The present analysis reframes African computer engineering as a problem of design under persistent intermittency, cost ceilings and maintenance scarcity rather than as delayed adoption of imported architectures. It asks why performance, energy and reliability models calibrated for stable grids and dense service networks mischaracterise feasibility across much of the continent and what alternative pathways remain viable when those assumptions are replaced. Through comparative synthesis of architectural, embedded-systems and networking scholarship alongside continental connectivity and development assessments, the article distinguishes three pathways — frugal embedded sensing, resilient edge connectivity and selective sovereign compute — and derives their trade-offs formally in terms of energy per task, availability under brownouts and lifetime cost. The contribution is a constraint-aware taxonomy with testable propositions and procurement criteria that link engineering choice to curriculum and maintenance capacity. The consequence is a practical basis for selecting architectures that remain operable where power and support are intermittent.