Datasets Platform, currently hosts over 5,500 non-personal public datasets across sectors including agriculture and health, reducing dependence on Western training data and the biases embedded in it. In agriculture, Bharat VISTAAR is designed to deploy local-dialect voice bots for personalised crop and weather advisories, bypassing literacy barriers entirely.
The Policy Architecture That Must Follow
The IndiaAI Mission, approved in March 2024 with a ₹10,371.92 crore outlay, is the central vehicle, but its components need to be operationalised with equity as an explicit design criterion. The IndiaAI FutureSkills pillar must accelerate the establishment of Data and AI Labs in Tier-2 and Tier-3 cities through ITIs, shifting the curriculum from basic coding toward AI literacy and prompt engineering. The MSDE-WEF India Skills Accelerator complements this by targeting the talent-retention challenge, aiming to ensure that the 4 million projected deep-tech jobs are filled domestically. The Anusandhan National Research Foundation (ANRF), backed by the Rs 1 lakh crore Deep-Tech Fund of Funds, is meant to catalyse private R&D investment so India becomes an innovator rather than a consumer of AI.
For the 490 million informal and gig workers most exposed to labour market disruption, portable social protection — health, pension, and unemployment benefits that travel with the worker across engagements — is a structural necessity, not a welfare afterthought. On the environmental side, data centre approvals must be conditioned on zero-liquid-discharge norms and the mandatory use of reclaimed wastewater, calibrated to regional ecological carrying capacity.
The governance layer ties this together. The Safe and Trusted AI pillar under the IndiaAI Mission, aligned with the proposed Digital India Act, must operationalise self-assessment frameworks to prevent algorithmic discrimination and deepfakes. The World Development Report 2026: The Promise of Artificial Intelligence recommends an Adopt-Adapt-Advance framework, prioritising locally adapted, cost-effective AI over expensive frontier models, a principle that maps directly onto what India's DPI model already does well. The question is whether the institutional will to execute at scale arrives before the window for equitable AI adoption closes.