They'll have Apple's (likely immense) license fees for at least a decade - Apple won't make another architecture change anytime soon. And the rest of ARM's business... it's not like there are any alternatives. China hasn't managed to create a competitor, and even if they do it may very well be sanctioned and so useless to the US and EU. Intel/AMD's x86 is a dead end for the high-efficiency, low power market that ARM targets. PowerPC, MIPS and SuperH have been dead for years. There's always the possibility that RISC or some other open ecosystem rises, yes, but ARM is decades ahead and it would take absurd amounts of investments that I don't see appearing any time soon given the interest rate hike.
ARMs behavior lately is pushing a lot of riscv investment, my prediction is that Apple will migrate away from ARM 10years from now to something they can self-control, like riscv with some Apple-only extensions
Apple has shifted its processor technology every 11 years for its desktop machines.
But it has kept with ARM on iPhone for the last 15, so...
Anyhow, it is hard to predict with accuracy that far out.
That said, the first indications of a shift would be Apple adopting the RISC-V chip on the low end, things like the Apple Watch and Air Pods. Apple Watch has been using a 7N ARM chip since 2018 at nearly the same speed -- it seems like it doesn't need to be getter faster at each generation like the iPhone or MacBooks do. If adoption happens in this peripheral areas, then it is likely to move upstream to other devices over time.
Apple was an original investor in ARM and used ARM processors for Newton and a bunch of things like, I think, some of the time capsules.
I believe they are using RISC V internally, for some small things that are invisible to the end user. It would make sense, just to get a feel for it in case it matters down the road.
Its crazy to think about where things will be 10 years from now. Will we have beached 1nm? Then what? Is there actually a feasible roadmap for after that?
RISC-V's software ecosystem really is close already. I'd be hard pressed to name something that doesn't work well on RISC-V yet, besides DotNet, which recently merged RISC-V support and thus is still quite green.
Hardware-wise, SiFive's designs destroy ARM's own in performance/power/area. Tenstorrent Ascalon is to be released 2024, competitive with projected Zen5.
Apple's license fees are a good revenue stream, but that doesn't really mean ARM will do any business. Those fees are baked in, fire all the engineers and just distribute the fees to share holders through dividends. The question is what is ARM designing today that Apple will use tomorrow, I think the answer is likely "not much".
They sold their ARM stake in their near-death phase. They have an architectural license.
At this point I assume their whole ARM portfolio has been designed from scratch and they are paying those license fees for patent licenses and bc they still use the ARM instruction set. Cheaper than switching.
