Arm is already priced as premium infrastructure (high-20s P/S), so the 5-year equity case is less “units explode” and more “Arm gets paid more per premium chip” (
Armv9 +
CSS), plus a real step-up in
server CPU royalty dollars as
hyperscalers increase core counts for always-on inference. The non-linear option is attaching new, recurring economics to “premium” designs (verification, outcome-linked performance tiers, certified subsystem ecosystems) that make Arm harder to substitute even if open ISAs improve. We assume some
multiple normalization, but not a collapse, because the model is asset-light and scales with partner shipments.