Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/samsung-should-acquire-globalfoundries.15553/
        )

    [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
)

Samsung Should Acquire GlobalFoundries!

Daniel Nenni

Admin
Staff member
Interesting article and graphics from DigiTimes. Both Intel and Samsung are IDM Foundries which is very different than the pure-play foundry players TSMC, UMC, Globalfoundries, and SMIC. This of course changed when Intel bought pure-play foundry Tower Semiconductor. Now Intel has a pure-play store front for their IDM foundry business. Brilliant move in my opinion. Not only does Intel get the appearance of a pure-play foundry, Intel also gets a wealth of pure-play experience from the Tower management and rank and file. It will be interesting to see how Intel retains the Tower Semiconductor talent. My hope is that the Tower management will assume control of IFS. If not there could be a mass exodus and it will be $5.6B not well spent, my opinion.

TSMC Samsung Intel CAPEX 2022.jpg


The CAPEX for 2022 is telling. If you exclude the memory CAPEX from Samsung, which in my estimation is 40%, TSMC is outpacing both Intel and Samsung for logic manufacturing capacity by a large margin. I also think TSMC is winning the matching CAPEX race with partnerships in Japan, Arizona, and others in process. TSMC also gets customer prepays for new fabs. So it is a race but TSMC clearly has the advantage.

The next big move would be Samsung acquiring Globalfoundries. That would certainly shake the foundry business up. Globalfoundries and Samsung partnered for 14nm so they have common footing there and Samsung definitely has the money to spend. Samsung has clearly stated on many occasions that they want to be the #1 foundry in the world. That is a very challenging task against just TSMC. Now with Intel back in the foundry game Samsung is really going to have to get disruptive and invest heavily.

Acquiring Globalfoundries would be a definite shortcut and would get Samsung's market share into the double digits (Samsung 8.7% + GF 5.7%). TSMC sits at 60%, UMC is 7.9%, SMIC is 5.7%, and Intel/Tower 1.6%.

Currently TSMC is #1 at $568.2B, Samsung is $82.9B, UMC is $75.5B, GF is $65.5B, and SMIC is $54.4B. I see zero chance of Samsung or Intel coming even close to TSMC in the next 10 years unless a serious disruption occurs.

TSMC and Intel expanding investments in chip manufacturing​

TSMC recently announced that its 2022 capital expenditure would reach US$44 billion, increasing 40% from 2021, and surpassing Samsung's US$3.8 billion. The news stunned the South Korea semiconductor industry. The "money war" has put Samsung in a tough situation.

Although Samsung did not disclose specific numbers regarding the capital expenditure for its foundry business, its capital expenditure for memory, foundry, and system ICs was about KRW 43.6 trillion (US$35.6 billion) in 2021. Sources at the South Korea semiconductor supply chain said TSMC's expanded investments would make it more difficult for Samsung to catch up with the Taiwanese giant in the foundry market.

Meanwhile, sources familiar with Samsung said the company will try to pay attention to the market developments in 2022 and continue its semiconductor investments. They said Samsung is considering investing in factories in Pyeongtaek-si, South Korea and Taylor, Texas, US. Market observers predict that Samsung's semiconductor investment would reach US$38 billion in 2022.

Meanwhile, sources at the Korea Semiconductor Industry Association said that if TSMC is considered a food court, Samsung should strive to become a Michelin star restaurant. In other words, the sources believe that Samsung can attract more clients by optimizing legacy nodes and adding investments in advanced processes.

In contrast, 70-80% of TSMC's capital expenditure will be used toward the development of 2nm-7nm advanced processes.

As TSMC and Samsung expand their semiconductor investments, Intel has stepped up its investments in the foundry business. The US tech giant recently announced plans to develop a Silicon Heartland in Columbus, Ohio with a US$20 billion investment to set up two new foundries in the beginning and eight foundries eventually. The company is expected to spend over US$100 billion on the project over the next decade.

TSMC, Samsung, and Intel are all planning to set up new factories in the US. Samsung will enter volume production of its 3nm process in the first half of 2022, deliberately ahead of TSMC's planned 3nm mass production in the second half of the year. Although Intel has yet to begin its 7nm process, industry analysts said the three companies will likely compete to achieve 2nm process production in 2025.


TSMC Samsung Intel 2nm Roadmap.jpg
 
Your insight on Intel acquiring Tower must use that to infuse a customer-facing foundry mindset is spot on.

Now, take that same insight to the case of Samsung and GFS. What are the odds of Samsung having the ability to evolve the qualities that GFS has, and to accept and nurture their best managers? Does Samsung have a track record of changing culture and outlook due to acquisitions?
 
Very interesting post Daniel, I completely agree that none of the fabs seem poised to catch up to TSMC for a while unless some unforeseen disruption occurs. Samsung in some ways reminds me of Intel during their 14nm++++ era: sort of goofy on the execution. While Intel at the time seemed more insincere each year that 10nm was further delayed, Samsung continues to skirt along the cutting edge while never reliably achieving anywhere near the performance & volume of TSMC's nodes. Given how lacking their leading nodes are compared to TSMC both in terms of performance/efficiency metrics and volume, Samsung needs to invest heavily if they're sincerely striving to get anywhere near TSMC's level of flawless, elegant execution. Given this, I can't help but feel they're lacking in fundamental areas of management and both, near and long-term execution at their fabs and they would do well to invest heavily in bettering those aspects of their fab business. Given that GlobalFoundries is arguably worse in these regards, would acquiring them really help in these crucial and fundamental areas? Needless to say I trust your judgement on this matter and would therefore be eager to hear more on how you think acquiring GlobalFoundries will help Samsung catch up to TSMC beyond just the pure numbers. Many thanks & cheers Daniel!
 
Samsung has made many foundry management changes over the years but it seems like rearranging the furniture versus doing a complete remodel. The same can be said for Intel on their other foundry attempts. Pat Gelsinger is betting his legacy on IDM 2.0 and acquiring Tower is a very big stake in the ground of the foundry business so things have definitely changed. Samsung should take this seriously. If I was the new Vice President of Samsung Foundry I would do something disruptive or he will be rearranged in a year or three. The previous Sr VP of Samsung Foundry was rearranged last year and is now a VP for Intel Foundry.
 
Interesting article and graphics from DigiTimes. Both Intel and Samsung are IDM Foundries which is very different than the pure-play foundry players TSMC, UMC, Globalfoundries, and SMIC. This of course changed when Intel bought pure-play foundry Tower Semiconductor. Now Intel has a pure-play store front for their IDM foundry business. Brilliant move in my opinion. Not only does Intel get the appearance of a pure-play foundry, Intel also gets a wealth of pure-play experience from the Tower management and rank and file. It will be interesting to see how Intel retains the Tower Semiconductor talent. My hope is that the Tower management will assume control of IFS. If not there could be a mass exodus and it will be $5.6B not well spent, my opinion.

View attachment 657

The CAPEX for 2022 is telling. If you exclude the memory CAPEX from Samsung, which in my estimation is 40%, TSMC is outpacing both Intel and Samsung for logic manufacturing capacity by a large margin. I also think TSMC is winning the matching CAPEX race with partnerships in Japan, Arizona, and others in process. TSMC also gets customer prepays for new fabs. So it is a race but TSMC clearly has the advantage.

The next big move would be Samsung acquiring Globalfoundries. That would certainly shake the foundry business up. Globalfoundries and Samsung partnered for 14nm so they have common footing there and Samsung definitely has the money to spend. Samsung has clearly stated on many occasions that they want to be the #1 foundry in the world. That is a very challenging task against just TSMC. Now with Intel back in the foundry game Samsung is really going to have to get disruptive and invest heavily.

Acquiring Globalfoundries would be a definite shortcut and would get Samsung's market share into the double digits (Samsung 8.7% + GF 5.7%). TSMC sits at 60%, UMC is 7.9%, SMIC is 5.7%, and Intel/Tower 1.6%.

Currently TSMC is #1 at $568.2B, Samsung is $82.9B, UMC is $75.5B, GF is $65.5B, and SMIC is $54.4B. I see zero chance of Samsung or Intel coming even close to TSMC in the next 10 years unless a serious disruption occurs.

TSMC and Intel expanding investments in chip manufacturing​

TSMC recently announced that its 2022 capital expenditure would reach US$44 billion, increasing 40% from 2021, and surpassing Samsung's US$3.8 billion. The news stunned the South Korea semiconductor industry. The "money war" has put Samsung in a tough situation.

Although Samsung did not disclose specific numbers regarding the capital expenditure for its foundry business, its capital expenditure for memory, foundry, and system ICs was about KRW 43.6 trillion (US$35.6 billion) in 2021. Sources at the South Korea semiconductor supply chain said TSMC's expanded investments would make it more difficult for Samsung to catch up with the Taiwanese giant in the foundry market.

Meanwhile, sources familiar with Samsung said the company will try to pay attention to the market developments in 2022 and continue its semiconductor investments. They said Samsung is considering investing in factories in Pyeongtaek-si, South Korea and Taylor, Texas, US. Market observers predict that Samsung's semiconductor investment would reach US$38 billion in 2022.

Meanwhile, sources at the Korea Semiconductor Industry Association said that if TSMC is considered a food court, Samsung should strive to become a Michelin star restaurant. In other words, the sources believe that Samsung can attract more clients by optimizing legacy nodes and adding investments in advanced processes.

In contrast, 70-80% of TSMC's capital expenditure will be used toward the development of 2nm-7nm advanced processes.

As TSMC and Samsung expand their semiconductor investments, Intel has stepped up its investments in the foundry business. The US tech giant recently announced plans to develop a Silicon Heartland in Columbus, Ohio with a US$20 billion investment to set up two new foundries in the beginning and eight foundries eventually. The company is expected to spend over US$100 billion on the project over the next decade.

TSMC, Samsung, and Intel are all planning to set up new factories in the US. Samsung will enter volume production of its 3nm process in the first half of 2022, deliberately ahead of TSMC's planned 3nm mass production in the second half of the year. Although Intel has yet to begin its 7nm process, industry analysts said the three companies will likely compete to achieve 2nm process production in 2025.


View attachment 658
Other than increasing Samsung foundry's market share, what's the benefits for Samsung to buy Globalfoundries?

There are issues:

1. Not much Samsung can learn from Globalfoundries on the leading edge nodes. Samsung knows more than GF does.

2. GF's main business is in mature nodes or specialty semiconductors. There are several strong incumbents compete in the same segment such as UMC, TI, NXP, STMicroelectronics, etc. What's the competitive advantage Samsung+Globalfoundries can achieve?

3. Samsung itself is huge already. Acquiring GF won't enhance Samsung's scale of economy too much, if any.

4. GF hasn't made profit yet. Their 2021 net income margin is -4% and adjusted net income margin is 0%. Due to the market condition and competition, I don't think GF can help Samsung to increase profit margin. If Samsung is not careful, GF can even drag Samsung's profit down.

5. At today's closing price, GF is valued at $26.44 billion. If Samsung pays 60% premium to buy GF and takes over about $7 billion GF's debt and obligation, GF acquisition will cost Samsung almost $50 billion. Is it worth it?

Will it be better for Samsung to spend $30 to $40 billion to build three leading edge fabs?
 
Samsung tends to invest heavily in downturns, and generally makes bigger decisions in crisis mode, and those conditions aren't present yet. So the lack of competitive response from Samsung is expected, for me. Samsung Foundry is profitable, and cautious, good things to be in my opinion.
TSMC remains a juggernaut, no change or surprise, yet they have to achieve dominance in HPC which is where the growth is. Samsung Foundry is pretty strong in HPC. I think that is what they mean with "Michelin" comments.
Intel seems to always do these mid-small size acquisitions that are distractions and never achieve scale within the corporate behemoth. I look at Tower Semi as more of the same. IFS is like that too, I don't expect Intel will be much of a Foundry player in 5-10 years. It takes longer than that. Samsung Foundry started in the 1990s, it takes much longer than people think to build a Foundry.
 
Personally I think IDM foundries are at a distinct disadvantage to pure-play foundries. It is all about the business model. If you are a systems company or fabless chip company and had an equal choice between an IDM foundry and a pure-play foundry, you would choose pure-play. Today it is not an equal choice. TSMC offers distinct advantages over Samsung and Intel foundry. TSMC has superior technology and a much larger ecosystem which are critical factors in choosing a chip manufacturing partner.

Intel purchasing Tower Semiconductor does two things: It is a statement that Intel is serious about the foundry business and Intel has a ready made pure-play foundry front end with an ecosystem and customers. Samsung buying GF would do the same. I realize that today it maybe cost prohibitive but I see no other way for Samsung to overtake TSMC in 2030 or even 2040 unless something disruptive happens.

How long has Samsung been in the foundry business and how many billions of dollars have they spent? By my count Apple was their first customer in 2005 so close to 20 years and hundreds of billions of dollars has been spent. Another $50B for GF does not seem like a lot in comparison and they get the pure-play front end, ecosystem, and customers.


Acquiring Tower saved Intel billions but accomplished a similar result. Great move by Pat Gelsinger.
 
"Acquiring Tower saved Intel billions but accomplished a similar result. Great move by Pat Gelsinger."

Does it make more sense for Intel to pay $50 billion to buy Globalfoundries instead of throwing $5.4 billion on Tower?

Intel 2022 cash flow is going to be negative according to Intel's own forecast. Intel must be choosy.
 
Last edited:
Does it make more sense for Intel to pay $50 billion to buy Globalfoundries instead of throwing $5.4 billion on Tower?
There's the belief that IF Intel were to attempt to acquire Globalfoundries, it would spark antitrust questions.

I'm no legal expert but I imagine getting China's approval wouldn't be a cakewalk given that Intel recently was forced to apologize for its Xinjiang supplier statement.
 
Hi Daniel - agree that Samsung or Intel highly unlikely to catch up with TSMC cap any time soon, even though there're some precedents in SEMI- think of AMD vs Intel - would one able to predict that to happen back in 2012 ;-)

you're saying that "...Currently TSMC is #1 at $568.2B, Samsung is $82.9B, UMC is $75.5B, GF is $65.5B, and SMIC is $54.4B."
Not sure of the source, but last time I've checked cap's look more like $619B, $24B, $31B and $27B for TSMC, UMC, SMIC and GR respectively, skipping Samsung foundry.
Regards
 
Sorry, that is estimated 2021 revenue.

Foundry Revenue 2021.jpg



Hi Daniel - agree that Samsung or Intel highly unlikely to catch up with TSMC cap any time soon, even though there're some precedents in SEMI- think of AMD vs Intel - would one able to predict that to happen back in 2012 ;-)

you're saying that "...Currently TSMC is #1 at $568.2B, Samsung is $82.9B, UMC is $75.5B, GF is $65.5B, and SMIC is $54.4B."
Not sure of the source, but last time I've checked cap's look more like $619B, $24B, $31B and $27B for TSMC, UMC, SMIC and GR respectively, skipping Samsung foundry.
Regards
 
"Acquiring Tower saved Intel billions but accomplished a similar result. Great move by Pat Gelsinger."

Does it make more sense for Intel to pay $50 billion to buy Globalfoundries instead of throwing $5.4 billion on Tower?

Intel 2022 cash flow is going to be negative according to Intel's own forecast. Intel must be choosy.

Intel buying Global Foundries makes no sense at all. Intel needs the money to invest in new process nodes and fabs. Global Foundries is not leading edge. If Intel wants to push forward with it's foundry plan it needs to invest in new process nodes and foundries. Not old tech especially old tech that isn't really making Global Foundries any money.

Intel if they want to commit to it also NEED, absolutely NEED to do is get rid of the share dividends and use that money to expand into greater capacity fabs in leading edge nodes if it wants any chance of eating into TSMC's business and competing.

I have to say that TSMC with 3nm tech will be huge if they can get it right. 120,000 wafers a month in 2023 for 3nm tech. I hear they are behind a bit with 3nm and are having difficulty. Not sure what the problem is but I've heard it's something that Intel sorted recently in their 10nm node. I think it was the use of Cobalt instead of copper in parts of their transistors, whereas TSMC is still using copper, and that's the reason TSMC is having difficulties in their 3nm products. Not sure if that's true or not at TSMC, but rumour mill.
 
Back
Top