The Impact of Unreliable Power and Internet on AI Research in Nigerian University Campuses: A PRISMA 2020–Compliant Critical Review of Campus Infrastructure Deficits and Their Effects on AI Development (2014–2026)
Abstract
Artificial Intelligence (AI) and Machine Learning (ML) research are intensely infrastructure- dependent. Successful deep learning training and experimental iterations require uninterrupted electricity for high-performance computing hardware and stable high-speed internet for accessing large datasets, cloud resources, and collaborative ecosystems. Nigerian university campuses represent the largest higher education system in Sub-Saharan Africa, yet they are characterized by severe grid power instability and bandwidth-constrained networks. This study presents a PRISMA 2020–compliant systematic and critical review mapping campus- level electricity and internet deficits in Nigeria and their downstream impacts on AI/ML research workflows between 2014 and 2026. Applying a rigorous search across Scopus, Web of Science, IEEE Xplore, and ACM Digital Library, we screened 405 initial records to synthesize a final corpus of 15 peer-reviewed studies meeting strict eligibility criteria. The synthesized evidence reveals that Nigerian university networks exhibit severe Quality of Service bottlenecks, with effective bandwidth allocations frequently falling below 0.5 Mbps per user, forcing 85–94% of researchers to rely on expensive personal cellular subscriptions. Power outages average 12 to 18 hours daily, leading to frequent interruptions of model training runs, high risks of hardware degradation, and complete reliance on techno-economic feasibility models over validated physical deployments. While direct causal metrics linking infrastructure to AI publications remain scarce— representing an urgent reporting gap—the structural bottleneck is shown to choke the data- ingestion and model-training phases of the AI research pipeline. We present a tiered, actionable governance roadmap spanning immediate "quick wins" (e.g., local dataset mirroring and safe- shutdown scripts) to long-term infrastructural designs (such as renewable-powered campus microgrids and zonal HPC centers) to salvage AI research capacity in low-resource national contexts.
Keywords
References
More Articles from INTERNATIONAL JOURNAL OF COMPUTER SCIENCE AND MATHEMATICAL THEORY
Author: Michael Arnold and Fabio Vitor
Author: Ngozi Samuel Uzougbo, Michael Ominyi, Cyril Chimelie Anichukwueze, Blessing, Chika Jones
Author: Lawal Ahmed Oladimeji, Achori Busayo, Akeju BusayoZainab, Saka Samson, Damilare, Mbah Demian Chidi, Runsewe Similoluwa Mayowa, Oladiti Luqman, Abiodun
Author: Okolo Clement, Eluemuno
Author: Chukumeka Gift Iroanwusi, Davies Isobo Nelson
