Plan infrastructure to power AI

OPINION By Godfrey Mutabazi Uganda has settled the argument that artificial intelligence (AI) matters. The harder question is whether the country is building the foundations that AI will require. At national scale, AI is not merely software. It depends on processors, data centres, high-capacity fibre, cooling, secure data and reliable electricity. Capability is advancing faster than the infrastructure needed to support it and it can be designed, financed and built. A computing hall can be running within two or three years; generation and grid connections often take far longer. If Uganda waits until demand becomes obvious, it may already be too late. The country is not starting from nothing. The national backbone infrastructure now connects more than 1,400 government and target-user sites. The National Data Centre hosts government cloud, backup and disaster-recovery services; by the end of 2024/25, the National Information Technology Authority of Uganda (NITA-U) recorded more than 300 applications for 100 entities. Last year, the Government launched a Data Centre Market Study to guide a third National Data Centre, scalable and AI-ready, with a planned primary site in Entebbe and disaster recovery in Jinja. Private interest is emerging too. Synectic Technologies has announced the Aeonian Project, a privately led high-performance computing development proposed at Karuma. Plans include a 10MW supercomputer and a larger modular campus drawing power from the Karuma Hydropower Plant. International firms have been associated with its design, but the facility is not yet operational, and its financing and timetable have not been fully disclosed. It should be treated as an important proposal, not existing national capacity. These foundations are encouraging, but they do not yet amount to a national compute plan. A market study, another government facility and a single private project will not establish how much capacity the economy will need as AI spreads through public services, finance, health, agriculture and industry, nor whether electricity, fibre and skills can grow alongside it. A modern AI data centre, built for continuous high-density computing rather than ordinary government hosting, sits beyond what the public purse can reasonably finance. The Government should turn existing studies into a co-ordinated 10-15-year National Compute and Data Centre Infrastructure Programme. Computational capacity should be treated as productive infrastructure. Because the state cannot underwrite such facilities alone, the programme should link demand to power, fibre, land, finance, regulation and skills, crowding in private operators and development partners rather than waiting for the Treasury to purchase an industry. The programme should map existing capacity and forecast demand under different growth paths. Uganda could model additional loads in modest, medium and large bands, planning figures rather than construction targets, and reconcile them with the market study so power, fibre and land can be phased with demand. It should identify compute zones with reliable electricity, high-capacity fibre, suitable land and room for expansion. They need not all sit in Kampala. Electricity and compute planning must be joined. Large AI sites need continuous, high-quality power and substantial cooling. Major investors can become anchor customers for new generation and transmission, so computational demand expands the electricity system rather than draining capacity needed elsewhere. Uganda should insist on efficient equipment, renewable power and water-conscious cooling. A site with abundant electricity but weak fibre cannot become a serious hub. Power and fibre corridors should be planned together, with redundancy built in against the failure of a cable or site. Uganda must also decide which computing capacity is strategically essential. That does not mean hosting everything at home. It means knowing which government systems must remain accessible, where backups should sit, and how essential services would continue through a cyberattack, power failure or network outage. The opportunity extends well beyond government hosting. Modern data centres can support cloud computing, cybersecurity, software, research and AI industries. Where public land, electricity or incentives make private projects possible, Uganda should negotiate measurable domestic returns: apprenticeships, research access, technology partnerships and affordable computing for researchers and start-ups. Support should depend on evidence of financing, technical readiness and implementation capacity. Universities cannot remain spectators. Computing infrastructure should feed engineering, computer-science and research programmes. Without engineers, network specialists, cybersecurity professionals and AI researchers able to operate complex systems, sophisticated machines may stand in Uganda while the higher-value work happens elsewhere. The country should not become merely a location where foreign-owned machines process foreign-owned intelligence. Co-ordination should use existing institutions rather than create another forum for meetings. NITA-U should lead a small group bringing together finance, energy, ICT, regulators, universities and industry. Its first task should be practical: establish current capacity, map requirements, identify sites and set out electricity and fibre needs, verify proposed investments and prepare bankable projects. Global System for Mobile Communications Association modelling suggests that appropriate digital reforms could add sh14.6 trillion in economic value and create 1.79 million jobs in Uganda by 2030. Data centres alone will not deliver those gains. Uganda has built important digital foundations. Readiness will be measured by whether these developments can scale, connect and generate value. Uganda must move from being AI-ready in principle to being infrastructure-ready in practice. Tomorrow's test is whether it has the power, compute, fibre and people to run that economy. The writer is an engineer