You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
In southern Taiwan, TSMC is quietly preparing what could become the largest fab expansion project in its history: a 507-hectare campus with plans for as many as 12 fabs, potentially dedicated to 1.4nm and even 1nm production.
The project is larger than TSMC's Arizona site and could define the company's Taiwan footprint for the next decade.
Yet building fabs is only half the story. The real challenge lies in navigating environmental reviews, ecological preservation, and the competing priorities of a society trying to balance semiconductor leadership with sustainability.
The battle over Shalun may ultimately reveal how far Taiwan is willing to go to keep TSMC's future anchored at home.
I think the TSMC's economists have determined that going for a "planned overcapacity" now is warranted on the peak of market upswing, given that the enormity of equipment costs now will deter late coming competitors.
On the latest investor call the big story was the increase in CAPEX for 2026 and the expected CAPEX for 2027. TSMC raised the CAPEX ceiling for 2026 from US$56 billion to US$64 billion. My guess would be US$64 billion will be spent if not more. We have been discussing this in the SemiWiki Forum…
I think the TSMC's economists have determined that going for a "planned overcapacity" now is warranted on the peak of market upswing, given that the enormity of equipment costs now will deter late coming competitors.
TSMC N3 is a cash cow since there was no competition. TSMC N2 looks to be the same although Intel 18-P is looking good. My concern is capacity. Intel 18A-P is going to be full of Intel products so how much foundry capacity can they really allocate? Especially if Elon Musk's Terafab is delayed and Intel has to take up the slack an 1.4nm.
Yet building fabs is only half the story. The real challenge lies in navigating environmental reviews, ecological preservation, and the competing priorities of a society trying to balance semiconductor leadership with sustainability.
TSMC N3 is a cash cow since there was no competition. TSMC N2 looks to be the same although Intel 18-P is looking good. My concern is capacity. Intel 18A-P is going to be full of Intel products so how much foundry capacity can they really allocate? Especially if Elon Musk's Terafab is delayed and Intel has to take up the slack an 1.4nm.
I am not talking about nearest competitor, but for things 7-10+ years down the line. The equipment will still be 10x the cost of optical sters, and they will still have the biggest factory on earth for those nodes.
How TSMC lost legacy 65nm+ nodes to mainland, despite having both insurmountable capacity, and cost advantage is a separate, and very interesting story, but in short, it was due to them simply refusing to compete for mainland clients, even when they had spare capacity, and general lack of commitment to maintaining dominance.
cwnewsroom.substack.com
