The “AMD trap” narrative feels more like a positioning reset than a structural breakdown. When expectations run hot, even good earnings can trigger a de-risking move.

What’s interesting is the rotation into “hardcore infrastructure.” ALAB and ANET are picks-and-shovels plays on AI scaling — connectivity, data flow, latency optimization. That’s structurally different from pure compute cyclicality. ON, meanwhile, is more tied to auto/industrial cycles, so it’s not the same AI beta.

The key question isn’t whether semis recover — it’s where operating leverage and pricing power are strongest in the next 12–24 months. If AI capex continues consolidating around hyperscalers, networking and interconnect may actually show more durable growth than GPUs themselves.

This feels less like a sector collapse and more like capital rotating toward the most defensible margins in the stack.