From Divide to Delivery: How AI Can Serve the Global South

Commentary by Anjali Kaur

Published October 8, 2025

Next week's World Bank and International Monetary Fund (IMF) meetings will focus on rebuilding economic resilience in a fragile global economy. Yet amid these discussions in Washington, one issue risks being overlooked : The AI infrastructure and governance frameworks being built right now will shape development trajectories for decades. Whether that transformation is designed with the Global South or merely delivered to it will depend on decisions being made now on infrastructure, governance, and priorities.

AI is not just technology, it's infrastructure. It depends on compute clusters, energy grids, and network connectivity-assets rooted in specific geographies that determine where value is created and risks accumulate.

Treating the Global South as a unified category obscures more than it reveals. China has invested heavily in building domestic AI capacity-from chips to cloud platforms. India generates roughly one-fifth of the world's data but holds only about 3 percent of global data center capacity; it is data rich but infrastructure poor.

These disparities will determine who captures value from AI and who gets left behind. The IMF warns that AI could exacerbate cross-country income inequality, with growth impacts in advanced economies potentially more than double those in low-income countries. South Asia alone has nearly 100,000 young people entering the labor market daily, with almost half the region's 1.8 billion population under age 24. That demographic momentum could be a dividend or a disaster. Without compute infrastructure and workforce transition strategies, AI may erode the very labor advantages that once underpinned growth.

The India-hosted AI Impact Summit in 2026 will test whether the international system can address these dynamics. Success depends on three issues: infrastructure access, governance influence, and local adaptation.

Infrastructure: Access as a Foundation, Not an Afterthought

You can't use AI if you can't access it, and current access levels are wildly unequal. Countries lacking infrastructure-including compute capacity, energy grids, and connectivity-cannot build their own models or process their own data domestically. The IMF identifies access to computing power and data infrastructure as a critical determinant of which countries benefit from AI. Compute capacity remains concentrated in advanced economies. Africa accounts for less than 1 percent of global data center capacity despite being home to 18 percent of the world's population. India would need to nearly double capacity by 2026 just to meet domestic demand. Most of South Asia, Southeast Asia, and Latin America depend on external infrastructure.

AI is also energy-intensive. Estimates suggest training a frontier-scale model can consume on the order of thousands of megawatt-hours-a burden most fragile power grids cannot support. For countries with weak grids-where outages are common and capacity constrained-hosting AI infrastructure means not only building servers, but upgrading generation, transmission, and resilience. These upgrades often require capital and technical capacity beyond what governments can mobilize quickly, making sustained AI workloads infeasible without external support.

The Global Digital Compact, adopted in September 2024, recognizes digital connectivity as foundational to development. The World Bank and IMF should reflect that mandate by moving AI infrastructure from the margins to the core of what they finance. A century ago, ports and roads underpinned growth. Today, distributed AI infrastructure deserves the same level of priority.

Governance: From Consultation to Co-Design

Infrastructure alone isn't enough: Even with access to compute, developing countries need influence over how AI is governed and deployed. Most governance frameworks are being written in Washington, Brussels, and Beijing. The risk is that priorities get set without participation from those who will implement and use these tools. Conversations about which AI applications matter-crop disease detection versus automated trading systems, climate early warning versus content moderation, maternal health diagnostics versus facial recognition-happen without Global South governments at the table.

India illustrates both the challenge and the opportunity. Major AI companies are offering hundreds of millions of Indian users free access to premium tools, positioning the country as a "high-pressure testing ground" for AI development. India has recognized the governance implications, and its debates over data localization reflect a determination to move from data supplier to governance partner. The 2026 AI Impact Summit offers India-and its partners across the Global South-an opportunity to advance a more equitable governance model.

The World Bank, and the IMF should move beyond consultation to genuine co-design-embedding Global South representation on advisory boards, shared steering of project criteria, and joint oversight of safeguards.

Adaptation: Building Tools That Fit Local Realities

Even with infrastructure and governance influence, AI must fit local contexts. Diagnostic models trained predominantly on data from advanced economies often show reduced accuracy when applied to underrepresented populations. Agricultural AI systems trained on industrial farming data frequently fail in smallholder contexts where soil and crop conditions differ fundamentally. Most successful use cases remain isolated pilots rather than scaled solutions.

Adaptation also requires social fit, or aligning AI adoption with economic realities and demographic pressures. In countries with large youth populations, AI adoption collides with employment expectations An estimated 54 percent of South Asian youth lack skills for decent jobs, with over 30 percent not in education, employment, or training. In sub-Saharan Africa, nearly three-in-four working young adults are in insecure work. Without deliberate strategies to align AI adoption with job creation, demographic dividends risk becoming demographic disruptions.

Delivering that alignment requires workforce-transition programs tied to World Bank-financed AI projects, concessional financing that favors applications augmenting human productivity, and country frameworks tracking employment impacts. The World Bank's Digital Government Readiness Assessment still does not channel financing toward local AI ecosystems-the startups, universities, and cross-border collaborations building context-appropriate tools. Development finance should prioritize these ecosystems. South-South collaboration-through regional research networks, shared platforms, and pooled pilots-is where adaptation happens and how countries generate AI applications that serve their own priorities, such as food security, resilient health systems, inclusive education, and climate adaptation.

Policy Integration: Making It Real

Turning these principles into practice requires integrating AI into the World Bank and IMF's core instruments. Here are four moves they can make now:

1. Include AI indicators in country frameworks. These indicators include compute capacity per capita, connectivity reach, local workforce development, and employment impact ratios. If resilience depends on digital capacity, progress must be tracked.
2. Reward digital investment with concessional finance and more favorable lending terms for countries that meet digital public goods benchmarks.
3. Co-finance regional compute hubs in Africa, ASEAN, and Latin America. Pooling resources can reduce costs by as much as half compared to isolated national builds.
4. Develop a Global South AI Development Fund co-governed by representatives from the regions. This would ensure local priorities drive investments rather than depending on external institutions to set the agenda.
   

The Road to India's AI Impact Summit

The India-hosted AI Impact Summit in 2026 will test whether principles translate into practice. India has demonstrated what locally designed digital infrastructure can achieve-the Aadhaar identification system, Unified Payment Interface, and India Stack demonstrate how public digital goods can drive inclusion at scale when built for local contexts.

For India, the summit is an opportunity to translate that experience into global AI governance leadership. India's dual position-generating massive data flows while still building domestic AI capacity-gives it credibility to broker conversations between advanced economies and developing ones.

The World Bank and IMF should arrive with deliverables, not rhetoric: cofinanced infrastructure for shared capacity, co-governance platforms giving developing countries decisionmaking power, and financing frameworks rewarding inclusive digital investments.

AI won't solve poverty by itself. But without intentional design, it risks concentrating value and control in ways that make other development challenges harder to solve. The question is whether the World Bank and IMF will tackle sovereignty and equity head-on. The infrastructure being built now, the governance frameworks being written now, and the financing priorities being set now will shape development trajectories for decades.

Anjali Kaur is a senior associate (non-resident) with the India and Emerging Asia Program at the Center for Strategic and International Studies (CSIS) in Washington, D.C.

Programs & Projects

Report by Noam Unger and Madeleine McLean - August 13, 2025