Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/tsmc-has-the-largest-market-share-in-the-global-semiconductor-manufacturing-industry.13693/
        )

    [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
)

TSMC has the largest market share in the global semiconductor manufacturing industry

Daniel Nenni

Admin
Staff member
A Counterpoints Research study has determined that TSMC has 28% of the market share in matured nodes (40nm and below), making it the world leader in the global semiconductor manufacturing market. UMC, SMIC and Samsung take second, third and fourth place with market shares of 13%, 11% and 10%, respectively.

A study conducted by market research firm Counterpoint Research sheds some light on the global semiconductor manufacturing situation. As expected, TSMC leads the market with a total share of 28%. Taiwan-based UMC ranks second with 13% market share, primarily due to its investment in legacy process nodes (40nm and above). SMIC comes in third with an 11% share, followed Samsung and Global Foundries at 10% and 7%, respectively.

TSMC's hegemony over the semiconductor market is hardly surprising, considering that it makes silicon for industry bigwigs such as Apple, Nvidia, and AMD. Intel is expected to join that party in 2022, once TSMC's 3nm process is fully functional. However, TSMC is unable to keep up with additional demand from the likes of AMD. That is why it has about US$28 billion set aside for expansion plans. A good chunk of that amount will go towards furthering its EUV (extreme-ultraviolet lithography) processes. The rest will likely be used to set up additional manufacturing plants to ramp up production.

Although China-based SMIC now has its in-house 7nm process node, its prospects look grim due to its addition to the U.S Government's entity list. If the decision isn't reversed, SMIC runs the risk of fading into obscurity. It may even have to divest some assets like Huawei to stay afloat. Samsung could occupy the market void left by SMIC, due to the US$10 billion investment in its 3nm node. AMD already plans on outsourcing some of its GPU and APU manufacturing to Samsung. Lastly, the report also prophesizes that Intel's initial 7nm yields will be less than ideal due to the lack of in-house EUV tools.

Global Semiconductor Market 2020.png

Global semiconductor market share (image via Counterpoints Research)
 
  • Like
Reactions: mef
I’ll wonder still for awhile, I think, why GF made the decision to end its 7nm program. I can think of ideas behind it - i.e. maybe they think skipping to the lower nodes gives them the edge while TSMC plods away at 7nm (once 7 is less leading edge). Or maybe they gave up that battle because TMSC was ahead of them with 7nm PDK development or something? Just seems like the decision was ill-timed in terms of market needs, but that might be a short-sighted view.
 
I’ll wonder still for awhile, I think, why GF made the decision to end its 7nm program. I can think of ideas behind it - i.e. maybe they think skipping to the lower nodes gives them the edge while TSMC plods away at 7nm (once 7 is less leading edge). Or maybe they gave up that battle because TMSC was ahead of them with 7nm PDK development or something? Just seems like the decision was ill-timed in terms of market needs, but that might be a short-sighted view.

GF did not succeed at 14nm, they ended up licensing it from Samsung. GF skipped 10nm and tried again at 7nm but did not succeed and Samsung did not license 7nm. TSMC is a fierce competitor in regards to TTM and if you are late to a node you miss out on the big customers, the yield learning and revenue that goes with it.
 
A Counterpoints Research study has determined that TSMC has 28% of the market share in matured nodes (40nm and below), making it the world leader in the global semiconductor manufacturing market. UMC, SMIC and Samsung take second, third and fourth place with market shares of 13%, 11% and 10%, respectively.

A study conducted by market research firm Counterpoint Research sheds some light on the global semiconductor manufacturing situation. As expected, TSMC leads the market with a total share of 28%. Taiwan-based UMC ranks second with 13% market share, primarily due to its investment in legacy process nodes (40nm and above). SMIC comes in third with an 11% share, followed Samsung and Global Foundries at 10% and 7%, respectively.

TSMC's hegemony over the semiconductor market is hardly surprising, considering that it makes silicon for industry bigwigs such as Apple, Nvidia, and AMD. Intel is expected to join that party in 2022, once TSMC's 3nm process is fully functional. However, TSMC is unable to keep up with additional demand from the likes of AMD. That is why it has about US$28 billion set aside for expansion plans. A good chunk of that amount will go towards furthering its EUV (extreme-ultraviolet lithography) processes. The rest will likely be used to set up additional manufacturing plants to ramp up production.

Although China-based SMIC now has its in-house 7nm process node, its prospects look grim due to its addition to the U.S Government's entity list. If the decision isn't reversed, SMIC runs the risk of fading into obscurity. It may even have to divest some assets like Huawei to stay afloat. Samsung could occupy the market void left by SMIC, due to the US$10 billion investment in its 3nm node. AMD already plans on outsourcing some of its GPU and APU manufacturing to Samsung. Lastly, the report also prophesizes that Intel's initial 7nm yields will be less than ideal due to the lack of in-house EUV tools.

View attachment 343
Global semiconductor market share (image via Counterpoints Research)
This research report is kind of strange and possibly misleading. For example if we look into 2020 revenue, TSMC is more than 7 times of UMC ($47 billion vs $6.4 billion) in terms of revenue. How can the market share chart shows TSMC (28%) vs UMC (13%), only two times?

Unless they are only talking about legacy nodes, anything older than 40nm. But in their original report, it spent significant efforts to discuss EUV, 7nm, 3nm, and Intel's outsourcing. What does that do with 40nm and older nodes? How do they measure the legacy nodes market share?
 
TSMC itself was a late comer for much longer than it was a leader.

I'd say few years old nodes, and shrinking position in mature ones are their Achilles heel.

It's ripe for the taking for any volume oriented fab ready to invest in a way which will make any sock market analyst unhappy.
 
Dear Daniel,

I see in the list that Tower Semi has 5% of the global semi manufacturing business. As a guy who lives in Israel, I wish this were true, but I find it very hard to believe that Tower is on this list and and a company like Intel is not? Can you please explain how these numbers were derived? Thanks Again! Yaakov
 
TSMC itself was a late comer for much longer than it was a leader.
I'd say few years old nodes, and shrinking position in mature ones are their Achilles heel.
It's ripe for the taking for any volume oriented fab ready to invest in a way which will make any sock market analyst unhappy.

TSMC certainly does not dominate mature nodes and I see no reason for them to invest money backwards. FinFETs are the future for power and area savings and TSMC dominates the foundry FinFET era. At some point in time the early FinFET processes will be considered mature nodes then TSMC will dominate mature nodes. It is only a matter of time.

Automotive is a good example. The old automotive companies had no reason to move up into more modern process nodes until Tesla disrupted the car industry. And I can assure you that Tesla is using FinFETs. The same can be said for Apple and the phone market. They are pushing the smart phones and tablets through the FinFET era very quickly.

Bottom line: I highly doubt very many new designs are targeting mature nodes so at some point in time the mature designs will fade into the sunset and TSMC will win even more market share, absolutely.
 
  • Like
Reactions: mef
Dear Daniel,
I see in the list that Tower Semi has 5% of the global semi manufacturing business. As a guy who lives in Israel, I wish this were true, but I find it very hard to believe that Tower is on this list and and a company like Intel is not? Can you please explain how these numbers were derived? Thanks Again! Yaakov

Remember Tower has two fabs in Israel and two fabs in the US so they are half and half. They are probably also counting the Tower/Panasonic fabs in Japan.

 
52% of tsmc at 5 nm goes to apple so there is not enough for anyone else.
Portland,

That research report is a bit misleading. Those market shares described in their charts are for 40nm process nodes or older.
 
Bottom line: I highly doubt very many new designs are targeting mature nodes so at some point in time the mature designs will fade into the sunset and TSMC will win even more market share, absolutely.
I am doing a 0.18µm TSMC MPW now and these are regularly sold out. I think you should change statement to that not many new high-volume designs are on mature nodes; and if you look at revenue it is the high volume designs that count.
 
I am doing a 0.18µm TSMC MPW now and these are regularly sold out. I think you should change statement to that not many new high-volume designs are on mature nodes; and if you look at revenue it is the high volume designs that count.

Why 0.18µm? Instead of something smaller?
 
I’ll wonder still for awhile, I think, why GF made the decision to end its 7nm program. I can think of ideas behind it - i.e. maybe they think skipping to the lower nodes gives them the edge while TSMC plods away at 7nm (once 7 is less leading edge). Or maybe they gave up that battle because TMSC was ahead of them with 7nm PDK development or something? Just seems like the decision was ill-timed in terms of market needs, but that might be a short-sighted view.
GF ended 7nm because they didn't have the $$ to keep funding the development, that's really all it was. It's very expensive and their investor was no longer willing to keep infusing cash.
 
Let's do some quick math re TSMC #1, UMC #2, SMIC #3:

TSMC total revenue in 2020: $45.51bn

TSMC foundry sales % for mature nodes (+40nm): ~28%

TSMC revenue for mature nodes: 45.51 * 0.28 = $12.74bn

- - - - - -

UMC total revenue in 2020: $6.314bn (NT$176821mn)

UMC foundry sales % for mature nodes (+40nm): 86%

UMC revenue for mature nodes: 6.314 * 0.86 = $5.43bn

- - - - - -

SMIC total revenue in 2020: $3.91bn

SMIC foundry sales % for mature nodes (+40nm): ~91%

SMIC revenue for mature nodes: 3.91 * 0.91 = $2.9bn


Of course, take that with a healthy grain of salt, since it's quick & dirty... but based on public financials SMIC had almost half the revenue for mature nodes than UMC, but in Counterpoint's analysis their market share is almost identical.
 
TSMC certainly does not dominate mature nodes and I see no reason for them to invest money backwards.
Deny it to immediate competition, and prevent future one. Delivering a knockdown punch to the entirety of 2nd tier fab sector will probably cost TSMC less than 1/10th of money they spent on 5nm.

There is not much risk for TSMC on the leading edge. Nobody can, or want to compete with them for a new node. For as long as they make big, and moneyed clients happy with each new node, they will keep scoring nearly all the premium for early node access.

On other hand, the risk of somebody closing in to compete with TSMC by repeating TSMC's path of steady growth in less juicy, but much deeper market for mainstream ICs.

At some point in time the early FinFET processes will be considered mature nodes then TSMC will dominate mature nodes. It is only a matter of time.
They wouldn't. 65nm, and down is inherently more costly.

65-55 nodes are last ones where ordinary planar CMOS, materials, simple masks, single patterning, and dry lithography works well.

Anything below needs an n-times more in cost for the five things above, without any prospect of becoming cheaper with time.

High volume ICs will stay on dry litho for a very long time, even new designs.

And as you seen above, even 200mm, and 180nm are alive and kicking, without any prospect of any sunset on the horizon. On the opposite, through the last few years, we've seen 200mm even starting to get new capacity.
 
Last edited:
Back
Top