When it begins production in early 2024, TSMC's fab in Arizona will be the foundry's first advanced fab in years located outside of Taiwan. According to recent reports, while the fab will operate in the USA, it will be at least partially built in Taiwan in order to save some money on setting it up and to ensure a smooth ramp-up of the production facility.
In a bid to ship parts of the fab from Taiwan to the U.S., TSMC is expected to use as many as 4,000 to 5,000 containers and spend around $110 million on transportation, reports Taiwan News. Han Tang, a cleanroom engineering firm from Taiwan, Evergreen Marine Corp., and Wan Hai shipping companies are expected to be involved in the project. It is unclear what exactly TSMC plans to build in Taiwan before shipping it to the USA, and the company hasn't commented on the matter.
Modern fabs use hundreds of different tools that are made all around the world and then installed and set up locally. ASML, which makes lithography scanners, installs EUV tools itself, and it is unlikely that TSMC will ship these scanners to Taiwan, have them assembled and set up, and then ship to the U.S.
There are also tools made by companies like Applied Materials, KLA, and LAM Research in the USA, which would also be shipped directly to Arizona. Meanwhile, there are custom-built tools used by TSMC that are made in Taiwan. For the contract maker of semiconductors, it might be easier to set them up in Taiwan and then ship to the USA rather than to set them up locally since developers of such equipment are in Taiwan.
TSMC has built its advanced fabs in Taiwan for many years. Furthermore, the company tends to produce chips using a particular node at the same fab (albeit in different modules/phases), so the company does not have a program akin to Intel's Copy Exactly that allows the chip giant to quickly set up production at a new location and maintain experience and fab equipment configurations across different sites. To that end, it is not going to be easy for TSMC to build a 5nm-capable fab in the U.S. and maintain the same fab configuration and yields as it does in its Taiwan-based fabs.
Perhaps this is why TSMC reportedly wants to at least partly build the fab equipment in Taiwan and then ship it to the U.S. via cargo ships. Another reason why TSMC reportedly opts for such an unusual way of building a production facility could be the fact that its fabs and cleanrooms are architected in Taiwan and are designed to be built by local companies.
TSMC does not comment on its fab plans and we cannot verify what exactly the company plans to build in Taiwan for its Arizona fab. Meanwhile, as ironic as it may sounds, but it looks like TSMC's first advanced U.S.-based fabs in years will be partly built in Taiwan.
The US has a problem with construction and infrastructure projects going awry. Where I live, we can't even build a new light rail line or replace a pair of bridges (seriously, it's been 10 years; why isn't this project finished?! The Romans were faster). In other cities, we've seen images of new construction buildings partially collapsing. I want to see the new fabs built ASAP, but I worry they might wind up like the new nuclear plants that are in permanent states of never being completed.
Even better, fly the line development engineers over for the price of a plane ticket and hotel rooms. Setting up a factory and running it are different jobs. Of course they will want to do as much "Pre-fab" pre-fab work as possible on their home ground. TSMC would be wise to copy exactly "Copy Exactly".
This is a large move for TSMC and will change them from a "Globally Important" company to a "Global and Globally Important" company. Even if it accounts for a small percentage of their whole operation I can see it drastically changing the corporate culture.
When my father-in-law was working at Atmel (much smaller semi-conductor firm) he had a peer in the same position as him in Korea operating the same testing equipment. The company found it efficient to have him fly to Korea and live there short term to fill in for temporary absences of his Korea counterpart (and less often vice-versa) rather than train someone up. The machine and larger line that it was apart of were so expensive (and profitable) that the labor costs to ensure it's continuous operation were almost irrelevant.
(And now the cynical take...)
TSMC's most important hires will likely be the lawyers, lobbyists, accountants and fixers that can make "subsidy farming" a bigger and better business than "crypto-mining". (and just as connected to the real world)