Nigeria Proposes African Compute Pooling at BUILD 2026 Conference

Nigeria has proposed creating a shared computing framework among African countries to enable researchers to train and run complex artificial intelligence models across borders, according to reporting by THISDAYLIVE. Speaking through Dr Olubunmi Ajala, the national director of the National Centre for Artificial Intelligence and Robotics, Communications Minister Bosun Tijani presented the proposal at the BUILD 2026 conference in Abuja. The initiative seeks to connect computational capacity across nations, including Kenya, Rwanda, and Senegal, to overcome severe national resource limits.

Tijani stated that while individual African states often lack the graphics processing units and server scale to train large language models independently, aggregating hardware for 24- to 48-hour cycles would allow researchers to run advanced workloads. According to the conference organisers, Africa is home to 18% of the world's population yet accounts for less than 1% of global research and development expenditure. Tijani identified electrical grid instability, insufficient broadband, restricted access to cutting-edge microchips, and heavy reliance on external cloud services as primary obstacles to African AI sovereignty.

To counter these bottlenecks, the Nigerian government highlighted cross-border projects, including a planned data park between Nigeria and Togo alongside expanded fibre connections. According to figures presented by philanthropic funder Luminate at the gathering, Nigerian entities currently spend roughly $850 million annually on overseas cloud hosting. Tijani noted that although Nigeria climbed from 103rd to 73rd on the Oxford AI Readiness Index over two years, sustained regional cooperation on physical infrastructure is necessary to maintain momentum.

What it means

For developing nations, aggregating computing resources regionally offers a practical defence against foreign cloud dependency and allows Global South researchers to develop indigenous AI models within tight capital and energy constraints.