The shape I'm hunting:

• A company the market decided was a commodity years ago

• Strong IP, captive customers, and manufacturing depth that takes a generation to build

• A narrative that has stayed wrong long enough that the valuation has hardened around it, building potential to snap with a violent reprice

Qualcomm might fit that shape.

The market filed them under "phone chips" and mostly stopped paying attention. By 2029, their own targets have two-thirds of revenue coming from non-handset sources. Automotive is already up 38% year-over-year with a $65 billion design win pipeline. The data center line is what I'm looking towards.

The thing I'm watching most is HBC. It's Qualcomm's new memory architecture built for AI inference. The claim is that it moves data dramatically faster while using a fraction of the power of what hyperscalers currently run. Satya Nadella named it on stage at investor day and said it was going onto Azure. That's the main external validation that gives signal that this is not pure hype, because as of right now the performance numbers are Qualcomm's own and haven't been independently tested yet.

Specs still need independent validation. Manufacturing partner unconfirmed. Both matter and I'm watching both.

When custom silicon ships to a hyperscaler this December and the first data center revenue line shows up on an earnings call in early 2027, every analyst who modeled a phone chip company has to ask what they're actually looking at. The label changes. The multiple follows.

SK Hynix looked like this before anyone was calling it that. More coming.