This study explores contemporary leading artificial intelligence accelerators and their cache design considerations, as part of a semester-long undergraduate research project. The undergraduate student researchers were identified as previous Computer Architecture students with strong fundamentals and were geared towards underrepresented groups. The team of two undergraduate students and the faculty members studied recent artificial intelligence processors and related accelerators from major companies, evaluated their design principles based on implicit target domains, and concluded main outcomes. The undergraduate research experience proved to be very instrumental to explore the core topics, with the student researchers having superior knowledgebase, in a very limited time period with limited resources. In addition, the cache design investigations have revealed that each major artificial intelligence processor cache components tend to follow their intended applications and related data as well as algorithmic processing requirements. The main research finding has shown that different companies have targeted optimizing their artificial intelligence accelerator cache units by following the domain specific architecture and organization specifications as well as implementation requirements.