Designing the AI Factory from the Rack Up

For years, AI infrastructure was measured at the scale of the data center. That's changed. In this episode, Neeraj Paliwal explains why the rack has become the indivisible unit of a modern AI system — designed as one system rather than a collection of servers — and why performance is no longer defined by the compute in any single box, but by how efficiently data moves, is stored, and is secured across the entire rack.

 And getting there depends on co-designing the silicon, memory, interconnect, and packaging together, with the IP foundation turning concentrated power into usable compute.