We are still in a data center driven investment era. As agents prevalence expands, more capacity will be built. Here are the “picks-and-shovels” winners: Delta Electronics is the standout on your chart — it shows up in PSUs, cooling fans, BBU systems, and CBU modules. As rack power densities climb with AI inference, the company that sells the power conversion, the fans, and the backup units is collecting tolls at every step. Vertiv — cooling fans, racks, busbars. AI racks run hot and power-hungry; cooling and power distribution is where the bottleneck physically lives. Memory makers (SK Hynix, Micron, Samsung) — inference is memory-bandwidth bound, and HBM is the chokepoint. Agent growth means more inference, which means more high-bandwidth memory. Samsung and Panasonic HD also show up in your chart's battery/capacitor columns, so there's overlap. Murata and TDK — they own the DC-DC converters and capacitors on your chart. Boring, essential, and present in every rack regardless of whose GPU is inside. The classic "sell shovels" play. One layer up: The hyperscalers, especially Microsoft, CPU is custom Azure silicon, and agents will live overwhelmingly on the big three clouds. If agent use explodes, cloud capex explodes with it, and Azure is positioned to capture a big share. Supermicro / Chenbro — chassis is more commoditized and lower-margin, so these benefit from volume but with thinner moats.