Chiplets for physical AI; rack-level considerations; verification complexity; sparse matrix workloads; AI data centers.