Author: Danil Kerimi
Site of publication: https://t20southafrica.org/
Type of publication: Policy Brief
Date of publication: 2025
Diagnosis
There is clearly a gap in the market, but is there a market in the gap?
Countries in the Global South face structural barriers in scaling AI infrastructure and capabilities, including limited access to advanced computer infrastructure, scarcity of locally relevant data, a deficit in technical and governance talent, and underdeveloped domestic markets for AI innovation.
Despite many global challenges and deepening international fault lines, 2024 was a transformational year for AI governance. Building on the work of the High-Level Advisory Body on Artificial Intelligence, the United Nations General Assembly passed two resolutions on AI, adopted the Pact for the Future 3 and the Global Digital Compact, and created the UN Office for Digital and Emerging Technologies (ODET).
Relying on the market forces or waiting for these countries’ internal capabilities to catch up will not yield positive results in an acceptable timeframe. Many AI funds have been announced in the first few months of the 2025 (e.g. at Paris and Kigali AI Summits), with some further ones being in the planning stage. As in many other areas of development, donor coordination is going to be crucial to avoid duplication, overprescription, short-terminism, or vendor lock in. From AI bonds to debt forgiveness, many lessons for AI capacity could be learned from other areas of development finance (e.g. health, agriculture, climate, etc). Innovative financing can be deployed not only to close the gaps in the market but to help build the markets by providing advanced commitments for certain computational requirements for example, among other mechanisms.
Innovative Finance needed to Finance the Innovation
To thrive, AI systems equally rely on in-kind assets such as datasets, compute power, and human expertise. If all the resources are spent on procurement from hyperscale’s little will be left on nurturing local public and private AI ecosystems. However, the traditional overseas development and the UN system has struggled to design an architecture adapted to the needs of in-kind contributors.
From countries in special circumstances to development poster children, all jurisdictions could benefit from taking a hard look at their AI needs, AI posturing, and AI ambitions to ensure they are rooted in broader development priorities, competitive advantages and the stages of AI maturity.
Most LMICs begin with cloud-based computers supported by donors, often from abroad, but as AI maturity grows so do the calls for transition to sovereign infrastructure with many arguments for data sovereignty, cost efficiency, and innovation control. Users (consumers and citizens) need training and access to models suited to their languages and contexts, developers (creators and enablers) require compute and data tailored to local problems, while governments need institutional capacity and peer learning to implement national strategies. However, without sustained and Unlocking AI Capacity in Low-and Middle-Income Countries (LMICs): Innovative Solutions for Compute, Competences, and Competitiveness 6 tailored financing mechanisms, these journeys on the maturity curve are unlikely to achieve sustainable progress.
From AI Divide to AI Dividend
The G20 AI Principles, adopted in 2019 and grounded in the OECD AI Principles, provide the Group views on the foundational framework for responsible stewardship of trustworthy AI. These include principles of inclusive growth, sustainable development, human-centered values, transparency, robustness, security, and accountability. The rapidly evolving nature of AI ecosystems suggests that the G20 Principles would benefit from enhancement in two key areas.
First, Recognition of Infrastructure and Skills Gaps. The existing G20 framework rightly stresses values-based AI but stops short of proposing concrete pathways for achieving those goals in diverse economic contexts.
Key elements of G20 leadership in this area could include standardization of in-kind contributions valuation, convening power for multistakeholder coalitions, as well as the financial innovation and risk sharing.
G20 can set frameworks that quantify the economic value of data, compute power, and expert time. This would allow non-monetary contributions to be equitably accounted for in funding mechanisms. It can support jurisdictions (perhaps going beyond countries to municipal or inter-national level) in enabling multi-stakeholder partnerships involving development banks, cloud providers, academia, and civil society to co-create AI capacity projects with LMICs.
The G20 can build on its Data Gaps Initiative to track AI readiness and capacity metrics, aligning them with Sustainable Development Goals (SDGs) and enabling targeted support to lagging regions. Innovative financing, as proposed in this brief.
A global financing mechanism, supported by G20 members and aligned with the United Nations system, should be established to coordinate blended finance for AI in LMICs. These mechanisms could recognize in-kind contributions as capital assets, support modular, demand-driven investments across infrastructure, data, and skills and provide tiered funding for AI capacity at individual, institutional, and national levels.
G20 could further support the United Nations System and the multilateral financial institutions in their efforts to support local AI compute and data hubs, to provide shared infrastructure and multilingual datasets. These should lower barriers for LMIC AI developers, respect data sovereignty through local governance, foster collaboration and trust across regions, and support LMICs in designing AI development trajectories.