But TSMC can delay 3nm until it has the clients. Enough of the "smaller" players will eventually want and pay for it. Apple is far more dependent on TSMC than TSMC is on Apple. Where else is Apple going to go to realistically? It would likely take 3-5 years to fund another competitor to bring it up to what is needed.
> Enough of the "smaller" players will eventually want and pay for it.
At the right price, everybody wants it, but TSMC wants a high price now, to recoup their investment on the new process. If they drop their price now, that will take a long time, and may even be never, as, over time, competition will drive down margins.
Apple may be fairly unique in having a “right price” that is a lot higher than that of other players while they also are willing to order hundreds of millions units.
> But TSMC can delay 3nm until it has the clients.
My impression is that they invested huge sums of money into the 3nm lines; you don't want to delay getting returns on such investments.
Considering that capital has a cost, it might be cheaper for them to forego the 6% increase (at least for now) and start returning some of their loans, rather than letting the interest accumulate. (Which Apple knows, of course, which is why they play hardball).
From what I hear apple's latest chips are already so far ahead of amd64 in performance and power consumption that they can easily wait a couple of years. I do not think that this is what they actually want to do, i am only saying that they can even afford that for a few years if they have to.
I am not sure that there is enough demand from smaller clients to 100% cover apple's usual capacity.
Apple's chips are better largely because they're the only one on TSMC 5. If apple went to say samsung they'd end up on a worse 5nm process while AMD takes up the 3nm capacity Apple just ditched.
The vast majority of Apple chips are phones, not Macs. Samsung is neck & neck and their top end phones are definitely competitive, performing better in many ways for many people.
Apple is about long-term. Short-term, yes, accepting the hike is the cheapest. However, it opens you up for future price hikes. Yeah, they can't just switch to Samsung or acquire and uplift GloFo, join venture with Intel also won't work. Long-term, though, they can start working on this.
That would mean having sub-optimal chips for 3-5 years in their products while Qualcomm/AMD could finally catch up.
TSMC having undisputedly best litography is advantageous to Apple. It means Apple can buy the whole production and nobody else will be able to compete with their chips. They will lose this if they will make the field more competitive by propping up a competitor.
Sure, but good luck with that endeavor. Samsung and Intel are struggling to keep up with TSMC despite decades of expertise in the field and very deep pockets, so starting from scratch even with infinite money isn't gonna be a picnic. And won't solve Apple problem in the short term.
If there is one thing modern Apple is famous for, it is to dislike any situation where they depend on one single external vendor for anything. Apple wants either a somewhat healthy competition (e.g. with batteries, screens or assembly where misbehaving suppliers can be replaced rapidly) or vertical integration - and their acquisition of Intel's mobile modem business shows they think in long terms anyway.
In any case, Apple investing into their own fab also makes sense from a geopolitical point. Apple is already beginning to diversify from China because of their shoddy covid policies causing delays all the time, and everyone can see the writing on the wall that says China is very high on the sanctions priority list of the West. Additionally, the threat of China invading Taiwan and TSMCs fabs getting blown up in a scorched-earth action or damaged by war is only growing bigger every day.
They struggle because Apple paid upfront for capacity that didn't exist. That allowed TSMC to buildup capacity and secure lithography machines from ASML.
Intel struggled because of piss-poor decisions by execs:
- Reduce stake in ASML
- Not use EUV for 10nm
- 14nm++++++++++
- That was fine (for some time) because Intel had no competition from AMD
- Thinking that 4 cores 8 thread is what consumer needs (if it weren't for AMD we'd still top of the line i7, or how they call it today i9, would still have 4 cores and 8 threads)
- Not seeing that new node is getting more and more expensive in terms of R&D and that pure-play fabs have gigantic advantage
When AMD Zen happened, Intel got caught with its pants down.
Samsung "struggling" because they couldn't outbid TSMC for EUV machinery (wonder where TSMC got money from...)
Capacity for cutting-edge nodes is limited purely on how fast EUV machinery can be delivered. Currently, it's limited by ASML production capacity.