Title: AI Maintenance Costs, Infrastructure Obsolescence, and the Challenge for African Educational Integration
Authors: Arinaitwe Julius, Ahumuza Audrey
Volume: 10
Issue: 3
Pages: 97-105
Publication Date: 2026/03/28
Abstract:
Background: The integration of artificial intelligence in African educational institutions has gained increasing policy momentum as a strategy for advancing equitable, quality education across the continent. However, the sustainability of such integration is threatened by the escalating recurrent costs of AI system maintenance and the pervasive obsolescence of digital infrastructure in African educational settings two structural challenges that have received insufficient empirical attention relative to their magnitude and consequence. Objective: This study examined the relationship between AI maintenance costs, infrastructure obsolescence, and AI educational integration outcomes in African educational institutions, with specific focus on the mediating role of institutional capacity. Methods: A quantitative cross-sectional survey design was employed, with data collected from 387 ICT directors, academic administrators, finance officers, and laboratory technicians across universities, technical and vocational institutions, and secondary schools in Uganda, Kenya, Nigeria, Ghana, and South Africa, selected through multistage stratified random sampling. A validated structured questionnaire was used for data collection. Univariate analysis described the distribution of key variables; bivariate Pearson correlation and independent samples t-tests examined pairwise associations and group differences; and Structural Equation Modeling (SEM), estimated via maximum likelihood with bootstrapped mediation testing (5,000 resamples), modeled the structural pathways among constructs. Results: Descriptive analysis revealed that technical staffing (M = USD 31,547) and hardware replacement (M = USD 22,417) were the costliest maintenance components, while composite AI maintenance cost burden (M = 3.84/5.00) and infrastructure obsolescence severity (M = 3.91/5.00) scores were high and AI integration sustainability perception (M = 2.61/5.00) was low across all institution types. Bivariate analysis demonstrated significant negative correlations between maintenance cost burden and integration sustainability (r = -0.614, p < 0.01) and between infrastructure obsolescence and integration sustainability (r = -0.658, p < 0.01), with institutional capacity positively correlated with sustainability (r = 0.702, p < 0.01). T-test analyses revealed that universities, privately funded institutions, and those in upper-middle income country contexts reported significantly higher sustainability perceptions, though all remained below the scale midpoint. SEM results confirmed excellent model fit (CFI = 0.961, RMSEA = 0.054) and revealed that institutional capacity significantly and partially mediated the negative effects of both AI maintenance cost burden (indirect ? = -0.199; 95% CI: -0.271 to -0.134) and infrastructure obsolescence (indirect ? = -0.216; 95% CI: -0.291 to -0.148) on AI integration sustainability. Conclusion: AI maintenance costs and infrastructure obsolescence constitute structurally embedded barriers to sustainable AI educational integration in Africa, operating both directly and through the erosion of institutional capacity. Coordinated multi-level policy responses encompassing dedicated recurrent maintenance funding, institutional capacity development, and the adoption of Africa-appropriate AI tool design and pricing frameworks are essential to realizing the transformative educational potential of artificial intelligence across the continent.