Economic Survey Urges Reality Check on India's AI Strategy

India’s approach to artificial intelligence must remain anchored in economic reality rather than imitate the capital-intensive models of advanced economies, the Economic Survey 2025–26 said on Thursday.

The Survey frames AI as a series of economic choices constrained by capital availability, access to advanced computing hardware, energy costs and institutional capacity. It argues that India’s comparative advantage lies not in replicating hyperscale foundation models but in deploying AI in ways that raise productivity across sectors, create dignified employment and retain value within the domestic economy.

The Economic Survey marks a shift from last year’s emphasis on preparedness toward a more sober assessment of deployment risks, financial sustainability and structural constraints. A McKinsey survey cited in the survey shows that 88% of firms globally reported using AI in at least one business function in 2025, though only a small share have fully integrated it at scale.

The Survey’s concern is not whether AI will be used in India, but how.

Why hyperscale won’t work for India

The Survey contrasts two global AI trajectories. In the US and parts of Europe, development has followed a top-down path driven by massive private capital, concentrated intellectual property and heavy investment in frontier models and data centers. That model, the Survey argues, rests on assumptions that do not hold for India.

India faces binding constraints in computing capacity, access to advanced GPUs, and long-term financing for energy-intensive infrastructure. Pursuing scale for its own sake would therefore be inefficient and potentially destabilizing. Instead, the Survey advocates a bottom-up, application-led approach focused on sector-specific AI systems deployed across health, agriculture, education, finance, manufacturing and public administration.

This approach plays to India’s strengths as the country ranks among the top global contributors to AI research output, and has one of the world’s most AI-literate workforces, the survey said. Its diversity and scale generate rich domestic datasets that could support application-specific models, though the Survey notes that this potential remains underutilized, with Indian firms accounting for just 2% of startups focused on curating training data globally.

The economic logic is explicit. Application-driven AI systems require less capital, generate faster productivity gains and are more likely to diffuse across firms and regions. They also offer a clearer pathway to employment creation, especially when combined with human-in-the-loop systems rather than full automation.

AI and the labor market

A central theme of the chapter is that AI’s economic impact will be shaped more by how labor markets and institutions adapt. The Survey rejects the idea of immediate, economy-wide labor displacement. Instead, it emphasizes gradual task reallocation, with AI augmenting human work in most sectors.

This places human capital at the center of AI strategy. The Survey calls for scaling earn-and-learn pathways, curricular flexibility and skilling systems aligned to sectoral demand rather than generic AI training. Foundational cognitive and socio-emotional skills are identified as essential, particularly in primary education, alongside continuous reskilling for workers exposed to AI-enabled change.

Crucially, the Survey warns against over-indexing on elite AI talent pipelines while neglecting broader workforce readiness. The economic dividend from AI, it argues, will depend on diffusion across small firms, state capacity and public service delivery, not on a narrow concentration of technical expertise.

Data is treated as a strategic economic asset. The chapter argues that value generated from India’s domestic data should accrue primarily within the country, even as cross-border data flows remain open. This requires governance mechanisms that ensure accountability and regulatory visibility without stifling innovation.

The risks of premature regulation

On governance, the Survey takes a clear position against premature or omnibus AI regulation. It proposes sequencing policy in three stages: enabling experimentation, selective scaling based on evidence, and only then introducing binding obligations where risks and asymmetries are most pronounced.

Oversight, it argues, should sit with existing sectoral regulators rather than a single overarching AI law. Graduated obligations should reflect scale and use-case risk, while the AI Safety Institute should focus on scenario testing, red-teaming and international coordination, with clearly defined non-negotiable boundaries for high-risk applications.

The Survey’s caution is grounded in global experience. It notes that early adopters in advanced economies scaled AI under conditions of cheap capital and weak regulation, locking themselves into energy-intensive architectures and large financial commitments with uncertain revenue pathways. In some cases, discussions of government backstops have already emerged to manage downside risk.

India, as a later adopter, benefits from hindsight. The Survey argues that delayed adoption is not a weakness but an opportunity to design more resource-efficient systems, avoid fragile dependencies in global GPU supply chains, and align AI deployment with public objectives from the outset.

Choices under constraint

The chapter closes by framing AI as a strategic choice rather than a technological inevitability. India must decide what to build domestically, what to source globally, what to regulate early and what to allow to evolve. Passive consumption of AI technologies, the Survey warns, is the riskiest position of all.

Access to advanced computing hardware will remain a constraint in the medium term, shaped by concentrated global supply chains. India’s strategy therefore relies on diversified international partnerships alongside the gradual build-up of domestic semiconductor and computing capabilities linked to the National Semiconductor Mission, it says.

The Survey’s core message is deliberately unspectacular. AI will not transform India through scale alone. Its economic value will come from disciplined deployment, institutional coordination and sustained investment in people.

Properly sequenced, a development-oriented AI strategy can raise productivity, strengthen resilience and expand opportunity without replicating the financial and environmental excesses now visible elsewhere, it adds.