FutureLens
Forecast intelligence
Forecast dossier

🤖 AI Divide: Will Global Inequality Surge Again?

A new UNDP report warns that artificial intelligence could reverse decades of convergence and reopen a "great divergence" between richer and poorer countries if left unmanaged.([undp.org](https://www.undp.org/asia-pacific/press-releases/ai-risks-sparking-new-era-divergence-development-gaps-between-countries-widen-undp-report-finds?utm_source=openai)) This forecast explores how global AI governance, compute access and skills investment might shape inequality over the next half-century.

Verdict: UNDP warns that AI could widen gaps between rich and poor states, reversing decades of convergence in income and human development (UNDP, 2025-12-02).([undp.org](https://www.undp.org/asia-pacific/press-releases/ai-risks-sparking-new-era-divergence-development-gaps-between-countries-widen-undp-report-finds?utm_source=openai)) The risk is highest where countries lack compute, skills, data and regulatory capacity to shape AI for their needs. Without targeted investment and global coordination, moderate divergence is more likely than renewed convergence.

Back to board
Date
Dec 2, 2025
Reliability
78
Harm potential
Medium

Scenario odds

Best Case

15%

Major economies treat the UNDP warning as a spur to act and create a funded global framework for inclusive AI development (for example, concessional compute access, skills programs and open models).([undp.org](https://www.undp.org/asia-pacific/press-releases/ai-risks-sparking-new-era-divergence-development-gaps-between-countries-widen-undp-report-finds?utm_source=openai)) Developing countries integrate AI into education, health and public services, raising productivity and state capacity. Inequality between countries stabilises and then slowly narrows as lagging regions leapfrog some legacy infrastructure stages.

Baseline

50%

Most countries publish AI strategies and a few multilateral funds emerge, but financing and implementation lag behind rhetoric.([undp.org](https://www.undp.org/china/press-releases/human-development-progress-slows-35-year-low-according-un-development-programme-report?utm_source=openai)) Advanced economies and a handful of large emerging markets capture most high-value AI industries and data platforms. Cross-country inequality edges higher, but not catastrophically, while pockets of success appear where local policy is strong.

Adverse Case

25%

Export controls, chip nationalism and proprietary frontier models lock most cutting-edge AI capabilities inside a small bloc of rich states and firms.([thestar.com.my](https://www.thestar.com.my/tech/tech-news/2025/12/02/ai-could-increase-divide-between-rich-and-poor-states-un-report-warns?utm_source=openai)) Poorer countries become primarily markets and data sources, with limited bargaining power and brain drain of top talent. Global inequality in income, skills and governance capacity rises sharply, fuelling instability and migration pressures.

Wildcard

10%

A powerful open-source or low-compute AI stack emerges unexpectedly, enabling many countries to adopt advanced capabilities without massive capital outlays. At the same time, AI-driven automation destroys key export industries for some low-income states before they diversify. The result is a patchwork: some late adopters surge ahead, while others face severe disruption and political upheaval.

Timeline projections

1-Year

🤖 Agenda-Setting Without Full Funding

Developments: Over the next year, UNDP's report and similar analyses will push AI and inequality onto agendas at forums such as the G20, UN General Assembly and regional development meetings.([undp.org](https://www.undp.org/asia-pacific/press-releases/ai-risks-sparking-new-era-divergence-development-gaps-between-countries-widen-undp-report-finds?utm_source=openai)) Several countries will launch or update national AI strategies that explicitly mention inclusion and human development but remain light on concrete budgets. Early pilot projects in social protection targeting, tax administration and health triage will start in a few low- and middle-income states with donor support.

Risks: Headline-grabbing fears around AI safety or deepfakes could crowd out attention to development-focused uses. Geopolitical tensions around chips and data flows may slow or complicate multilateral cooperation on inclusive AI infrastructure. Domestic politics in donor countries could constrain new funding, leaving strategies under-resourced and eroding credibility.

Outlook: Awareness of the AI inequality risk will rise faster than capacity to address it. Symbolic commitments will outnumber funded projects. The main uncertainty is whether early pilots generate enough evidence to justify scaling.

2-Year

Early Implementers and Emerging Fragmentation

Developments: Within two years, a cluster of middle-income countries with strong digital infrastructure will move ahead with scaled AI deployments in tax, customs, agriculture extension and education. Multilateral development banks will begin to package AI-related infrastructure and skills loans alongside traditional projects, but coverage will be patchy. A small number of cross-border AI initiatives in Africa, Latin America or South Asia will seek to share compute and models across several states.

Risks: Implementation capacity gaps may lead to failed or harmful deployments, from biased targeting systems to brittle automated decision tools that trigger backlash. Early failures in high-profile pilots could stigmatise AI in public administration and reduce political space for further experimentation. Fragmentation of standards and incompatible regional frameworks may raise barriers for smaller states trying to integrate into AI-enabled value chains.

Outlook: Some pathfinder countries will demonstrate credible, beneficial AI use in public services. Others will struggle, reinforcing perceptions that AI mainly benefits already capable states. The global policy conversation will shift from "if" to "how and for whom" AI is deployed.

3-Year

Diverging AI Capacity Curves

Developments: By year three, AI skills and institutional learning will diverge clearly between countries that invested early and those that treated AI mainly as a branding exercise. Universities and technical institutes in several emerging economies will graduate sizeable cohorts of AI practitioners linked to domestic public and private projects. A few least-developed countries will rely heavily on outsourced platforms from foreign providers or NGOs, gaining access to some functionality without building deep local capacity.

Risks: Reliance on foreign AI platforms could create new dependencies, with pricing, data sovereignty and service continuity controlled abroad. States with weak data protection and accountability frameworks may see AI deployments amplify corruption, discrimination or surveillance. Growing performance gaps between AI-enabled and lagging states might drive capital and talent flight, further entrenching inequality.

Outlook: The direction of travel toward moderate divergence will be clearer. Countries with deliberate capability-building strategies will show measurable productivity and service gains. Others risk locking into client relationships with foreign AI vendors and limited bargaining power.

5-Year

Structural Shifts in Trade and Services

Developments: In five years, AI will reshape trade in services, with data-rich countries exporting AI-enabled logistics, finance, creative and professional services across borders. Some commodity-dependent economies will begin using AI in resource management and revenue tracking, modestly improving governance and bargaining positions. Regional AI hubs in at least three developing regions will attract investment and talent, though often serving global firms as much as local needs.

Risks: Countries unable to plug into AI-enabled services trade may see their current export baskets lose value without obvious replacements. Concentration of AI infrastructure and platforms could create new systemic risks if a few providers experience outages, cyberattacks or governance failures. Political backlash against perceived "data extraction" or algorithmic colonialism could disrupt cooperation and investment flows.

Outlook: AI will be embedded enough to influence comparative advantage and trade patterns. Moderate divergence between AI-integrated and lagging economies will likely be visible in income and human development indicators. Whether this gap grows or stabilises will depend on the breadth of regional hubs and local value capture.

10-Year

Locked-In Advantages for Early Movers

Developments: After a decade, countries that combined digital infrastructure, education reform and regulatory capacity will have entrenched AI ecosystems spanning government, finance, health, agriculture and manufacturing. Their firms will occupy key rungs in global AI supply chains, from data labelling and model tuning to specialised applications. Human capital flywheels will spin faster as experienced AI professionals train new cohorts and found startups.

Risks: Latecomer countries may find it difficult to catch up because platforms, standards and ecosystems are already consolidated. Some AI-intensive states could experience internal inequality spikes if benefits accrue mainly to urban professionals, triggering political volatility that undermines outward success. Climate shocks or conflicts could abruptly degrade the data and infrastructure foundations that AI systems rely upon in vulnerable regions.

Outlook: Early movers will likely enjoy durable AI-enabled advantages in productivity and state capacity. Many others will depend on imported tools with limited scope to adapt them to local needs. The danger is a stratified global system where a minority of countries set AI rules and a majority absorb them.

20-Year

AI as Core Development Infrastructure

Developments: In twenty years, AI will be as foundational to development as electricity or telecommunications, embedded in education, health, taxation, social protection and urban planning. States that mastered AI-enabled governance and industrial policy will be substantially richer and more resilient, with stronger administrative capacity. A second wave of adopters may narrow the gap somewhat by importing mature, cheaper AI technologies and building around them.

Risks: A large group of chronically under-resourced states could remain stuck in low-AI equilibria, reliant on foreign platforms for critical services and vulnerable to external shocks. International institutions might struggle to manage cross-border impacts of concentrated AI power, from tax base erosion to security vulnerabilities. Technological breakthroughs such as highly autonomous systems could radically alter labour markets in ways that disproportionately harm late-developing economies.

Outlook: AI will be deeply woven into the fabric of state and market institutions. Unless there is sustained redistribution of capabilities, the divergence flagged by UNDP could become entrenched. Still, technological diffusion and regional cooperation could prevent the most extreme stratification outcomes.

50-Year

Long-Run Stratification or Managed Convergence

Developments: Over fifty years, AI and successor technologies will likely transform how knowledge, capital and governance capacity move across borders, potentially redefining what "development" means. Regions that built robust human capital, institutions and adaptable infrastructure could repeatedly upgrade their technological base and remain competitive. New alliances among developing countries might create alternative AI ecosystems, including shared public digital infrastructure and open knowledge commons.

Risks: If current trajectories harden, a small club of countries and corporations could control most advanced cognitive infrastructure, with weaker states locked into peripheral roles. Irreversible climate damage, demographic stress or large-scale conflict could intersect with AI-driven inequality to produce cascading humanitarian crises. Conversely, radical breakthroughs such as general-purpose autonomous systems might make existing models of national development obsolete, invalidating historical expectations.

Outlook: The long-run range of outcomes is wide, but path dependence suggests today's AI choices will matter greatly. Proactive governance and capability-building can keep the door open to later convergence. Neglect and fragmentation raise the odds of a hierarchically stratified global AI order.

Planning prompts to verify

  1. Create national AI development strategies tied to closing human development gaps, not just boosting GDP.
  2. Negotiate international support for compute, open models and skills in least-prepared countries through a dedicated AI inclusion facility.
  3. Embed hard distributional metrics in AI and industrial policies, tracking who gains by income level and region.