Executive summary

This report examines the artificial intelligence (AI) dependency risks facing ten middle-income economies: Brazil, Colombia, India, Indonesia, Kenya, Mexico, Nigeria, South Africa, Thailand, and Vietnam. The analysis reveals that these nations participate in a global AI economy where the foundations—including services, hardware supply chains, and investment networks—are dominated by the United States and China. For example, 30 of the 42 most heavily used generative AI service providers globally are headquartered in the United States. China accounts for 33 of 50 first-place supplier positions for AI-relevant imported goods across the sample. The report suggests that full technological independence is neither achievable nor desirable, proposing instead a strategy of ‘managed interdependence’ to preserve policy autonomy and avoid lock-in to dominant powers.

Overview

Middle-income economies are active participants in the digital era but often operate in isolation from the technological foundations they rely upon. While many start-ups are homegrown, their managerial insights, financial capital, and technological inputs are frequently drawn from abroad. The study focuses on the situation as of the first quarter of 2026, selecting sample countries that have at least 100 active AI start-ups and participate in the OECD.AI Policy Observatory. The research highlights the need for these nations to diversify their sourcing and strengthen international cooperation to improve economic resilience.

AI in middle-income countries: Sectoral growth and dependency

A consistent tension exists between the potential gains from global AI networks and the structural vulnerabilities created by such engagement. Limited access to cutting-edge infrastructure risks deepening inequality, as AI advances may widen the gap between advanced and developing countries. Successful navigation of this transition requires economies to move beyond defensive postures toward active engagement. This involves using policy tools to ensure that integration generates local skills development and innovation capacity rather than merely serving as a passive market for foreign AI deployment.

AI sector leadership context

Scale and intensity in the internet economy vary widely. While India and Brazil show significant internet traffic volumes, the provision of AI services remains narrowly concentrated. Globally, the 42 most used generative AI services originate in just ten nations. Middle-income economies do not host any of the top generative AI service providers. Furthermore, leading AI bots and web crawlers are almost exclusively of United States origin, with only one originating from China. This concentration creates a risk of exploitative relationships for website operators in middle-income countries who receive fewer referrals back to their data sources.

AI trade dependency

Trade dependency is evaluated across five AI-related product groups, including computers, integrated circuits, and semiconductor fabrication equipment. Nearly three-quarters of the 50 analysed product-country pairings exhibit high trade concentration. China is the top supplier in most high-concentration cases involving large import flows valued at US$1 billion or more. In the services sector, official statistics are often masked by a statistical artifact; Ireland appears as the leading information technology services supplier for five of the ten sample economies because United States multinationals route revenues through Irish subsidiaries. True exposure to United States origins is therefore substantially higher than reported.

AI start-ups in 10 middle-income economies

The sample countries field approximately 9,554 active AI start-ups. India’s ecosystem is the largest with ,5458 firms, followed by Brazil with 1,964. AI start-up employment is a significant economic factor, estimated at 7.1 per cent of the total ICT sector employment across nine of the sample countries. In Vietnam, this figure reaches 12.3 per cent. Most firms are small-scale, with 11–50 employees, placing them structurally in the application and integration tier which is most exposed to platform dependency. Geographic concentration is also pronounced, with single hubs often accounting for the majority of national firms.

Investors in AI start-ups

Only 15 per cent of the sample start-ups have known outside investors. The investment landscape is characterized by a striking asymmetry between the United States and China. While Chinese entities act as infrastructure providers, they are nearly absent as equity investors in these start-ups. Conversely, the United States dominates the cross-border accelerator channel, accounting for nearly 60 per cent of such placements. This influence shapes which founders gain visibility and which technology stacks are validated, potentially creating long-term structural dependencies beyond simple financial flows.

AI policy and international economic relations

All ten sample countries have adopted national AI strategies, yet a gap remains between policy commitment and implementation capacity. According to the Government AI Readiness Index, middle-income economies underperform consistently on economy-wide AI infrastructure and resilience. While some countries like Thailand and Mexico score relatively high on preparedness indices, all lag significantly behind the G7 average. The binding policy constraint is increasingly execution rather than intent, requiring targeted investment in human capital within government and infrastructure seeding.

Thinking outside the box: Advanced economy partners and the path to managed interdependence

Cultivating alternative partnerships with advanced economies such as Canada, Japan, and the United Kingdom offers a path toward diversification. Canada provides leading AI research and a model for international development cooperation through programmes like AI4D. The European Union offers a developed risk-based governance template through the EU AI Act. Japan brings leadership in specialist manufacturing and foundation models for physical AI, while the United Kingdom provides expertise in AI safety and significant compute capacity. These partnerships aim for mutual benefit, helping middle-income countries capture gains while building improved institutional foundations.

Conclusions

The report recommends four key actions for middle-income economies: investing in domestic AI governance to close shortfalls in infrastructure and human capital; maintaining access to frontier technology from the US and China while actively diversifying ties with third-party advanced economies; engaging in international fora like the WTO and UN to shape emerging governance architectures; and deepening strategic engagement with partners seeking similar diversification. Advancing institutional quality, including predictable rule of law and reliable intellectual property protection, is identified as the most sustainable form of leverage available to these nations.