TSMC charging 6% more for the 3nm process seems reasonable. Article notes Apple accounts for 25% of their revenue and says this gives them leverage over TSMC but aren't they the only 3nm player in town?
I imagine the negotiations largely centre around timeline and quantity guarantees, as well as capability. Apple is cutting back iPhone Pro production [1]. If 3nm won't enable features that sell more iPhones, or if the market is shifting from a feature focus to one on costs, 6% more for 3nm may make sticking with five more profitable.
[1] https://www.bloomberg.com/news/articles/2022-09-28/apple-iph...
No, iPhone Pro demand is very strong. The regular iPhone demand appears soft or didn't ramp up. None is this is a surprise since early adopters would likely be the same who buy the Pros.
Besides with Perf/watt gap Mac/Iphone and rest Apple can wait it out for at least a year or maybe even longer.
Also worsening economic crisis will hit TSMC much sooner than Apple.
It's almost as if TSMC asking or an extra 6% when inflation is at an all-time high is reasonable, and Apple doesn't really have a viable alternative...
https://www.usinflationcalculator.com/inflation/historical-i...
On the other hand they have more money than they could possibly spend, so insourcing everything for vertical integration isn't out of reach.
Even posturing to begin that process would knock tens of billions off TSMC's market cap and spike its funding costs, to say nothing about complicating its contract negotiations with every other customer. They did it, after all, to Intel.
But all this is likely just negotiations and Apple won't be making a move (where to anyway?). Maybe instead of 6% per chip Apple invests a bit more elsewhere in the company to keep production costs as low as possible.
Isn't this misleading? If Apple doesn't take the fab capacity then somebody else will.
Is there? Seems like we're heading for a recession and reduced demand - at higher prices I wouldn't be surprised existing customers cut back on existing capacity.
Perhaps lower demand for consumer devices[1], but at Google and Amazon's scale, getting more efficient Tensor cores, ARM processors, and AI accelerators for their respective clouds will save them some OpEx by requiring less power (and cooling). It may also save CapEx as well, by reducing the need to build additional datacenters by getting +10% more compute[2] with the existing DC footprint, power and cooling budgets.
Server costs are amortized over long periods - longer than whatever recession may or may not happen. Also, if they can lock-in TSMC for years, they'll continue to reap benefits long after the recession is over and shift capacity to Kindles or Pixel devices when consumer demand recovers.
1. Which may explain Apple developing cold feet.
2. Back of envelope calculations. I do not know what node mixes Google and Amazon have in their data centers, but I'm guessing it averages-out between 10 to 15nm. How much can they benefit by replacing their oldest processors with 3nm parts?
There are many more possible reasons why Apple had dibs, including being able to guarantee specific volumes, or committing to pay large non-cancelable amounts as a deposit, and could be as pedestrian as TSMC preferring to deal with a single large order rather than 2 or 3 less large ones as they work out the kinks in a bleeding edge process.
You mean TSMC threaten? The only other huge players willing to pay a premium for state of the art node like that are GPUs, and they just got the whole crypto mining rug pulled out from under them so they may not be interested in paying top dollar any more.
Quantity varies though.
I also think the ability to improve tier prices as fast as they have is directly related to apple’s volumes.