I don't think anyone was under the impression that it's cost effective to manufacture in the US. The US push to get TSMC in the US is 1000% about risk mitigation... and I have a feeling that the nay-sayers from ROC might be concerned about losing political significance with their western trade partners.
> ...and I have a feeling that the nay-sayers from ROC might be concerned about losing political significance with their western trade partners.
This.
I live in Taiwan, and while I love the country and hope that it never falls to the CCP, I do believe that a primary reason TSMC's manufacturing remains so heavily concentrated in Taiwan is to increase the stakes for the West.
Very few Taiwanese I know believe that Taiwan is capable of repelling a Chinese invasion or attack without the support of the US, and protecting the Western world's access to semiconductors is a big incentive for the US to think twice about letting Taiwan fall.
While TSMC folks will tell you that their operation is so heavily concentrated for manufacturing and supply chain efficiency/optimization reasons, it is very unusual for a company as big and important as TSMC to have so little diversification in the geographical footprint of its operations.
> nay-sayers from ROC might be concerned about losing political significance with their western trade partners
It's more than that. TSMC's competitive advantage is their volume. It allows them to experiment and iterate faster than everyone - some say it's a big reason why Intel fell behind. Moving production into the US and out of TSMC's hands kills that.
> Cost effective over a long time horizon may not look cost effective over a shorter one
And cost effective over a shorter horizon may not look cost effective over a longer one. Just ask all those species that specialized to a niche then went extinct when that niche disappeared.
This problem will solve itself in the next few years when China retakes Taiwan. There's not a whole lot anyone can do about that, although I predict quite a bit of saber rattling. Open warfare is not an option (China has nukes), and "sanctions" are self-defeating, since we import far too much. If they leave an "out" (which they'd be wise to leave) the US establishment will take it, although the rhetoric will be quite severe I'm sure.
Indeed re-shoring high end silicon is an unambiguous bet on exactly this eventuality - there'd be no reason to spend the billions otherwise.
What are the reasons for his reservations about it? The article doesn't seem to cover that. Why does he call it impossible? I was under the impression the idea is to supplement the current supply chain, not replace it completely.
>"Chris Camacho, chief executive of the Greater Phoenix Economic Council that signed the agreement Tuesday, said his group is working to co-locate as many TSMC suppliers and other associated companies to the area as possible. He said about 40 companies are currently evaluating Arizona for possible investment, but declined to name them due to confidentiality agreements."
Apple contracts TSMC for their silicon in Taiwan, not China; some of the larger older operations are in mainland China but I understand that's the extent of it and only applies to the auto industry so the trade tensions don't impact this aspect of Apple's products or TSMC or Taiwan actually. They're also working on a partnership with Sony in Japan.
Current TSMC dominance is shortsighted happenstance that is not in any major power bloc's interest, least of all US. Yes the silicon shield protects TW, and exploiting it as silicon lance can inhibit PRC, but ultimately it's better for PRC for TSMC fabs to be within PRC rocket force ranges, on TW soil without explicit US security commitment than say Samsung Fab on SK soil. US could be pressing PRC a lot harder if not for these fabs.
There's also the broader consideration of East Asian IC supply chains in general, PRC/TW/JP/SKR accounts for ~75% of global IC capacity. Losing TSMC in war over TW will set the world back 1-2 nodes and 10 years, losing regional fabs in a broader war over TW will knock out decades worth of capex. Global semiconductor industry is 500B market, trillions in terms of industries affected, most of which benefits west. TW GDP is ~600B. TSMC concentrating fabs on TW is more liability than prize in that regard. East Asian semi supply chain in aggregate is more or less IC mutually assured destruction, and everyone of them can be directly threatened / leveraged by PRC if Sino-US competition goes hot.
TW losing semi leverage in medium / long term is forgone. The interesting development ahead is whether US will be able to hold on to IP dominance in the space and convince others to block tech to PRC. Recent high level talks in this regard is happening while US is pressuring non-US fabs for supply chain info to direct investments in US for the semi reshore effort. Ultimately goal is to undermine east asian semi industries the same way US killed Japan's semiconductor in the 80s. Right now every party is reshoring fabs while working towards de-americanized supply chains because everyone wants to sell to PRC and fears US sanctions that still gives preferential exemptions to US companies. EU/JP/KR car industry were not happy about TW/TSMC preferencing US companies while barginning for political gains with semi supply. Ergo recent pressuring of TSMC to spread fabs around with carrot and stick. While fabs are raking it during covid due to extreme demand and double booking, that will level off eventually. Latest Semiconductor Industry Association report is guestimating self sufficient local supply chains will cost at least 1T and 35-65% increase in prices. The strategic IC moat is eventually going to exclude a lot of smaller countries. Who will want TW as part of their blocs' supply chain if their fabs are existential risk? Queue Morris Chang.
Do we have the same concerns with Samsung being nestled so close to the mainland and North Korea? I myself don't imagine they'd do so well without US support either. If we're talking geopolitics, if Taiwan falls then frankly I expect Korea, part/all of Japan and the Philippines are next. If that impacts those foundries then I'd be curious how to regard former and current TSMC management's comments at that point in time/circumstance.
They have explicit US security commitment. Which potentially constrains how much US can respond in TW crisis. Roping in US assets in JP or SKR for TW defense will come with extreme costs. It's a risk US wants to mitigate, hence talk about reshoring entire supply chain even if extremely difficult.
Western countries can build up those bombed facilities in 6mths. The issue is just money which USA has plenty. China such move will doom her to be split apart. Every regional countries want that to happen even both Russia and NK (historical she is also due for such event). It would be incredibily stupid for CCP to do such a thing. It would be like they never read or heard of Sun Tzu. 1 or 2 leaders maybe, entire CCP chain of command?
Lead time for semi equipment is already years. Supply of qualified labour and expertise is constrained. Money is not the issue. Would take decades to rebuild existing capex. China broke apart meme negates the fact that when united, Chinese dynasties typically last 300 years if they survive first 50. Threatening east Asian supply chain is incredible smart because it resets world to PRC semi levels and forces US to fight within 1st island chain where PRC is strongest. It's massive deterrence to prevent fighting in the first place, per Sun Tzu.
This isn't the first time Morris Chang said this, even when he was still the CEO of TSMC.
> I was under the impression the idea is to supplement the current supply chain, not replace it completely.
To over simplify the problem and answer. There are lots of synergy having production all in the same place. From cost benefits, human resources, operation, and capacity planning. Breaking down these benefits will mean cost increase regardless the Fab is in US or not. You also need to factor in other things like testing and packaging. The whole supply chain, not just the foundry.
Some of them aren't even solvable with Tax Breaks. I still think TSMC should stick to Taiwan and not build Fabs in US.
But then this is currently done because of politics and National security. I would not be surprised if Apple Fab with Intel some time in 2025 - 2030. Just like how Apple helped TSMC, they will also fund Intel's Custom Foundry to a new level.
>But then this is currently done because of politics and National security. I would not be surprised if Apple Fab with Intel some time in 2025 - 2030. Just like how Apple helped TSMC, they will also fund Intel's Custom Foundry to a new level.
I should write it better. I am predicting Apple will Fab with Intel sometime in 2025 - 2030. And Fab their own Modem, possibly WiFi / Bluetooth with TSMC.
it was generalized as "supply chain" being regarded as "incomplete" leading to overpriced/uncompetitive positions; other articles on the topic with feedback from TSMC leadership add more detail relating to labor supply, inferior regional transportation/infrastructure/education and related; certainly will be interesting to watch how the Arizona regions and the various projects develop the next few decades
Shipping from a Singapore fab to SZ will always be cheaper than shipping from a Chandler fab. The real investment needs to be on board assembly in the NAFTA region.
The comments in the kube-system thread are spot on. I love how HN readers have quickly identified some of the real reasons Chang keeps making these claims.