Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/u-s-chip-production-is-expensive-korean-chipmakers-find.17933/
        )

    [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
)

U.S. chip production is expensive, Korean chipmakers find

Daniel Nenni

Admin
Staff member
Samsung Electonics' chip CEO Kyung Kye-hyun receives a ″Samgsung Highway″ road sign from Williamson County Judge Bill Gravell at Samsung's Taylor plant construction site on Jan. 15. Kyung announced the chipmaker's plans to complete construction within the year 2023. [SCREEN CAPTURE]

Samsung Electonics' chip CEO Kyung Kye-hyun receives a ″Samgsung Highway″ road sign from Williamson County Judge Bill Gravell at Samsung's Taylor plant construction site on Jan. 15. Kyung announced the chipmaker's plans to complete construction within the year 2023.

Korean chipmakers are facing a multitude of challenges in building and expanding capacity in the United States, after committing billions of dollars to the efforts as the U.S. government initiated a push to increase domestic manufacturing of semiconductors.

They have faced high costs of running factories and hiring workers in the United States, a reality made only worse by inflation. Competition for skilled labor and engineers is particularly fierce as others — Apple, Intel, Qualcomm and Taiwan Semiconductor Manufacturing Company (TSMC) — are hiring and often paying top dollar for talent.

Samsung Austin Semiconductor, a Samsung Electronics subsidiary based in Texas, has been recruiting new employees for 429 positions, mostly in the Fab Engineering, Infra Technology Team and Production & Systems departments.


The hires are for its plant in Austin and for the company's new Taylor plant, which is in the process of being built. Samsung Electronics is filling the spots early to have a workforce in place when the facility opens later this year.

TSMC is recruiting new employees for 61 positions at its newly completed Arizona plant, scheduled to begin production in 2024, under the condition of receiving a job training in Taiwan for up to 12 months. Academic requirements vary from high school diplomas to doctorate degrees.

At a semiconductor industry event in Washington on April 18, U.S. Secretary of Commerce Gina Raimondo said the United States has to get more serious about semiconductor manufacturing and develop the workforce needed. Raimondo forecast a shortage of about 100,000 technicians over the next few years and said the chip industry has to do something different to succeed.

Semiconductor Equipment and Materials International President Ajit Manocha went on further during the same event to say that the frequently-used shortfall estimate of 300,000 workers by 2030 understates the problem in America.

Manocha said the United States will require 500,000 or 600,00 more people in the industry to achieve success.

Taiwan addressed the talent shortage in a white paper on the Taiwanese semiconductor industry released by the Taiwan Semiconductor Industry Association (TSIA) in March. The TSIA pointed out that the number of science, technology, engineering and math (STEM) students is decreasing year after year due to the low birth rate.

In Japan, the Asahi Shimbun said Kyushu will suffer a shortage of 1,000 workers each year if the island loses its workforce to TSMC’s new plant in Kumamoto Prefecture.

Countries are lowering the bar to attract foreign workers to meet the semiconductor industry’s demand for labor.

Wang Mei-hua, Taiwan's minister of economic affairs, earlier this month said companies in Taiwan can now recruit graduates directly from the world’s top 500 universities ranked by the Education Ministry. In the past, university graduates needed either a master’s degree or two years of work experience to work in Taiwan.

Japan revised its immigration laws in February to allow graduates from the top 100 global universities to stay in Japan and apply for jobs to work for high-tech companies.

Intel and Global Foundries, along with other U.S. chipmakers, sent a letter to the U.S. Congress last August and requested the government to “exempt eligible immigrants with doctorate and master’s degree in STEM fields from annual green card country caps.”

Korea began issuing the E-7-S Visa in January to host people working for advanced technology companies and started giving graduates from universities ranked top 500 by QS World University Rankings bonus points for visa approval.

It is difficult for semiconductor companies to find the right people in Korea because the most academically successful tend to go into medicine, said Kim Jung-ho, a professor of electrical engineering at KAIST, adding that the government needs to spend three to 10 times more money for the talent hunt.

Increasing costs are a major bump in the road for chipmakers.

Samsung Electronics initially expected to spend $17 billion on the construction of the Taylor, Texas plant, but industry sources expect costs to snowball by $8 billion to $10 billion.

New plants of other chipmakers are facing the same cost-related issues.

The JoongAng Ilbo requested an evaluation of 20 plants by 20 experts, and half the facilities were encountering problems.

Ten out of 20 plants — built by Samsung Electronics, SK hynix, TSMC, Intel, Rapidus (Japanese chip consortium) and Chinese chipmakers — were rated “worrisome” in more than three categories. The Texas chip plant of Samsung Electronics had more than three hurdles to overcome, including high production costs and a lack of labor.

The nine categories evaluated were: production cost; geopolitical and environmental risk; semiconductor equipment and facility delivery; future demand; subsidies and tax credits; water and electricity supply; governmental risks, such as excess profit sharing; and labor supply.

The current semiconductor industry is “changing from a structure of an efficient division of labor to an inefficient, bloc structure,” said Kim Yong-suk, a professor of electronic engineering at Sungkyunkwan University.

The labor division system has "shattered," according to Cho Joong-hwee, a professor of embedded system engineering at Incheon National University.

TSMC and Intel share the cost and labor concerns of Samsung Electronics. For Samsung and TSMC plants, risks were assessed as high in the “governmental risks” category relayed to the excess profit-sharing provision of the CHIPS and Science Act. Such “guardrails” pose bigger threats to foreign companies than to local companies, according to some analysts.

Semiconductor production plants are expected to break apart from the cluster in Korea and Taiwan and be dispersed throughout the world — in the United States, Europe and China — within a couple of years.

“Companies build factories on foreign soil despite the high construction cost because the market is there,” said Park Jea-gun, president of the Korean Society of Semiconductors and Display Technology (KSDT) and a professor of electronics engineering at Hanyang University.

“They need to maximize their tax credits to avoid losses,” Park added.

The U.S. semiconductor act requires companies to disclose corporate secrets and return excess profits, and this is on top of high construction costs that are higher than government subsidies being received, said a semiconductor industry source who requested anonymity.

“The pressure on companies will decrease if the government pulls off a deal to ease these conditions,” the source said.

Some experts say the new paradigm for the chip industry can be an opportunity for “K-semiconductors.”

“The semiconductor value chain was a production structure maximized in economy and efficiency, but the feature dissolved when countries began pursuing domestic production,” said Kim Yang-paeng, a senior researcher at Korea Institute for Industrial Economics and Trade. “Semiconductor price increases in the process will benefit Korean companies.”

KSDT’s Park also said the Korean semiconductor market’s global revenue will increase by 80 percent after a decade since efforts by Samsung Electronics to capture the application processor market by building plants in the United States will eventually lead to the growth of Korea’s memory chip market.

 
I wonder if for training for the Taylor plant if workers will get sent to the Austin fab, or to their 3GAE/GAP fab(s)? Maybe they go to Austin for basic training and get shipped out to SK when it is time to transfer 3GAP to Taylor?
 
Y-A-W-N.

I can't believe TSMC and Samsung are both willing to say publicly that their construction costs for the US plants were so much higher than planned. Apparently multiples higher. This doesn't reflect well on the management acumen and rigor of either company.

600,000 more trained people in the US? Really? The Semiconductor Industry Association (SIA) estimates total semi employment in the US at 277,000 jobs. But that includes design and related activities. Total Intel US employment in OR and AZ combined is about 35,000 people, and a bunch of those are not working directly chip design or production.

Something about the numbers in that article smell funny.

Intel's been complaining for decades about the H1B and immigration programs, and I agree with that. Some of the smartest people in the world want to live here and we tell them to wait in a long line.
 
Last edited:
I think they are wasting their time on the wrong species. As I see it, and most people I know do, people are getting dumber. Raccoons on the other hand, through natural selection are getting smarter.

They also don't like money as much as people, but do love marshmallows far more. With proper training, and plenty of marshmallows, you could have a very motivated and low cost labor force.

Until they form unions. I mean, you've seen them gathering in groups for just about anything. What's after marshmallows? Peanut butter? S'mores?

There would be some problems, but I think far fewer than humans, who actually complain they are getting paid peanuts. Raccoons, on the other hand, place the proper value on peanut products, as noted above.
 
I can't believe TSMC and Samsung are both willing to say publicly that their construction planning for the US plants were so much higher than planned. Apparently multiples higher. This doesn't reflect well on the management acumen and rigor of either company.
I think TSMC can somewhat be forgiven for this given their relative inexperience with global manufacturing and US operations. Samsung less so, however my gut says that is mostly related to how fast this shell has gone up compared to say TSMC/intel AZ.

On a side note I would love to see a racoon in a bunnysuit and having little foot covers on their paws
 
I think TSMC can somewhat be forgiven for this given their relative inexperience with global manufacturing and US operations. Samsung less so, however my gut says that is mostly related to how fast this shell has gone up compared to say TSMC/intel AZ.
I'm not buying it. You can hire experts in the US construction industry, and anyone can read salary surveys. And TSMC has built a US facility in Camas, WA (WaferTech), and it didn't go well. Apparently they learned nothing from that experience.

 
I'm not buying it. You can hire experts in the US construction industry, and anyone can read salary surveys.
A fair counter point

And TSMC has built a US facility in Camas, WA (WaferTech), and it didn't go well. Apparently they learned nothing from that experience.

That is an OLD 8in fab. Fab design and presumably regulations/the US labor market have changed since the 90s making lessons from wafertech harder to draw direct conclusions from. It is also my understanding it is run as an independent subsidiary. In my opinion this was/is a mistake as it buries any lessons that could have been learned. I could be wrong on this point, but because TSMC has basically forgotten this venture and pretends it never happened, I would assume it is pretty hard to take lessons from it.

TLDR because of TSMC's hand's off nature and the age of the fab I think the wafertech experiment would be less helpful than the Samsung Austin. Regardless in general I agree with the thesis that the numbers don't add up, just that maybe TSMC has better excuses if incompetence is the culprit behind these numbers.
 
That is an OLD 8in fab. Fab design and presumably regulations/the US labor market have changed since the 90s making lessons from wafertech harder to draw direct conclusions from. It is also my understanding it is run as an independent subsidiary. In my opinion this was/is a mistake as it buries any lessons that could have been learned. I could be wrong on this point, but because TSMC has basically forgotten this venture and pretends it never happened, I would assume it is pretty hard to take lessons from it.

TLDR because of TSMC's hand's off nature and the age of the fab I think the wafertech experiment would be less helpful than the Samsung Austin. Regardless in general I agree with the thesis that the numbers don't add up, just that maybe TSMC has better excuses if incompetence is the culprit behind these numbers.
I'm pretty sure all international operations of all companies are operated as subsidiaries. For example, Intel Israel has co-CEOs and operates as a subsidiary. Intel Ireland is also a subsidiary. This is all legal mumbo-jumbo for tax, liability, and legal reasons I don't claim to understand the details of, and I'm sure would put me to sleep if I tried. If you do a search for any major company and their subsidiaries you'll get a list for each.

WaferTech is still an active fab (called Fab 11 within TSMC) apparently operating in the 16-35nm market, and TSMC claims it's one of six manufacturing sites located worldwide. So this whole TSMC story about "surprises", "inexperience", and "unforeseen expenses" sounds an awful lot like a "my dog ate my homework" story.

Perhaps my expectations for TSMC are too high just based on the excellence of their Taiwan operations, and I don't give Intel enough credit for its ability to build and operate international manufacturing sites.
 
Last edited:
Is 'Korea JoongAng Daily' a reliable source?

Because this:
> Samsung Electronics initially expected to spend $17 billion on the construction of the Taylor, Texas plant, but industry sources expect costs to snowball by $8 billion to $10 billion.

would be an incredible cost overrun if true.
 
Is 'Korea JoongAng Daily' a reliable source?

Because this:
> Samsung Electronics initially expected to spend $17 billion on the construction of the Taylor, Texas plant, but industry sources expect costs to snowball by $8 billion to $10 billion.

would be an incredible cost overrun if true.
"Industry sources" are always IMO an indicator of useless speculation to get clicks or promote an agenda. I know Samsung Electronics culture pretty well, and an overrun of that magnitude would be very embarrassing to the executives involved, and they would probably lose their positions in disgrace. There would be forced retirements for the most senior ones. I think they're smarter than this.
 
Back
Top