This paper explores the convergence of neuromorphic computing and sustainable AI, proposing novel architectures specifically designed for resource-constrained environments. Despite significant advances in artificial intelligence, current models face substantial energy consumption challenges, particularly in edge computing and IoT applications. We introduce a hybrid neuromorphic framework that combines spike-based processing with selective precision computing to achieve substantial energy efficiency while maintaining computational performance. Our experimental results demonstrate up to 87% reduction in energy consumption compared to conventional deep learning implementations, with minimal accuracy trade-offs. We further propose adaptive power scaling techniques that respond dynamically to computational demands. This approach represents a significant step toward sustainable AI systems that can operate effectively in environments with limited power resources. published by the Journal of Biodiversity and Environmental Sciences | JBES