Computer-aided design (CAD) has become essential for hardware product development in our industrial age. However, increasing complexity, shorter lead times, and cost pressures present new challenges. While generative AI has gained significant attention and transformed various business functions, its application in engineering design with CAD remains underdeveloped. Our research aims to explore why generative AI has not yet reached its potential in CAD, despite its prominence in other fields, by identifying key challenges through case studies and a literature review. These challenges include small datasets, difficulty representing mixed data types, proprietary file formats, and lack of advanced CAD modeling commands. We propose future developments such as high-quality datasets, a vendor-neutral format, novel neural network architectures, and expanded generative methods.