Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/chipmaker-tsmcs-ceo-says-us-investment-driven-by-customer-demand.22241/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Chipmaker TSMC's CEO says US investment driven by customer demand

hist78

Well-known member
1741273623028.png


"TAIPEI, March 6 (Reuters) - Taiwanese chipmaker TSMC (2330.TW), opens new tab is expanding investment in the United States because of large U.S. customer demand, its CEO said on Thursday, adding that its production lines there are already fully booked for this year and the next two years."

"TSMC's overseas investment plans also include projects in Japan and Germany.

While TSMC plans to build three new production lines in the U.S. over coming years, it will build 11 more new production lines in Taiwan this year, Wei told reporters."

""We will continue to build them (in Taiwan) because it is still not enough," he said."

  • - Announced $100 billion U.S. investment earlier in week
  • - TSMC says U.S. investment won't impact domestic expansion
  • - U.S. President Trump has threatened to impose tariffs on chips

 
Correction, U.S. President Trump has threatened to impose tariffs on chips but not on wafers. TSMC will build packaging in the US to avoid tariffs. :ROFLMAO:

"its CEO said on Thursday, adding that its production lines there are already fully booked for this year and the next two years.""

It's interesting, but not surprising, that TSMC's Phoenix fabs are fully booked from 2025 to 2027. TSMC builds its capacity based on customer demand, not speculation. Simply put, this is the only way to succeed in the leading edge foundry business, which requires billions in capital investment.

In contrast, Intel's Ohio fab project, which began in 2022 and may not come online until 2030–2031, is the opposite example.
 
"its CEO said on Thursday, adding that its production lines there are already fully booked for this year and the next two years.""

It's interesting, but not surprising, that TSMC's Phoenix fabs are fully booked from 2025 to 2027. TSMC builds its capacity based on customer demand, not speculation. Simply put, this is the only way to succeed in the leading edge foundry business, which requires billions in capital investment.

In contrast, Intel's Ohio fab project, which began in 2022 and may not come online until 2030–2031, is the opposite example.
This is the biggest reason (but far from the only reason) why I'm mystified by Craig Barrett's call for re-hiring Gelsinger as CEO. I always thought of Barrett as one of the smartest people I've ever met, but not seeing the folly of the fab overbuilding strategy makes me wonder.
 
I’m wondering if the '26 and '27 bookings mentioned here are cancellable or require prepayment.
CC Wei stated in Taiwan’s presidential press conference that they are fully booked.
The Taiwan president's motivation may be to counter concerns raised by pro-China media and opposition parties,
which claim that the Taiwanese government yielded to U.S. pressure and compromised the country’s interests.
 
While TSMC plans to build three new production lines in the U.S. over coming years, it will build 11 more new production lines in Taiwan this year, Wei told reporters."


11 fabs in 1 year! 1 per month, with February for Lunar New Year.

I'll make some predictions:
These will be the most sophisticated fabs in the world, built at the lowest cost, with the best yields.

No business has built this many fabs in such a short time before.
 
11 fabs in 1 year! 1 per month, with February for Lunar New Year.

I'll make some predictions:
These will be the most sophisticated fabs in the world, built at the lowest cost, with the best yields.

No business has built this many fabs in such a short time before.
Probably 9 fabs and 2 advanced packaging plants and could mean projects in progress, need not to be finished this year.
There are rumors that they're cutting TW fab plans and move to the US, which could be the reason why he said that.
 
3.66 TW to 1 US fab ratio is what he's saying I think, as a gentle response to Chinese (BS) and Taiwanese (not-exactly-BS) concerns.

When you build this many fabs and make improvements with each fab, as TSMC does, it creates a wide moat with competitors. It's secret sauce, but not that secret--Third party vendors can be read into their facility spec, after a vetting process, and I was 12 years ago. There were things in that spec which exceeded Intel at the time, although Intel-TSMC were parity in broad strokes, and exceeded Samsung by a decade at least. So when I extrapolate from then until now, TSMC progress beyond Intel and Samsung is likely to be decisive.

One advantage TSMC has over competitors is chemically self-cleaning facilities which removes the sources of defects, providing a strong response when defects increase, and yields fall, as they constantly do. This extends beyond typical chamber cleans, far upstream. And some cleans are completely novel and unknown to Samsung or Intel. Or at least, they were.
 
I think every company has something unique that the competitor doesn't have be it TSMC/Intel/Samsung
In Intel's case, its unique something is incompetence. For years, Intel was the king. How the fck did they fck it up so badly? It has become a shit-tier company with a market cap of $90.68B (as of 03/07/2025). Hock Tan should buy out Intel, tear it apart, and rebuild it.
 
One advantage TSMC has over competitors is chemically self-cleaning facilities which removes the sources of defects, providing a strong response when defects increase, and yields fall, as they constantly do. This extends beyond typical chamber cleans, far upstream. And some cleans are completely novel and unknown to Samsung or Intel. Or at least, they were.
Are you talking about self cleans of the clean room or in-situ chamber self cleans (the terminology we are used to might be different)? In my personal experience the former seems like it would be of little value given how infrequently defects I have seen came from outside the chamber or Mom entity. And I thought the latter was industry standard. On older equipment sure maybe not super common. But the newer stuff has many built-in features to enhance in-situ cleans.
 
In Intel's case, its unique something is incompetence. For years, Intel was the king. How the fck did they fck it up so badly? It has become a shit-tier company with a market cap of $90.68B (as of 03/07/2025). Hock Tan should buy out Intel, tear it apart, and rebuild it.
It's not incompetence just bean counting and complacency they didn't stop their R&D but the cultural rot was bad and btw you need to have talent to destroy the moat of TSMC+Nvidia combined.
 
Back
Top