The reference hardware I have been handling since January has no product name printed on it anywhere, and the people who could tell me what it will eventually be called are not allowed to. There is a name internally. It gets filed once the lawyers stop arguing about it, which by itself tells you Qualcomm thinks the thing is worth owning as a brand rather than shipping as another part number.
The name is the least interesting part. What I keep coming back to is that they are putting the cellular modem and the application processor on the same die and treating that as the product, not as a way to shave a dollar off the bill of materials.
Right now the accepted wisdom runs the other way. You buy a good application processor from TI or Intel, you buy a baseband from whoever, you wire them together, and you get to swap either one next cycle without redesigning the other. Best of breed, modular, safe. Every handset maker I talk to defends it in roughly those words, and the argument is not stupid: integration means you are married to one vendor’s radio roadmap, and if their modem slips, your phone slips with it.
I still think it loses, and not for the reason Qualcomm will put on a slide. The pitch will be cost and board area. Fine, those are real. The actual argument is power, and it is not the static kind you can measure by adding up datasheet numbers. Two chips in two power domains cannot coordinate. The application processor has no idea the radio is about to wake for a paging cycle, the radio has no idea the CPU just went idle, and so both of them sit at a conservative operating point because neither can afford to guess wrong about the other. Put them on one die under one power manager and that guessing stops. That is where the battery life comes from, and it is very hard to retrofit onto a two chip design later.
The part that surprised me is how little of this shows up in benchmarks. On a bench, running one thing at a time, an integrated part looks unremarkable next to a fast discrete application processor. The gap only opens when the radio is doing real work at the same time as the screen and the CPU, which is the condition an actual phone lives in permanently and almost no review ever reproduces. I expect this platform to be systematically underrated by exactly the people whose job it is to evaluate it.
Whether Qualcomm can execute is a separate question and I am not sure yet. Designing your own ARM compatible core instead of licensing one off the shelf is an enormous amount of rope to hand yourself, and the schedule I have seen is optimistic in the way engineering schedules are always optimistic in the first quarter of a project. Ask me in eighteen months whether the silicon landed anywhere near where they said it would.
I am leaving the operator side of this alone, because what integration does to carrier certification cycles deserves its own post and I do not have a clean answer yet.
If this works, the interesting consequence is not that Qualcomm sells more chips. It is that the modem stops being a component you shop for and starts being a thing you either have or do not, and a fair number of application processor vendors currently sleeping soundly are going to find that out late.