Building, Not Borrowing: Financing Compute for the Developing World

For most of the developing world, artificial intelligence arrives as an import, and the divide of the coming decades will be about compute rather than connectivity. This article argues that compute access, largely absent from digital-development financing frameworks, is now the binding constraint, and that grant dependence, value leakage and the high cost of capital in institutionally thin markets make the problem worse. It sets out a Dual Core Finance model that capitalises a Sovereign AI Development Fund through concessional anchor capital and community-operated velocity capital, completed by Algorithmic Impact Bonds, Complementary Asset Portfolios, governance de-risking and Minimum Viable Innovation Engines. Drawing on the Neural Network (AI) Lab of the Ministry of Foreign Affairs of Bangladesh and the AI-assisted transformation of its 85 missions, it closes with five recommendations for the WSIS+20 and Global Digital Compact agenda. Written in a personal capacity; the views expressed are those of the author and do not necessarily represent the views of the Ministry of Foreign Affairs or the Government of Bangladesh.