IMF AI Impact on Global Economy: Mind the Gap Analysis

The IMF’s Warning on AI-Driven Global Inequality

In one of the most consequential economic research papers of 2025, the International Monetary Fund has issued a stark warning: artificial intelligence will exacerbate cross-country income inequality, disproportionately benefiting advanced economies while leaving low-income countries further behind. Published as IMF Working Paper WP/25/076, The Global Impact of AI: Mind the Gap represents the most rigorous quantitative analysis to date of how AI will reshape the global economic landscape.

Authored by a team of seven IMF economists — Eugenio M. Cerutti, Antonio I. Garcia Pascual, Yosuke Kido, Longji Li, Giovanni Melina, Marina Mendes Tavares, and Philippe Wingender — the paper employs a multi-sector dynamic general equilibrium model of the global economy to quantify AI’s impact across different country groups. The findings are both revelatory and deeply concerning: the estimated growth impact in advanced economies could be more than double that in low-income countries.

This is not abstract theorizing. The IMF’s analysis carries enormous weight in shaping global policy debates, influencing everything from development aid allocation to trade agreements and technology transfer frameworks. As our analysis of the McKinsey State of AI 2025 shows, the enterprise AI revolution is accelerating — but this IMF paper reveals that acceleration is profoundly uneven across the globe.

The IMF Three-Factor Framework for AI Impact

At the heart of the paper is a powerful three-factor framework that determines how AI affects any given country. This framework moves beyond simplistic narratives about AI adoption rates to capture the structural conditions that shape whether a nation benefits from or is disadvantaged by AI.

The first factor is sectoral exposure to AI — how exposed a country’s economic sectors are to AI adoption and disruption. This varies significantly based on economic structure. Service-heavy economies with large financial, professional services, and technology sectors have higher exposure to AI’s transformative potential. Agriculture-heavy economies, by contrast, may see less immediate impact from current AI capabilities, though this could change as agricultural AI advances.

The second factor is AI preparedness — a country’s readiness to integrate AI technologies into its economy. This encompasses digital infrastructure (broadband, cloud computing access, data center capacity), human capital (STEM education, AI literacy, technical talent pools), institutional capacity (regulatory frameworks, intellectual property protection), and the broader business environment that enables AI deployment.

The third factor is access to essential data and technologies — whether countries can actually obtain the data and AI technologies needed to benefit. This is a critical bottleneck, particularly for developing nations that may lack access to foundation models, proprietary datasets, computing resources, and the technical partnerships necessary to deploy AI at scale.

AI Sectoral Exposure Across Global Economies

The sectoral exposure dimension reveals a fundamental asymmetry in how AI will affect different economies. Advanced economies, with their large service sectors — particularly in finance, professional services, information technology, healthcare administration, and education — have significantly higher sectoral exposure to AI. These are precisely the sectors where large language models, machine learning systems, and AI automation are having the most immediate impact.

Consider the contrast: in the United States, services account for approximately 77% of GDP, while in many Sub-Saharan African economies, agriculture still represents 20-40% of economic output. Current AI technologies are far more adept at automating knowledge work, financial analysis, customer service, and administrative processes than at transforming subsistence agriculture or informal economies.

This creates a paradoxical situation. The economies that are already the most productive are also the ones most exposed to AI’s productivity-enhancing effects. Rather than serving as an equalizer — bringing productivity gains to less developed economies — AI in its current form amplifies existing advantages. The World Bank’s Digital Development initiative has highlighted this structural challenge, noting that digital transformation benefits are concentrated in economies with existing digital infrastructure.

AI Preparedness: The Digital Infrastructure Divide

The AI preparedness factor is where the gap between nations becomes most stark. The IMF has previously developed an AI Preparedness Index (featured in the October 2024 World Economic Outlook) that ranks countries on their readiness for AI integration. The pattern is clear: advanced economies cluster at the top, while low-income countries lag significantly across virtually every dimension of preparedness.

Digital infrastructure is the most visible component of the preparedness gap. Advanced economies benefit from ubiquitous high-speed internet, extensive cloud computing infrastructure, data center networks, and mature telecommunications systems. Many developing countries still struggle with basic internet connectivity — the International Telecommunication Union estimates that approximately 2.6 billion people remain offline globally, the vast majority in low-income nations.

Human capital represents an equally critical dimension. The AI talent pool is extraordinarily concentrated in a handful of countries. The United States, China, the United Kingdom, Germany, Canada, and India account for the overwhelming majority of AI researchers and engineers. Low-income countries face acute shortages of technical talent, with limited university capacity in STEM fields and significant brain drain as skilled professionals migrate to higher-paying markets.

Institutional capacity — the regulatory frameworks, data governance structures, and intellectual property regimes that enable responsible AI deployment — varies enormously across nations. While the European Union has implemented the AI Act, the United States has established executive orders on AI safety, and China has implemented comprehensive AI regulations, many developing countries lack even basic regulatory frameworks for AI governance. As our analysis of state-level AI policy dynamics demonstrates, even within advanced economies, AI governance capacity varies dramatically.

Access to Data and AI Technologies: The Bottleneck

Perhaps the most concerning dimension of the IMF’s framework is access to essential data and technologies. Even countries with strong sectoral exposure and improving preparedness may be unable to benefit from AI if they cannot access the foundational technologies and datasets required.

The AI technology landscape is dominated by a small number of large technology companies — primarily based in the United States and China — that control the foundation models, computing infrastructure, and proprietary datasets that underpin modern AI capabilities. For developing countries, accessing these technologies typically requires either significant financial resources (to license enterprise AI platforms), technical capacity (to deploy and customize open-source models), or strategic partnerships with technology providers.

Data access presents its own challenges. AI systems require large, high-quality datasets for training and deployment. Many developing countries lack the digitized data infrastructure to generate the datasets needed for locally relevant AI applications. Furthermore, data sovereignty concerns and cross-border data flow restrictions can limit countries’ ability to leverage global datasets, while the absence of robust data governance frameworks may discourage technology companies from operating in these markets.

The compound effect of these access barriers is significant. A country may have sectors that could benefit enormously from AI, may be investing in digital infrastructure and education, but still be unable to capture AI’s benefits because it cannot access the technologies themselves. The UNCTAD Technology and Innovation Report has documented this technology access gap extensively, calling for international frameworks to ensure equitable technology distribution.

Growth Impact Disparity: Advanced vs Low-Income Economies

The paper’s headline finding — that AI’s estimated growth impact in advanced economies could be more than double that in low-income countries — emerges from feeding all three dimensions (exposure, preparedness, access) into a multi-region dynamic stochastic general equilibrium (DSGE) model. This is not a back-of-envelope calculation; it is a sophisticated macroeconomic simulation that captures the interactions between sectors, countries, and economic channels.

The mechanism is intuitive when understood through the three-factor framework. Advanced economies have higher sectoral exposure (more AI-ready sectors), better preparedness (infrastructure, talent, institutions), and superior access to data and technologies. Each advantage compounds the others: better infrastructure enables more effective AI deployment, which generates more data, which improves AI performance, which attracts more investment — a virtuous cycle that widens the gap with each iteration.

For low-income countries, the dynamics run in reverse. Lower sectoral exposure means fewer opportunities for AI-driven productivity gains. Weaker preparedness means those limited opportunities are harder to capture. And restricted technology access means that even the most promising applications may be out of reach. The result is a vicious cycle where AI widens rather than narrows the global income distribution.

This finding challenges the optimistic narrative that AI will be a great equalizer — giving developing countries access to the same capabilities as advanced economies. While AI does offer transformative potential for developing nations (in healthcare diagnostics, agricultural optimization, financial inclusion, and education), the IMF’s analysis suggests that without deliberate intervention, the net effect will be divergence, not convergence.

The Inverse Balassa-Samuelson Effect of AI

One of the paper’s most novel contributions is the identification of what the authors describe as an inverse Balassa-Samuelson effect. The traditional Balassa-Samuelson effect is a well-established economic phenomenon: productivity gains concentrated in the tradable sector (manufacturing, exports) drive real exchange rate appreciation in fast-growing economies, as wages in the more productive tradable sector push up wages — and therefore prices — in the non-tradable sector (services, local goods).

AI inverts this mechanism. Unlike previous waves of technological progress that primarily boosted productivity in manufacturing and tradable goods, AI has a particularly large impact in the non-tradable sector — services such as finance, healthcare administration, legal services, education, and professional consulting. When productivity gains are concentrated in non-tradables rather than tradables, the traditional exchange rate adjustment channel is disrupted.

The implications for international economics are significant. If AI-driven productivity gains are concentrated in non-tradable services, the expected pattern of exchange rate appreciation in fast-growing economies may not materialize as strongly as historical models would predict. This could affect everything from trade competitiveness calculations to current account projections and the effectiveness of exchange rate policies as adjustment mechanisms.

For policymakers, this finding means that conventional macroeconomic frameworks may need updating to account for AI’s sectoral composition of productivity gains. Central banks, finance ministries, and international institutions that rely on traditional productivity-exchange rate relationships for forecasting and policy analysis should incorporate AI’s non-tradable sector concentration into their models.

Regional AI Vulnerability Analysis

The IMF paper covers multiple regions, with particular attention to Africa, Asia and Pacific, the Caribbean, and the Middle East as areas where AI’s uneven impact will be most acutely felt. Each region faces a distinct configuration of the three-factor framework, creating different vulnerability profiles.

Sub-Saharan Africa faces perhaps the most challenging combination: low sectoral exposure (large agricultural and informal sectors), limited preparedness (digital infrastructure gaps, talent shortages), and restricted technology access (distance from AI development hubs, limited venture capital). The risk is that AI accelerates growth in advanced economies while African economies continue to rely on sectors where AI has less immediate impact.

Asia and Pacific presents a more heterogeneous picture. Countries like Singapore, South Korea, and Japan rank highly on AI preparedness and are likely to capture significant AI-driven growth. But large developing economies in South and Southeast Asia face substantial challenges in translating AI potential into broad-based economic gains, particularly in rural areas where digital infrastructure remains limited.

Small island developing states in the Caribbean and Pacific face unique vulnerabilities: their small market sizes, limited technical talent pools, and geographic isolation compound the structural challenges of AI adoption. For these nations, international cooperation and technology transfer mechanisms are not optional — they are existential necessities. Research from the OECD AI Policy Observatory further documents these regional disparities and the policy interventions needed to address them.

Policy Implications for Global AI Equity

The IMF paper’s findings carry profound policy implications at multiple levels — national, regional, and global. The central policy challenge is clear: how to ensure that AI’s transformative potential benefits all countries, not just the already wealthy.

At the national level, the paper’s framework implies that developing countries must invest simultaneously in all three dimensions: building sectors that can benefit from AI, strengthening preparedness infrastructure, and securing access to technology. This requires coordinated strategies across education, infrastructure, regulation, and industrial policy — a tall order for countries with limited fiscal space and competing development priorities.

At the international level, the findings argue for robust cooperation mechanisms to prevent AI from becoming another axis of global divergence. This includes technology transfer agreements that give developing countries access to AI capabilities, international standards that prevent digital protectionism, development finance targeted at AI infrastructure, and capacity-building programs for AI governance and deployment.

The IMF itself has a role to play. As the guardian of global financial stability, the Fund can incorporate AI-related structural factors into its surveillance and lending frameworks, advise member countries on AI-ready macroeconomic policies, and advocate for international cooperation on AI governance. The paper’s DSGE model provides a quantitative foundation for these policy discussions — moving the debate from qualitative concerns about AI inequality to specific, measurable projections that can inform resource allocation.

For more context on how financial institutions are navigating AI adoption, our coverage of the FSB monitoring of AI adoption in financial services provides complementary analysis.

Bridging the Global AI Divide: Path Forward

The IMF’s Mind the Gap paper serves as both a diagnostic and a call to action. The three-factor framework — sectoral exposure, preparedness, and access — provides a clear lens for understanding why AI’s benefits will be unevenly distributed. The quantitative finding that advanced economies could see more than double the growth impact of low-income countries quantifies what many have suspected but few have rigorously modeled.

The crucial insight is that while improvements in AI preparedness and access can mitigate these disparities, they are unlikely to fully offset them. This means that even ambitious programs to build digital infrastructure, train AI talent, and improve technology access in developing countries will not eliminate the AI growth gap entirely. The structural advantages of advanced economies — their service-heavy economic composition, deep capital markets, established technology ecosystems, and large pools of digital data — create a persistent advantage that targeted interventions can reduce but not erase.

This does not mean that efforts to bridge the AI divide are futile — far from it. The difference between a world where AI doubles the growth gap and one where targeted interventions narrow it to a smaller premium is enormous in human terms. It translates to millions of jobs, billions in economic output, and the life prospects of generations in developing nations. The IMF paper makes the economic case for action; the moral case hardly needs stating.

For policymakers, business leaders, and international organizations, the message is clear: AI governance cannot be an afterthought or a luxury reserved for wealthy nations. It must be a central element of global economic policy, with the same urgency and institutional commitment that the international community has brought to challenges like trade liberalization, financial stability, and climate change. The gap is already widening. The time to act is now.

Explore the full IMF Working Paper through the interactive experience above. For additional analysis of AI’s economic implications, see our coverage of NVIDIA’s State of AI Report 2026.