The integration of renewable energy sources, like solar power, is crucial for achieving sustainable computing infrastructure, especially in off-grid systems. The variable nature of solar energy often leads to surplus generation, particularly at midday, that cannot be efficiently stored or consumed. To address this challenge, we propose an AI-enabled demand response system that dynamically scales server utilization based on real-time solar availability. The system is also adaptable for integration with grid-connected networks for broader energy optimization. Leveraging recent advances in Artificial Intelligence (AI) and drawing from the peer-reviewed literature, reputable conference proceedings, and industry white papers, we evaluated an AI-driven framework for server scaling and resource allocation in Large Language Model (LLM) training. Real-time data on energy production and consumption, as well as battery storage, were collected from various sources, including Battery Management Systems (BMSs), Maximum Power Point Tracking (MPPT) units, and smart inverters. Supplementary solar forecasts were likewise integrated from third-party services (e.g., Solcast). These data were fed into a Tools AI Agent Node, and implemented within n8n, a low-code workflow automation platform, using the Ollama chat model gpt-oss:20b. This agent evaluates whether surplus energy would otherwise be unutilized for computational workloads and, if so, decides to use our proposed tool, activating or shutting-down the selected RAM-only provisioned servers optimized for LLM execution, effectively repurposing otherwise-wasted energy. Preliminary evaluations demonstrated high operational reliability (99%), near-real-time responsiveness (<1s latency), and accurate surplus energy detection. Workloads were successfully executed aligned with solar availability, validating the operational stability of the system. This research demonstrates that AI-driven demand response can repurpose surplus solar output into a valuable resource for sustainable computing, contributing to energy-efficient data centers and a broader transition toward renewable-powered infrastructures.