I wonder how this affects TSMCs business in the long run and if this is a move that is specifically geared towards foundries to come under US ownership. Interesting regardless.
Apple is single handedly bankrolling the latest and greatest from TSMC. If it wasn’t for them, the latest and greatest would likely be even more expensive/out of reach for “some companies”.
That's the whole point of the recent semiconductor investment boom, though. Reduce dependence on Taiwanese silicon so that we can care less about the geopolitics of the region.
The west essentially is getting tired of playing bodyguard to Taiwan; because as has been evident with the Ukraine invasion, the west is not willing to go to war to protect peace.
Russia and China cannot win a conventional war against the West. For the West, it wouldn't be easy, it wouldn't be quick, but it would be a certainty. Maybe it would even take a couple of wars, like Europe in the 20th century. The only way for those two terrible countries to win is if they use nuclear weapons, and use them quickly. Western democracies know this, and act accordingly.
US fabs will mostly service NATO and partners nations, i don't see them being able to compete in the global south on price given Chinese infrastructure, living costs of labor and domestic market size advantages which can fuel the R&D of Chinese semi designer studios and fabs.
TSMC doesn’t have much to fear in the near future.
Edit: I see where the confusion lies.
https://www.forbes.com/sites/patrickmoorhead/2022/08/10/inte...
This was announced two years ago. Intel hasn’t proven that it can be a third party manufacturer for anything close to cutting edge
Intel announced they would be producing chips for MediaTek two years ago.
https://www.forbes.com/sites/patrickmoorhead/2022/08/10/inte...
Intel has its 2nm node (called 20A, followed by 18A six months later) lined up for production to start in the first half of 2024, which compares to TSMC’s equivalent node (called N2) which is scheduled for the second half of 2025. 18A will be half a node denser than N2, while also being 12 months earlier to market.
As to IFS as a concept, it's only recently been revamped in 2021 after a lackluster attempt in 2013 as you mentioned. So far they've announced partnerships with Arm, MediaTek, Qualcomm, AWS, and the DoD. That's a pretty heavy hitting pack, not including all of Tower Semiconductors customers. Intel wouldn't be building another 3 fabs if it wasn't serious this time about building capacity for other companies.
- see: https://en.wikipedia.org/wiki/2_nm_process#:~:text=The%20ter...
https://en.m.wikipedia.org/wiki/Semiconductor_device_fabrica...
It would be interesting to hear what the consensus view is on Intel actually being able to meet this timetable. Surely a strong degree of scepticism is warranted given their recent record.
"We all take two years to develop one generation, how come you guys can do it in one or one-and-a-half year?" And they asked if some of your customer transfer technology to you or what not? And I told him, "No," I told him that, "That's not true." I think he probably implied we steal technology from customer, the way he talk.
And I say, "I'll tell you why." I said that, "When we develop one node, basically you have some learning cycles. First, you do some simulation. And you have some idea, then you run wafers to prove that. So, you run a group of wafers according to simulation and you have some splits. The wafer runs through the fab, they come out and you measure them, you analyze them, and you try to improve and you run this again. This again, you run. So, this is learning cycle." At that time, "It takes about six learning cycle, roughly, to complete one generation." Of course, you had some short loops and not just one. I said that, "My R&D wafer in the fab run much faster than yours, because my R&D engineer works three shifts and you only work one shift. So, your R&D wafer move eight hours a day, my work/move 24-hours a day. So, my wafers go three times faster, even if you are twice smarter than me, I still beat you up." <laughter>
https://www.computerhistory.org/collections/catalog/10279267...
Previous leadership was seriously mismanaging things. The following comment sheds some light on their stagnation during the 10nm/'Intel 7' node development period:
Intel has a long way to go to prove they can catch up.
https://www.forbes.com/sites/patrickmoorhead/2022/08/10/inte...
Intel isn’t dabbing cutting edge chips from them
Like their 7nm process has been delayed, what? 4 times? Have they even shipped it? Its the same story they had with the 10nm process too. IIRC they were behind like 5-6 years on that one too.
TSMC 7nm HPC: 66M Transistors/mm^2
TSMC 7nm Mobile: 96M Transistors/mm^2
Intel 7: 100M Transistors/mm^2
TSMC 7nm+: 115M Transistors/mm^2
Remember that unlike the others, Arm won't be an actual customer here. They don't require any capacity themselves.