Key Details
The IMF working paper “AI and Economic Divergence in Asia” uses country-calibrated simulations to examine how economy-wide AI adoption could affect growth and inequality across Asia-Pacific economies.
Finding | What the Paper Shows | Why It Matters |
|---|---|---|
Uneven adoption | Capital-rich, highly skilled and productive economies are positioned to adopt AI at scale earlier | Existing economic strengths may determine which countries receive the first productivity gains |
Pressure on late adopters | Investment by early adopters could raise global demand for capital and financing costs | Emerging economies may face weaker investment and growth before adopting AI at scale |
Long-term growth | AI could eventually add 0.5–1 percentage point to annual growth in advanced Asian economies and 0.2–0.8 percentage points in emerging and developing economies | Delayed adoption does not eliminate the potential benefit, but postpones it |
Structural reforms | Sustained improvements in skills and productivity could advance adoption by one or two decades for some late adopters | Education, infrastructure and efficient capital allocation become part of AI policy |
Skill inequality | High-skilled workers generally receive larger wage gains; some lower-skilled workers may initially lose | Aggregate growth can coexist with wider wage disparities |
Generational inequality | Older, asset-owning households benefit earlier from higher returns to capital than younger, wage-dependent workers | AI may redistribute income between capital owners and workers |
Policy trade-off | Transfers can reduce inequality but may also affect labour supply and output | Redistribution must reflect national fiscal and demographic conditions |
India’s position | India’s young and expanding workforce delays capital deepening in the model and raises the need for employment-creating adoption | AI strategy must consider both productivity and the availability of work |
AI Could Produce Divergence Before Catch-Up
The paper examines AI adoption at scale — economy-wide use supported by investment in computing infrastructure, data centres, specialised chips, cloud systems, software and organisational capabilities. This is different from individual firms simply subscribing to existing AI tools.
Advanced Asian economies such as Japan, Korea and Singapore are better positioned for early adoption because they combine higher capital intensity, stronger productivity, larger pools of skilled workers and ageing labour forcesthat increase the value of labour-saving technology.
Many emerging and developing economies start from a different position. Lower capital intensity, skills gaps and rapidly growing labour forces make an immediate shift towards capital-intensive production less profitable in the model.
There is also an international spillover: investment by early adopters could increase global demand for capital and interest rates, raising financing costs for countries that have not yet adopted AI at scale. In the mild and moderate scenarios, emerging and developing economies experience annual pre-adoption growth around 0.5–0.6 percentage points below the no-AI baseline, with investment-to-GDP ratios around 5–6 percentage points lower. These are modelled scenarios, not forecasts.
Structural Readiness Determines When the Benefits Arrive
The simulations suggest that late adoption is not predetermined. A sustained reform programme that raises skills and improves the productivity of labour and capital can bring forward adoption and increase the eventual growth dividend.
The foundations extend well beyond AI-specific investment:
higher education and vocational training to expand the skilled workforce;
digital and physical infrastructure capable of supporting capital-intensive deployment;
more efficient allocation of finance towards productive investment;
stronger labour productivity;
greater organisational capacity within firms; and
regulatory systems that enable safe and productive AI diffusion.
Importantly, the paper finds these capabilities to be complementary. Skills without infrastructure — or finance without suitable workers and productive firms — may not be sufficient for economy-wide adoption. Isolated reforms therefore produce smaller gains than improvements across several structural constraints.
AI May Widen Inequality within Economies
AI-driven growth is not distributed evenly in the model. High-skilled workers benefit more because the technology complements their capabilities and raises the productivity of capital used alongside their work. Lower-skilled workers may initially experience weaker gains or temporary wage declines, although wages eventually rise above the no-AI baseline in several scenarios.
The model also identifies a generational divide. Older households tend to own more financial assets and therefore benefit earlier from higher returns to AI-related capital, while younger households depend more heavily on wages and receive gains more gradually.
Targeted transfers reduce inequality more effectively in the simulations, but impose greater output costs than universal transfers under the model's assumptions. The paper presents this as a policy trade-off, not a prescription.
The authors also caution against interpreting these results too literally: actual effects will depend on occupations and tasks, and some generative-AI applications may disproportionately improve the productivity of less-experienced workers.
India Faces a Distinct Employment Challenge
India is treated as a young economy with a growing labour force, giving it a different AI transition from ageing Asian economies seeking to compensate for worker shortages. India must simultaneously capture AI-driven productivity gains and create or improve employment opportunities for new labour-market entrants.
The simulations place India behind the leading advanced Asian adopters under faster AI scenarios. However, the modelled adoption dates represent an illustrative ranking of structural readiness, not calendar predictions. Indian firms are already using AI; the model concerns the later threshold at which capital-intensive AI becomes pervasive across the economy.
For India, the implication is therefore not to delay AI adoption. It is to strengthen the complementary conditions — skills, infrastructure, finance, firm capabilities and regulation — that allow adoption to spread while ensuring that productivity gains translate into broader employment and income growth.
What Is Capital Deepening?
Capital deepening occurs when the amount or productive quality of capital available per worker increases. In the paper, economy-wide AI adoption requires greater investment in computing, digital infrastructure and complementary business systems, raising the role of capital in production.
Policy Relevance
The paper connects AI preparedness with India’s wider development priorities rather than treating model development or computing capacity as sufficient measures of progress.
Link AI policy with employment strategy: Adoption should be assessed by whether it raises worker productivity, expands complementary occupations and creates viable entry routes for young workers.
Build complementary capabilities: Skills, electricity, connectivity, digital infrastructure and access to finance determine whether firms beyond a small technological frontier can use AI productively.
Focus on diffusion across firms and regions: Wider gains depend on adoption by smaller enterprises and less-developed regions, not only investment by large technology companies.
Prepare for occupational transitions: Training and employment-support systems should respond to changing tasks within clerical, professional, manufacturing and service occupations.
Track distribution, not only aggregate output: Monitoring should examine wage gains, employment, capital income and outcomes across skill, age, gender and regional groups.
Treat the numerical results cautiously: The model assumes economy-wide adoption, highly mobile capital and a single aggregate technology; it does not capture India’s influence as a large economy, sectoral differences, capital controls or the possibility of becoming an AI producer.
Follow the Full Paper Here: IMF Working Paper AI and Economic Divergence in Asia