India risks a temporary growth drag as Asia’s early AI adopters bid up global capital costs, potentially undermining Viksit Bharat 2047 unless India accelerates structural reforms to hasten economy-wide AI adoption and mitigate unequal wage and generational impacts.

Main Finding

The commentary applies Che, Xin, and Yoshida (IMF, 2026) to India and concludes that economy-wide AI adoption can widen both cross-country growth gaps and within-country inequality. Because AI acts as capital-deepening, early adopters raise global demand for capital and interest rates, creating an interim growth and investment penalty for capital-scarce, young economies like India unless structural reforms (skilling, productivity, capital-market deepening) accelerate adoption timing. Without such reforms, India risks a pre-adoption drag during the critical 2025–2047 window; with reforms, adoption can be advanced and turned into a sustained growth accelerant.

Key Points

• Mechanism of divergence
  • AI modeled as a more capital-intensive, skill-biased frontier technology. Adoption occurs when an economy's effective capital-to-labour ratio crosses a profitability threshold.
  • Early adopters (capital-rich, ageing, skill-intensive economies) surge investment, raising global interest rates and raising financing costs for others.
  • Late adopters face a two-phase path: a pre-adoption headwind (lower investment & growth) and a post-adoption catch-up boost.
• Within-country distributional channels
  • Skill bias: low-skilled labour’s share and (in some scenarios) absolute wages can fall during transition; high-skilled wages rise.
  • Generational bias: rising capital share benefits older, asset-holding cohorts earlier than younger workers.
• India’s position
  • Calibrated as a young, labour-abundant EMDE with lower capital intensity and a smaller high-skilled share (in percent terms) despite large absolute skilled pools.
  • Likely faces delayed economy-wide adoption relative to Asian advanced economies and is therefore exposed to the interim global-capital-cost headwind.
• Quantitative magnitudes (regional/aggregate from IMF paper)
  • Pre-adoption growth drag ~0.5 percentage point annually vs. no-AI counterfactual in EMDE regional aggregation.
  • Investment-to-GDP ~5–6 percentage points lower during the pre-adoption phase.
  • Ambitious structural reform simulations can advance adoption by ~1–2 decades and substantially amplify gains (benchmarked to past East Asian transformations).

Data & Methods

• Core model: small open-economy overlapping-generations (OLG) framework with endogenous technology choice between incumbent and AI-intensive frontier technologies.
• Sample: 15 Asia–Pacific economies, calibrated to country-specific capital intensity, skill composition (three education-based skill classes), total factor productivity (TFP), and demographic profiles.
• AI scenarios: mild, moderate, accelerated AI-progress paths (common global technological frontier; adoption timing varies by country).
• Policy experiments: counterfactual no-AI path; simulations of structural reforms (raising high-skilled share, labour productivity, capital-specific productivity) and redistributive transfers.
• Key modeling abstractions/caveats: single-sector aggregation per country (uniform adoption date), stylized three-way skill proxy (education-based), ordinal rather than calendar forecasts for adoption dates.

Implications for AI Economics

• Cross-country divergence:
  • Demonstrates a robust channel—global interest-rate spillovers from early adopters—through which AI can increase international growth divergence even when global gains are positive.
  • Timing of adoption is endogenous and policy-sensitive; structural reform can materially alter ranking and timing.
• Distributional trade-offs:
  • AI-driven capital deepening can produce transitory (and for some groups persistent) declines in low-skilled wages and favor older cohorts via capital income—necessitating targeted redistribution and active labour-market policies.
  • Fiscal design matters: targeted transfers narrow inequality more but can be more distortionary; broad-based, low-distortion instruments (leveraging India's DBT infrastructure) plus retraining may be preferable given fiscal constraints.
• Methodological priorities for the field:
  • Need for sector-heterogeneous models that allow staggered, partial adoption across sectors (information- and capital-intensive vs. manual services and agriculture).
  • Better task- and occupation-level calibration (beyond education-based skill bins) to capture differential exposure to generative-AI and automation.
  • Integration of sovereign and cross-border capital-market dynamics into growth-and-distribution models of technology diffusion.
• Policy relevance:
  • For countries with young labour forces (like India), the critical policy lever is accelerating structural fundamentals (skilling, TFP, capital-market depth) to advance adoption timing so the post-adoption gains outweigh the interim headwind.
  • Complementary policies to preserve inclusion: scalable retraining, job matching, broad-based transfers via existing digital public infrastructure, and careful fiscal design to avoid large efficiency costs.
• Research implication:
  • Empirical work should estimate the magnitude of global capital-cost spillovers from early AI investment and test heterogeneous sectoral adoption patterns to inform more granular, country-specific policy prescriptions.