Nigeria’s proposal to share artificial-intelligence computing resources with other African countries is a smart and welcome step. It tackles a basic constraint head-on: many researchers and institutions across the continent simply don’t have enough computing power to train or run advanced models at a useful scale.
Nigeria’s Communications Minister Bosun Tijani proposed at the BUILD 2026 conference in Abuja on October 21, 2026, that African countries combine their computing capacity so researchers can access resources no single nation can provide alone. The proposal targets GPU-sharing arrangements with Rwanda, Senegal, and Kenya, but remains an ambition rather than a functioning system; its value will depend on practical agreements on access, costs, data protection, and the electricity and connectivity that underpin any compute infrastructure.
READ ALSO: Malawi’s Digital Inclusion: GSMA Roadmap Links Connectivity to Economic Opportunity
Africa’s position in the global AI economy is structurally marginal. Despite housing roughly 18% of the world’s population, the continent accounts for less than 1% of global data centre capacity, according to the Africa Data Centres Association and the World Economic Forum. Active capacity stands at approximately 360MW, compared to 55GW globally; even if all announced projects materialise, Africa is projected to maintain rather than increase its global share as hyperscale expansion accelerates elsewhere. This deficit translates directly into constrained research capacity: African researchers face longer project timelines, higher costs, and dependence on external platforms that are not designed for African languages or locally relevant problems.
Tijani’s proposal is specific: Nigeria would connect its GPUs with computing resources funded by the Gates Foundation in Rwanda, and with capacity in Senegal and Kenya. The logic is that while Nigerian capacity alone “might not be big enough” to train a large language model, aggregating resources across countries over 48 hours could enable researchers to complete demanding tasks. The minister acknowledged that infrastructure electricity, broadband, cloud computing, and access to advanced semiconductors remain the binding constraint, and that chip supply constraints, high costs, and geopolitical considerations compound the challenge.
The compute gap is not a theoretical problem. The World Economic Forum reports that African researchers, startups, and SMEs already face 7 million GPU hours of unmet demand for model training over the next three years, while only 5% of AI innovators have reliable access to advanced compute. This is a sign of latent demand held back by the absence of infrastructure rather than a lack of interest. For smaller African buyers, global supply dynamics favour hyperscalers, leaving them facing steep prices and long delivery times in some markets; the cost of a single high-end GPU rivals the annual income of an average citizen.
Nigeria’s contribution to any regional network rests on domestic initiatives already underway. The Nigeria AI Scaling Hub, launched in 2025 and hosted by Lagos Business School, received up to $7.5 million over three years from the Gates Foundation for technical assistance, compute infrastructure, policy support, and partnerships. The hub aims to provide free access to national AI computing infrastructure during its initial phase, reducing the cost of training models for participating innovators. Separately, Nigeria has launched a national AI Scaling Hub to provide computing resources to startups, public institutions, and researchers, a practical foundation, though not proof that a continent-wide system is in place.
Tijani identified electricity, broadband, cloud computing, and access to advanced chips as the key constraints. About 20 million Nigerians remain without internet connectivity, and the government is pursuing fibre-optic expansion. On electricity, Nigeria is exploring a proposed Nigeria–Togo data park in Egbe, where approximately 60% of available power capacity remains unused. The global context matters: the IEA estimates data centres consumed about 415 TWh of electricity globally in 2024, around 1.5% of worldwide consumption—a figure that illustrates the energy intensity of AI infrastructure and the need for African compute plans to be coordinated with power-sector planning.
Nigeria’s AI readiness has improved measurably, according to Tijani: the country’s position in the Oxford AI Readiness Index rose from 103rd to 73rd over two years. The government’s AI strategy aims to place Nigeria among the top 25% of countries in AI readiness. This progress provides context for the shared compute proposal; it is not a substitute for national investment but an attempt to leverage limited capacity more efficiently. However, the strategy itself acknowledges structural challenges: limited and unreliable infrastructure, low broadband penetration, R&D investment at approximately 0.2% of GDP against a global average of 2.2%, and a shortage of skilled AI professionals.
A regional computing network requires rules before it can deliver reliable access. The report identifies five governance areas: capacity and access (what hardware is available, how requests are prioritised), costs (who pays for electricity, maintenance, connectivity, and upgrades), data protection (whether data stays in its country of origin and what safeguards apply when it crosses borders), security and accountability (who is responsible for breaches or downtime), and long-term funding (how the system is maintained after grants end). Without these arrangements, shared compute risks remaining a conference ambition rather than a dependable service.
The conference’s focus on digital sovereignty is not simply about keeping all data within national borders. It is about having meaningful control over how AI systems are built, hosted, and governed, and avoiding dependence on a single external provider. A shared African system could strengthen that control if it enables local researchers to work with local datasets, develop tools for African languages, and choose where sensitive data is processed. But sovereignty also requires the capacity to maintain infrastructure and expertise; a system relying entirely on imported hardware, foreign technical support, and short-term funding may expand access without building lasting local capability.
Countries do not all need identical facilities. A coordinated network could specialise capacity, share technical knowledge, and make better use of existing infrastructure, but this requires common technical standards and transparent access rules. The BUILD 2026 conference, convened by the Nigeria AI Collective and supported by Luminate, AI Collaborative, Ford Foundation, and MacArthur Foundation, brought together government, researchers, civil society, businesses, and development partners. The Collective’s role is to align interests across sectors, but its impact should be judged by what continues after the event: funded research, usable infrastructure, clear standards, and deployed services.
The shared compute proposal’s success will depend less on the announcement than on its design: reliable power, affordable access, secure data governance, sustained financing, and transparent regional rules. The measures that matter are concrete: the amount of computing capacity available to African researchers, the number and range of institutions using it, the cost and waiting time for access, system uptime and energy cost per task, and the proportion of infrastructure and expertise maintained by African institutions.
If those pieces are built, shared compute could become a practical asset for African research and innovation; if not, it remains an ambition without dependable capacity. As Tijani framed it, the proposal connects “the little” each country has, but bridging the gap between connecting resources and building an ecosystem takes years, not conference sessions.
