This paper presents a pilot initiative aimed at promoting equitable access to artificial intelligence (AI) technologies in public education through the reuse of obsolete computer hardware. In response to the growing environmental challenges posed by electronic waste and the increasing demand for computational resources driven by AI models, the study explores the refurbishment of a 15-year-old computer to support the deployment of modern deep learning systems. Specifically, the YOLOv11n object detection model and the lightweight language model Gemma 3:1b were tested under constrained hardware conditions. The study employed both quantitative and qualitative methods to evaluate model performance, focusing on inference time, resource consumption, and semantic quality of outputs. The results indicate that with appropriate configurations and lightweight models, recycled devices can support AI-based educational tools, particularly for children aged 5–6 years. The initiative not only extends the useful life of computing equipment but also contributes to reducing the digital divide and enhancing inclusive education in low-resource environments.