← Back to Overview

Sources →

Global Inequality in AI: Detailed Data

The Central Paradox

Growth Rates:

Lowest income countries: Growth rate in AI adoption
Highest income countries: Growth rate / 4
Emerging economies growing 4x faster

Absolute Usage:

Singapore: 4.6x expected usage (AI Usage Index)
Nigeria: 0.20x expected usage
23x disparity (4.6 / 0.20)

Interpretation:

Poor countries adopting faster (percentage growth)
Rich countries using more (absolute levels)
Gap may be widening despite convergent growth rates

Regional Inequality Metrics

Anthropic AI Usage Index (AUI)

Definition: Multiplier of expected usage based on population and GDP

High-Usage Countries:

Singapore: 4.6x expected
Canada: 2.9x expected
Interpretation: Wealthy, technologically advanced nations

Low-Usage Countries:

India: 0.27x expected
Nigeria: 0.20x expected
Interpretation: Large populations, lower per-capita usage

Regional Patterns:

Wealthy English-speaking countries: High AUI
Large emerging economies: Low AUI (despite high absolute numbers)
Sub-Saharan Africa: Lowest AUI

Adoption Rate Gaps

Global North vs Global South (IMF Data)

Adoption Rates:

Global North: 23%
Global South: 13%
Gap: 10 percentage points (1.77x multiplier)

Trend Direction:

Gap status: Unknown (need longitudinal data)
Growth rates: Global South faster
Prediction: Gap may narrow in % terms but widen in absolute capability

Regional Adoption Snapshot

Regions (Estimated Based on Available Data):

North America: High (ChatGPT 59.7% market share base)
Europe: High (included in Global North 23%)
East Asia (developed): High (Singapore 4.6x AUI)
East Asia (developing): Medium-High (India growth leader but 0.27 AUI)
Latin America: Medium-High (Brazil growth leader)
Middle East: Medium (limited data)
Sub-Saharan Africa: Low (Nigeria 0.20 AUI, but 180% YoY growth)

Emerging Economy Leadership

Growth Champions

India:

Leading Gemini growth in Global South
Part of 22% of new account activations
Despite 0.27 AUI (low absolute usage)
Large absolute user base due to population

Brazil:

Co-leader with India for Gemini growth
22% of new account activations
Latin America’s AI hub

Mexico:

Highest adoption rates globally (per Modern Diplomacy)
Highest trust levels in AI

South Africa:

Highest adoption rates globally
Most active engagement
Leading Sub-Saharan Africa

Africa (Overall):

Gemini usage: 180% year-over-year growth
Fastest growing region despite lowest base

Why Emerging Economies Lead Growth

Hypothesized Factors:

Leapfrogging: Skipping desktop/PC era directly to mobile AI
No legacy infrastructure: Nothing to replace or retrain
Greater marginal utility: AI helps bridge resource gaps (education, healthcare info)
Mobile-first populations: Smartphone penetration enabling AI access
Educational gaps: AI fills absence of traditional educational resources
Language barriers reduced: AI handles multiple languages better than human translators
Youth demographics: Younger populations = higher AI adoption

Evidence:

Departure from historical technology adoption patterns
Previous pattern: Wealthy → Poor (sequential)
AI pattern: Simultaneous adoption, faster growth in poor countries

Infrastructure Barriers

Digital Access Gaps

Asia-Pacific Offline Population:

25% of population has no internet access
No internet = no cloud AI access
Absolute barrier to adoption

Gender-Technology Gap:

Women in South Asia: Up to 40% less likely to own smartphone than men
Smartphone = primary AI access device for emerging economies
Gender gap compounds regional gap

Implications:

Even with fast growth, 25% completely excluded
Gender inequality within regional inequality
Infrastructure must improve for sustained adoption

Device Access

Smartphone Ownership:

Critical for emerging market AI access
ChatGPT mobile vs desktop usage higher in poor countries (inferred)
Gemini mobile visits: 368.8M of 1.2B total (31%)

Connectivity Quality:

Bandwidth affects AI utility (slower responses)
Cost of data affects usage frequency
Urban-rural divides within countries

Economic Impact Projections

IMF Working Paper Findings

AI Growth Impact:

Advanced economies: X (baseline)
Low-income countries: <X/2 (less than half)
Ratio: >2x differential

Mechanisms:

Complementary infrastructure: AI needs skilled workers, capital investment
Sectoral composition: Advanced economies in AI-friendly sectors
Implementation capacity: Technical expertise concentrated in wealthy nations
Data availability: Rich countries have better data for AI training
Regulatory frameworks: Wealthy nations can afford AI governance

Warning:

Could reverse decades of narrowing development inequalities
Technology typically narrows gaps over time (e.g., mobile phones)
AI may be different due to complementarity requirements

Speed of Change

UN Development Programme Assessment:

AI adoption: Months (not decades)
Previous tech waves: Years to decades for global diffusion
Many countries lack:
- Infrastructure to support AI
- Skills to utilize AI effectively
- Governance systems to regulate AI

Implication:

Countries can’t gradually adapt
“Catch up” may be impossible at current pace
Risk of permanent AI have/have-not divide

Education & Skills Gaps

Literacy & AI

Reading Literacy:

Required to use text-based AI
Sub-Saharan Africa: Lower average literacy
Barrier even with device access

Digital Literacy:

Understanding how to prompt AI
Evaluating AI outputs critically
Unequally distributed globally

English Language:

ChatGPT, Claude, Gemini best in English
Non-English speakers face quality degradation
Compounds inequality for non-English regions

Educational Infrastructure

University Access:

Anthropic study: 1M university student conversations
Assumes university enrollment
Poor countries: Lower tertiary enrollment rates
AI education benefits concentrated among educated elite

K-12 Integration:

Gemini: 14.5M students via Google for Education
Requires institutional licenses and support
Wealthy school districts vs poor districts within countries

Platform-Specific Inequality

Gemini (Google)

Growth Leadership:

India and Brazil: 22% of new activations
Africa: 180% YoY growth
Strong emerging market presence

Reasons:

Google ecosystem prevalence (Android)
Integration with Google Search (familiar)
Free tier availability
Multilingual capabilities

ChatGPT (OpenAI)

Growth Metrics:

Lowest income countries: 4x growth rate of highest income countries (by May 2025)
700-800M weekly active users (global)

Accessibility:

Free tier widely available
Mobile app in many countries
Brand recognition even in poor countries

Claude (Anthropic)

Geographic Distribution:

Singapore: 4.6x AUI (highest)
Nigeria: 0.20x AUI (lowest)
Larger inequality than other platforms

Interpretation:

More concentrated in wealthy countries
Enterprise/professional focus = wealthy country bias
Coding use case = educated elite bias

Microsoft Copilot

Enterprise Focus:

90%+ Fortune 500 (wealthy country corporations)
Microsoft 365 requirement = wealth barrier
Least accessible to poor countries

Implication:

Workplace AI benefits accrue to rich countries
Widens productivity gaps between nations

Comparative Regional Analysis

North America

Strengths:

Highest market share (ChatGPT 59.7%)
Corporate headquarters (OpenAI, Anthropic, Microsoft)
Early access to new features
Infrastructure fully supports AI

Usage Patterns:

High absolute usage
Both consumer and enterprise adoption
Diverse use cases

Europe

Strengths:

Included in Global North 23%
Strong regulatory frameworks (GDPR, AI Act)
High infrastructure quality

Challenges:

Language diversity requires multilingual AI
Regulatory complexity may slow innovation

East Asia (Developed: Japan, South Korea, Singapore)

Strengths:

Singapore: 4.6x AUI (highest globally)
Advanced infrastructure
High digital literacy

Patterns:

Technology-embracing culture
Rapid adoption curves

South Asia (India, Pakistan, Bangladesh)

Paradox:

India: Growth leader + 0.27 AUI
Massive absolute numbers + low per-capita

Barriers:

25% of Asia-Pacific offline
Women 40% less likely to own smartphones
Infrastructure gaps

Opportunities:

Huge market potential
Leapfrogging possibilities
Youth demographics

Latin America

Leaders:

Brazil: Co-leader in Gemini growth
Mexico: Highest adoption rates + trust

Characteristics:

Mobile-first populations
Younger demographics
Language advantage (Spanish/Portuguese well-supported)

Sub-Saharan Africa

Extreme Contrasts:

Lowest AUI (Nigeria 0.20x)
Fastest growth (180% YoY for Gemini)

Barriers:

Infrastructure most lacking
Lowest internet penetration
Highest offline percentage

Potential:

Youth bulge (young population)
Mobile leap-frogging (skipped desktop era)
Massive untapped market

Middle East & North Africa

Data Gap:

Limited specific data in sources
Likely medium adoption
Oil-rich countries (UAE, Saudi) likely high AUI
Poorer countries likely low AUI

Inequality Dimensions

1. Access Inequality

Internet connectivity: 25% of Asia-Pacific offline
Device ownership: Women 40% less likely to own smartphones
Geographic: Urban vs rural (data unavailable but likely)

2. Usage Inequality

Singapore 4.6x vs Nigeria 0.20x (23x gap)
Global North 23% vs Global South 13%
Enterprise access concentrated in wealthy countries

3. Capability Inequality

Advanced economies: AI + complementary inputs = high impact
Low-income countries: AI alone = limited impact
Multiplier effect favors wealthy nations

4. Impact Inequality

Economic growth: Advanced economies >2x benefit
Labor market: Wealthy countries automate, poor countries lack jobs to automate
Educational: AI helps educated more than uneducated

Policy & Intervention Implications

Infrastructure Investment Priorities

Critical Gaps:

Internet connectivity (25% still offline)
Affordable devices (smartphone access)
Bandwidth quality (for real-time AI)
Electricity reliability (for sustained usage)

Recommended Actions:

Subsidized connectivity in underserved areas
Low-cost device programs
Public access points (libraries, schools)

Education & Skills

Curriculum Integration:

AI literacy in K-12 education
Prompt engineering skills
Critical evaluation of AI outputs

Language Support:

Develop AI models in local languages
Translation improvements for underserved languages
Culturally relevant training data

Governance & Regulation

Balance Required:

Protect users without blocking access
Enable innovation without exploitation
Ensure safety without creating barriers

Capacity Building:

Support regulatory expertise in poor countries
Share best practices globally
Prevent regulatory arbitrage

Measurement Challenges

Defining “Usage”

Metrics Vary:

Active users (but frequency unknown)
Messages sent (but quality/length varies)
Time spent (data unavailable)

Comparability:

Different platforms use different metrics
Self-reported vs measured usage
Free vs paid tiers (quality difference)

Country-Level Data Gaps

Missing:

Most countries not in published studies
Within-country variation ignored
Urban vs rural splits unavailable

Needed:

Comprehensive country rankings
Subnational data
Longitudinal tracking

Research Questions for Further Study

Causality: Does AI adoption increase inequality, or does existing inequality determine adoption?
Sustainability: Will fast growth in poor countries sustain or plateau?
Leapfrogging: Can poor countries skip stages and reach parity?
Use case differences: Are poor countries using AI for different purposes than rich countries?
Quality gaps: Even with equal usage, do rich countries get better AI experiences?
Second-order effects: How does AI-driven productivity in rich countries affect poor countries?

Data Gaps

Comprehensive country coverage: Only select countries in studies
Within-country inequality: No urban/rural, rich/poor regional data
Usage quality: Access ≠ effective usage
Longitudinal trends: Need multi-year tracking to see if gaps widen/narrow
Impact measurement: Economic outcomes from AI usage by region

Summary: The Inequality Dilemma

The Optimistic View:

Emerging economies growing 4x faster
Mobile-first populations leapfrogging
AI providing educational/health resources where lacking
Market forces creating incentives to serve poor countries

The Pessimistic View:

Absolute usage gaps remain enormous (23x)
Economic impact differential >2x favors rich
Infrastructure barriers exclude 25% entirely
Speed of change prevents gradual adaptation
Complementarity requirements favor wealthy nations

The Likely Reality:

Both occurring simultaneously
% convergence + absolute divergence
Creating new forms of inequality
Intervention needed to prevent worst outcomes

Critical Metric:

Not just who uses AI
But who benefits economically from AI
Early evidence suggests: Wealthy benefit more