World Bank AI Foundations Report 2025: Key Findings on Digital Progress and Trends

Understanding the World Bank AI Foundations Report 2025

The World Bank Digital Progress and Trends Report 2025: Strengthening AI Foundations represents the most comprehensive data-driven assessment of the global artificial intelligence landscape published to date. Released as the second installment in the Digital Progress and Trends series, this landmark report examines how AI is reshaping economies and societies at an unprecedented pace, transforming how billions of people learn, work, and live across every continent.

At its core, the World Bank AI foundations report introduces a powerful analytical framework built around four essential pillars—the “Four Cs”—that determine a nation’s readiness to participate in the AI revolution. These pillars—connectivity, compute, context, and competency—provide both a diagnostic tool for understanding where countries stand today and a strategic roadmap for building inclusive, effective AI ecosystems that deliver broad-based economic benefits.

The report arrives at a critical juncture. Artificial intelligence capabilities are advancing faster than most governments can adapt, creating a widening gap between nations that lead AI innovation and those struggling to access even basic digital infrastructure. For policymakers, business leaders, and development practitioners, understanding these dynamics is essential for crafting strategies that harness AI’s transformative potential while ensuring no population is left behind. Readers interested in how AI research translates into productivity gains may also explore our analysis of NBER research on AI and productivity.

The Global AI Divide: Innovation and Adoption Trends

One of the most striking findings in the World Bank AI foundations report is the sheer concentration of AI innovation in a handful of wealthy nations. High-income countries account for a staggering 87% of all notable AI models developed globally, control 86% of AI startups, and attract 91% of venture capital funding directed toward artificial intelligence—all while representing just 17% of the world’s population. This concentration of innovation creates a self-reinforcing cycle: countries with more resources develop more advanced AI systems, which in turn attract more talent and investment.

Yet the picture is not entirely bleak for the developing world. Middle-income countries are emerging as increasingly active participants in the AI ecosystem, particularly on the adoption side. By mid-2025, more than 40% of ChatGPT’s global web traffic originated from middle-income nations, led by Brazil, India, Indonesia, and Vietnam. This surge in consumer adoption signals a genuine appetite for AI tools and suggests that with the right foundations in place, developing economies can rapidly integrate AI into their economic fabric.

The report also highlights a significant trend in AI employment. Generative AI job vacancies surged nine-fold between 2021 and 2024 globally, with approximately one in five of these positions located in middle-income countries. This data suggests that the AI workforce is beginning to decentralize, creating pathways for developing nations to build skilled AI workforces and strengthen their positions in the global digital economy. However, business and government adoption of AI remains in its nascent stages across most developing countries, indicating substantial room for growth.

Connectivity: The Gateway to AI Participation

The first pillar of the Four Cs framework—connectivity—addresses the most fundamental prerequisite for AI participation: reliable, affordable internet access coupled with dependable energy infrastructure. Without connectivity, populations cannot access AI-powered services, businesses cannot deploy AI solutions, and governments cannot leverage AI for public service delivery. The World Bank AI foundations report presents compelling data showing that while internet access continues to expand globally, massive disparities persist.

In high-income countries, 93% of the population uses the internet regularly. This figure drops to 54% in lower-middle-income countries and plummets to just 27% in low-income nations. These connectivity gaps mean that billions of people remain effectively excluded from the AI revolution before it even reaches their doorstep. The disparities extend beyond mere access to encompass affordability, connection speed, and data consumption patterns, all of which significantly impact the quality of digital participation.

Encouragingly, satellite internet technologies are opening new possibilities for closing remaining connectivity gaps, particularly in remote and rural areas where traditional terrestrial infrastructure has proven economically unfeasible. The report emphasizes that a cohesive policy approach combining public investment, private sector participation, and regulatory reform is essential to expand affordable connectivity—including the reliable electricity that powers it—and lay the groundwork for AI-ready economies. For a broader perspective on how digital innovation hubs are tackling infrastructure challenges, see our coverage of the BIS Innovation Hub initiatives.

Compute Infrastructure and the New Digital Electricity

The World Bank AI foundations report describes compute as “the new electricity in the AI era”—essential for powering AI applications but profoundly unevenly distributed across the globe. Computing resources, encompassing AI chips, data centers, and cloud computing platforms, determine a nation’s capacity to train, deploy, and scale AI systems. Without adequate compute infrastructure, even the most innovative AI strategies remain theoretical.

The numbers reveal a dramatic imbalance. High-income countries host 77% of global co-location data center capacity as of June 2025. Upper-middle-income countries hold just 18%, lower-middle-income countries account for a mere 5%, and low-income countries possess less than 0.1% of global capacity. This compute divide means that developing nations must either build expensive domestic infrastructure or rely on international cloud service providers, each option carrying distinct strategic and economic implications.

The supply side presents additional challenges. Compute resources are concentrated among a small number of technology firms, creating dependencies that extend well beyond traditional trade relationships. While many countries access compute through importing cloud services, this trade remains highly imbalanced. The report highlights that governments face a critical strategic decision: whether to invest in building domestic compute capacity through data centers and AI chips, or to focus on securing affordable, reliable access to international cloud infrastructure. Both approaches have merits, and the optimal strategy varies by country context.

Context: Why Local Data Shapes AI Capabilities

The third pillar of the Four Cs framework—context—addresses a frequently overlooked dimension of AI readiness: the availability of locally relevant, high-quality data for training and fine-tuning AI models. The World Bank AI foundations report makes clear that AI model capabilities fundamentally depend on the quantity, quality, and diversity of training data, and current data ecosystems heavily favor wealthy, English-speaking nations.

English still dominates AI training datasets, creating significant gaps in the ability of AI systems to serve populations speaking other languages or operating in different cultural contexts. This linguistic bias means that AI applications may perform poorly or even produce harmful outputs when deployed in contexts that differ substantially from their training data. For developing countries, the challenge is not just having enough data, but having data that accurately reflects local economic, cultural, and institutional realities.

However, the report identifies encouraging opportunities. Emerging data formats—particularly video and audio content—create new avenues for low- and middle-income countries to contribute to and benefit from the global data ecosystem. The growing data industry shows strong investor appetite for diverse data assets, suggesting market incentives are beginning to align with the need for more representative training data. Developing countries are increasingly navigating the landscape of open-source and proprietary AI models, seeking approaches that allow adaptation to local contexts while leveraging global technological advances.

Competency: Building AI Skills for the Future Workforce

The fourth pillar—competency—addresses what may be the most critical long-term determinant of AI readiness: human capital. The World Bank AI foundations report presents sobering data on the global skills divide. Less than 5% of the population in low-income countries possesses basic digital skills, compared with 66% in high-income nations. This gap in foundational digital literacy, let alone advanced AI competencies, represents a massive barrier to inclusive AI participation.

Digital skills have become a baseline requirement across many occupations, and AI is rapidly reshaping labor markets on every continent. The report documents that AI-related jobs are growing faster in middle-income countries than in high-income countries, suggesting a window of opportunity for developing nations to build competitive AI workforces. Generative AI skills in particular are expanding beyond traditional ICT roles, permeating sectors from healthcare and agriculture to finance and creative industries.

Yet many low- and middle-income countries face dual challenges in building and retaining AI talent. On the supply side, there is often a lack of quality, industry-aligned education and training programs that can produce graduates with relevant AI skills. On the retention side, brain drain continues to deplete the talent pools of developing nations, as skilled professionals migrate to wealthier countries offering higher salaries and better research infrastructure. Addressing these challenges requires coordinated investment in education systems, vocational training, and incentive structures that make it attractive for AI talent to work on local problems.

Small AI: Affordable Solutions for Developing Economies

Perhaps the most optimistic finding in the World Bank AI foundations report is the rise of “Small AI”—affordable, lightweight AI applications designed to run on everyday devices like mobile phones without requiring major underlying digital infrastructure. Unlike large-scale AI systems that demand enormous compute resources and sophisticated infrastructure, Small AI solutions can be deployed quickly and cheaply, making them accessible to populations and businesses that would otherwise be excluded from the AI revolution.

Small AI applications are already delivering measurable impact across developing economies. In healthcare, mobile AI tools help doctors and community health workers analyze patient data, screen for diseases, and make more informed treatment decisions in settings where specialist medical expertise is scarce. In agriculture, AI-powered applications provide farmers with crop disease diagnosis, weather predictions, and market price information, helping them optimize yields and incomes. In education, AI tutoring systems are expanding access to quality learning in communities where trained teachers are in short supply.

The report positions Small AI as a potential leapfrogging strategy for developing countries. Just as mobile banking allowed millions to bypass traditional banking infrastructure and access financial services directly through their phones, Small AI enables countries to harness AI’s benefits without first building the massive data center infrastructure that characterizes AI deployment in wealthy nations. This approach does not eliminate the need for foundational investments in the Four Cs, but it provides a pathway for countries to begin capturing AI’s benefits while those longer-term investments mature. For insights on how AI is diffusing across different economies, our analysis of Microsoft’s AI diffusion research offers complementary perspectives.

Open-Source AI and Democratizing Innovation

The World Bank AI foundations report identifies open-source AI technologies as a critical enabler of broader participation in the global AI ecosystem. While the development of frontier AI models remains heavily concentrated in high-income countries and a small number of large technology companies, the open-source movement is creating pathways for developing countries to access, adapt, and build upon these technological advances without starting from scratch.

Open-source AI models allow developers and organizations in any country to download, study, modify, and deploy sophisticated AI systems at minimal cost. This democratization of technology is particularly powerful for developing countries that lack the resources to train large AI models from the ground up. Instead, they can fine-tune existing open-source models using local data, adapting them to serve local languages, address local challenges, and reflect local institutional contexts.

The practical implications are significant. Open-source AI enables local innovation ecosystems to flourish, as developers can experiment with and improve upon global technologies without prohibitive licensing costs. It also reduces dependency on any single commercial AI provider, giving countries greater strategic autonomy in their AI development paths. The report emphasizes that supporting and participating in open-source AI communities should be a strategic priority for developing countries seeking to build sustainable AI capabilities. According to the OECD AI Policy Observatory, countries that actively engage with open-source AI tend to develop more robust and resilient AI ecosystems.

Policy Frameworks for Inclusive AI Ecosystems

The World Bank AI foundations report makes a compelling case that proactive government policies are essential to ensure the AI divide does not grow wider. Without deliberate policy intervention, market forces alone are likely to concentrate AI’s benefits in already-privileged countries and populations, exacerbating existing inequalities rather than reducing them.

The report outlines several priority areas for government action. First, countries must develop national AI strategies that are grounded in realistic assessments of their current Four Cs readiness and tailored to their unique economic, social, and institutional contexts. Cookie-cutter approaches imported from high-income countries are unlikely to succeed in environments with fundamentally different infrastructure, skills, and data landscapes.

Second, regulatory frameworks must balance the need to encourage AI innovation with the imperative to manage risks. This includes data governance policies that protect privacy while enabling the development of locally relevant training datasets, competition policies that prevent monopolistic control over critical AI infrastructure, and consumer protection measures that safeguard populations from harmful AI applications. The UNESCO Recommendation on the Ethics of AI provides a valuable reference framework for countries developing their regulatory approaches.

Third, the report emphasizes the importance of international cooperation. No country can build a complete AI ecosystem in isolation, and developing countries in particular benefit from partnerships that facilitate technology transfer, capacity building, and shared governance frameworks. The World Bank itself commits to helping countries harness AI for inclusive and sustainable development through technical assistance, financing, and knowledge sharing.

Implications for Digital Transformation Strategy

The findings of the World Bank Digital Progress and Trends Report 2025 carry profound implications for organizations and governments navigating digital transformation in the AI era. The Four Cs framework provides not just a national-level diagnostic but also a practical lens through which businesses, educational institutions, and development organizations can assess and improve their own AI readiness.

For organizations operating in or serving developing markets, the report’s emphasis on Small AI solutions offers a pragmatic starting point. Rather than waiting for perfect infrastructure conditions, forward-thinking organizations can begin deploying lightweight AI applications that deliver immediate value—improving decision-making, enhancing service delivery, and creating new efficiencies—while simultaneously advocating for and investing in the foundational infrastructure improvements that will enable more sophisticated AI applications over time.

The report also underscores the strategic importance of data as a competitive and developmental asset. Organizations that invest in building high-quality, locally relevant datasets position themselves to develop AI applications that outperform generic solutions in their specific contexts. This data advantage is particularly significant in sectors like agriculture, healthcare, and financial services, where local conditions vary dramatically and one-size-fits-all AI models often fall short.

Ultimately, the World Bank AI foundations report delivers a message that is both urgent and hopeful. The AI divide is real and growing, but it is not inevitable. With strategic investments in connectivity, compute, context, and competency—combined with smart policies, international cooperation, and a willingness to embrace innovative approaches like Small AI—developing countries can not only participate in the AI revolution but shape it to serve their own development priorities. The window of opportunity is open, but it will not remain so indefinitely. For organizations committed to making complex research accessible and actionable, tools that transform static reports into interactive experiences play an increasingly vital role in bridging the knowledge gap.