Carbon fiber composite materials are intensively used in many manufacturing domains such as aerospace, aviation, marine, automation and civil industries due to their excellent strength, corrosion resistance, and lightweight properties. However, their increased use requires a conscious awareness of their entire life cycle and not only of their manufacturing. Therefore, to reduce waste and increase sustainability, reparation, reuse, or recycling are recommended in case of defects and wear. This can be largely improved with reliable and efficient non-destructive defect detection techniques; those are able to identify damages automatically for quality control inspection, supporting the definition of the best circular economy options. Hyperspectral imaging techniques provide unique features for detecting physical and chemical alterations of any material and, in this study, it is proposed to identify the constitutive material and classify local defects of composite specimens. A Middle Wave Infrared Hyperspectral Imaging (MWIR-HSI) system, able to capture spectral signatures of the specimen surfaces in a range of wavelengths between 2.6757 and 5.5056 µm, has been used. The resulting signatures feed a deep neural network with three convolutional layers that filter the input and isolate data-driven features of high significance. A complete experimental case study is presented to validate the methodology, leading to an average classification accuracy of 93.72%. This opens new potential opportunities to enable sustainable life cycle strategies for carbon fiber composite materials.