Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/geopolitical-shifts-reshape-semiconductor-landscape-taiwans-foundry-assembly-and-test-shares-to-drop-to-43-and-47-respectively-in-2027.18920/
        )

    [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] => 2021370
            [XFI] => 1050270
        )

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

Geopolitical Shifts Reshape Semiconductor Landscape, Taiwan's Foundry, Assembly and Test Shares to Drop to 43% and 47%, Respectively, in 2027

Daniel Nenni

Admin
Staff member
By 2027, Two Leading Countries to Command 72% of Foundry and 68% of Assembly and Test Markets

SINGAPORE, 3 October 2023
– IDC’s latest report, The Impact of Geopolitics on Asia's Semiconductor Supply Chain: Trends and Strategies, unveils significant changes in the global semiconductor landscape. With the implementation of the chip acts and semiconductors policies by various countries, semiconductor manufacturers have been required to set "China + 1" or "Taiwan + 1" production plans. This overhaul has driven a new global layout for the foundry and assembly/test industry, leading to a regional development in the semiconductor industry chain.

"Geopolitical shifts are fundamentally changing the semiconductor game. While immediate impacts might be subtle, long-term strategies are focusing more on supply chain self-reliance, security, and control. The industry operation will move from global collaborations to multi-regional competitions. says Helen Chiang, Asia Pacific semiconductor research lead and Taiwan country manager.

Major industry players are making strategic moves. In terms of Foundry, TSMC, Samsung, and Intel, are spearheading advanced processes in the United States, which will gradually exert influence in the foundry field. Meanwhile, even as China grapples with the development of advanced processes, its mature processes have developed rapidly under the impetus of its domestic demand and national policies. Based on the categorization by production location, China's proportion of overall industrial areas will continue to increase, reaching 29% in 2027, an increase of 2% from 2023, and Taiwan's market share will fall from 46% in 2023 to 43% in 2027. The United States will make some gains in the advanced process part, and its share for 7nm and below is expected to reach 11% in 2027.

getfile.dyn


In terms of semiconductor assembly and test, given the influence of geopolitics, technological development, and talents, leading integrated device manufacturers (IDM) in the United States and Europe have begun to invest more in the Southeast Asia market, and OSAT companies have begun to shift their attention from China to Southeast Asia. Thus, Southeast Asia is projected to play an increasingly important role in the semiconductor assembly and test market, especially in Malaysia and Vietnam, which will be the key areas that deserve special attention in the future development of this field. Southeast Asia's share of the global semiconductor assembly and test will reach 10% in 2027, while Taiwan's share will decline to 47% in the same year from 51% in 2022.

getfile.dyn


For more information on this IDC Report, please contact Helen Chiang at hchiang@idc.com. For media inquiries, please contact Angel Wu at anwu@idc.com or Miguel Carreon at mcarreon@idc.com.

 
Do the foundry market projections for the US in 2027 include Intel's internal capacity? Or is it just the foundry side of the industry without any IDMs share?
 
Much (most) of Samsung's foundry is internal, correct. Intel will claim to be foundry internally as well. (Correct me Dan if I am wrong)

The graphs do not show major change IMO. Intel will claim foundry for internal work in 2027, TSMC will ship 80K wafers per month in 2027. Ohio will not have shipped a production wafer to an external customer by 2027 based on current plans.
 
Much (most) of Samsung's foundry is internal, correct. Intel will claim to be foundry internally as well. (Correct me Dan if I am wrong)

The graphs do not show major change IMO. Intel will claim foundry for internal work in 2027, TSMC will ship 80K wafers per month in 2027. Ohio will not have shipped a production wafer to an external customer by 2027 based on current plans.
The US share for 2027 doesn't add up even if Intel's internal capacity is counted under foundry.
 
The US share for 2027 doesn't add up even if Intel's internal capacity is counted under foundry.
good point. you might be right. So how many total Intel 7 wafers are made (WSPM) in US today? what about 7 and below in 2027? your rough guess is fine
 
Much (most) of Samsung's foundry is internal, correct. Intel will claim to be foundry internally as well. (Correct me Dan if I am wrong)

The graphs do not show major change IMO. Intel will claim foundry for internal work in 2027, TSMC will ship 80K wafers per month in 2027. Ohio will not have shipped a production wafer to an external customer by 2027 based on current plans.


80K wafers (assume it's 12-inch equivalent) per month is too low. TSMC shipped about 14 million wafers in 2022.
 
Speaking of geopolitics, with what is happening in Israel, I've heard that one of Intel's fabs is in code red.
 
I am talking about the US TSMC Fab only. what do you think it will ship in 2027

Thank you. I think it will have a good chance to exceed 80k 12-inch wafers per month in 2027. TSMC Phoenix fab is a Gigafab site under its operation model. According to TSMC, a typical TSMC Gigafab produces at least 100k of 12-inch wafers per month and has the most efficient operating cost.
 
Thank you. I think it will have a good chance to exceed 80k 12-inch wafers per month in 2027. TSMC Phoenix fab is a Gigafab site under its operation model. According to TSMC, a typical TSMC Gigafab produces at least 100k of 12-inch wafers per month and has the most efficient operating cost.
I would bet TSMC AZ fab will not exceed 80kwpm in 2027 under current circumstance: slow construction (HVM, max 20kwpm, delayed to 2025) and higher cost. It could be 40-60kwpm in 2027 as strategic investment.
 
I would bet TSMC AZ fab will not exceed 80kwpm in 2027 under current circumstance: slow construction (HVM, max 20kwpm, delayed to 2025) and higher cost. It could be 40-60kwpm in 2027 as strategic investment.

There are actually two fabs (N5/N4 and N3) under construction at TSMC Phoenix. Both of them should go online before the end of 2027. 80,000 wafers total per month from those two fabs is probably the minimum to achieve a Gigafab level of operation.
 
There are actually two fabs (N5/N4 and N3) under construction at TSMC Phoenix. Both of them should go online before the end of 2027. 80,000 wafers total per month from those two fabs is probably the minimum to achieve a Gigafab level of operation.
As far as I remembered, there are actually 6 phases(fabs) planned. If we check TSMC's history in WaferTech(tsmc's subsidiary) in Washinton State, it never expanded due to the cost and culture conflicts, although TSMC had talked about GigaFab, cluster synergy for years. TSMC AZ fab was the outcome responding to geo-political conflict, but with inherit high cost and it seems the lesson learned from WaferTech repeats again. If the subsidies and risk mitigation can not be justified in two years. I can not find the reason to pour in more money there.
 
As far as I remembered, there are actually 6 phases(fabs) planned. If we check TSMC's history in WaferTech(tsmc's subsidiary) in Washinton State, it never expanded due to the cost and culture conflicts, although TSMC had talked about GigaFab, cluster synergy for years. TSMC AZ fab was the outcome responding to geo-political conflict, but with inherit high cost and it seems the lesson learned from WaferTech repeats again. If the subsidies and risk mitigation can not be justified in two years. I can not find the reason to pour in more money there.
wafertech and TSMC AZ are not even remotely close biz models. If the AZ Fab is not successful, that would be tragic for the US and for AZ and for US Fabs. I think it will ramp as soon at TSMC actually needs more wafers (2025) and be successful. I expect the largest customer could easily be Intel in 2026.
 
There is a comment from a former CAMAS employee that is quite interesting:

 
As the new ESMC (TSMC, Bosch, IFX and NXP joint venture) will produce the first wafers in 2027-2028, and the expansion of GF (if they get funding) will also be due by the same time...I wonder how is Europe share changing beyond 2027.
 
Back
Top