New rack-based AI acceleration hardware is being positioned as a cost-effective and straightforward way to power AI inference workloads ...