Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/samsung-tsm-and-intel-the-odd-man-out.16371/
        )

    [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, TSM and Intel the odd man out

Arthur Hanson

Well-known member
With the economy getting hurt by bad management, war and covid will TSM and Samsung with their massive financial and technological strength render Intel the odd man out? Will semis become a duopoly headed by Samsung and TSM? Any aid should be an investment and not a grant for all players. Any thoughts or comments appreciated. Does anyone see any chance of Intel returning to the days of "Intel Inside"?
 
Samsung have lots of challenges too.
DRAM: Micron gradually leading over Samsung.
Phone: Chinese phone players catching up.
LCD/OLED: Give up LCD. OLED challenged by Chinese players and also mini LED.
Foundry: Latest nodes no major customers. Exynos SoC falters. At least Intel new nodes has orders from itself.

Also DRAM & Phone business cash inflow are the foundation of Samsung Foundry CAPEX.
 
Last edited:
If you asked this 15 years ago, Intel going down would also mean Samsung and TSMC lose a lot of revenue..

Today though there are a lot of other entire 'advanced node' markets that participate without Intel at all.. Mobile/Phone/iPad market for starters..

That said - it's extremely hard to predict how long this market volatility/looming pendulum of supply/demand swings will last (Vanguard in 2019, 2020, and 2021 predicted we'd have an entire decade of below standard returns for the US stock market -- just 3-5%, but fairly consistently through the 2030s), but I agree with VCT.. Intel appears to have enough internal products on the horizon of it's own to justify several more nodes.. And if they ended up with an insane level of overcapacity they could (eventually) pull some orders back from TSMC to some of those nodes, i.e. chipsets, I/O dies, GPUs..

On the aid vs. grant - that's a rabbit hole, but all three receive both from their hosting countries already.. the US, Korea, and Taiwan all supply some level of aid to these companies, whether it be education programs for the future employees, discounts on housing, tax incentives, etc..
 
Intel can only grow its revenue by 1% from 2020 to 2021 and shrunk its revenue in Q1 2022 despite almost all major semiconductor companies in the world were getting double-digit growth for the same period of time. Those companies are in all kinds of market segments utilizing advance, mature, specialty, or very old manufacturing technologies.

It proves currently Intel's most urgent problem is not on how advance the manufacturing technologies Intel is using. The problem hurting Intel greatly is Intel doesn't offer the product that market needs.
 
I remembered CEO of ASML said the advanced node is too expensive for 3 players. There will be a odd man out, but all 3 being his dear customer he wouldn't predict who will be the man out.

Safe to assume it won't be TSMC :).

Samsung vs. Intel; Intel moves primarily logic on the most advanced nodes.. while Samsung has a lot of DRAM and other tech. They also don't really have any GPUs on the cutting edge nodes at this point ("8nm" for current Nvidia; and it looks like Nvidia went TSMC for N5). The console APU market is basically TSMC also. Server and workstations split between TSMC and Intel. TSMC should have more than half of "Mobility" silicon since they have Apple exclusively + at least some Android..

I don't know for sure but at least right now I'd bet Intel is pushing more "revenue" through the latest nodes than Samsung. OTOH Samsung can subsidize their latest nodes with the rest of the (much) larger company.

THE ASML CEO still may have just made a generic comment - eventually the equipment / nodes get so expensive the # of players continues to reduce like they have been for a while..
 
You think it will ever get down to 1? (presumably TSMC)

(cue Highlander meme here)
I think five is more likely than one.

1. TSMC
2. Samsung is unlikely to just give up like GF. When Samsung wants to be a leader, this is what they're capable of:


3. Intel already has EUV machines, and IMO will figure out how to be one of the leaders again. And, even though I don't like the CHIPs Act, I still believe Congress will figure out something to subsidize fabs, eventually. Look at the US track record with subsidizing agriculture; the total annual spending is somewhere between $15B-$30B, depending on how and what you count. And while I'm not a Gelsinger fan, you've got to admit the man is driven to succeed.

4. Xi's confrontational strategy with basically the entire western world is costing the Chinese people a lot of economic success. And I suspect Xi's well-reported micro-management style may be unpopular with his own bureaucracy. So a Chinese leader could emerge who has different view of success than Xi, and plays nice with western countries, especially with the US, and then SMIC can get licenses to advanced semi equipment like ASML's.

5. I'm thinking it's too early to count Europe out. Especially if Trump gets elected to a second term, who didn't make any friends there with his policies, but I don't think his election is a necessary catalyst. Europe strikes me as tired of playing second fiddle to China and the US. Throw enough money and smart people at a manufacturing problem and it's solvable. Europe has a lot of smart people, and money can be printed (so to speak).
 
Last edited:
I think five is more likely than one.
Peter Wennink, ASML CEO, made that statement BEFORE Intel announced their 'IDM 2.0' strategy although I personally doubt the market can support five players. The reason is that only a few large companies (like Apple, Qualcomm, Nvidia and AMD) will be able to afford

1. TSMC - Although they've stated they don't rely on one company, an anchor customer (Apple) leaving for Intel would be disastrous. Unlike Intel or Samsung, they don't have another business to fall back on. They've got to be laser focused on maintain their leadership in yield, power efficiency, performance, cost. Tonight's earnings call will be very interesting in regards to TSMC's view on the current market and their competitors.

2. Samsung - It's reported that their 3nm GAA technology has poor yields and a Chinese customer (maybe?). It will be interesting to see their next 3nm iteration next year. I suspect that Samsung will be facing challenges:
a. (short-term) DRAM/NAND prices are forecast to decline this year due to less demand.
b. Foundry just lost their two biggest customers to TSMC: Qualcomm and Nvidia
c. Micron/SK Hynix providing fierce competition and Samsung's leadership in NAND/DRAM has slipped.
d. (long-term) YMTC is growing fast
e. (long-term) Intel Foundry is entering the market

3. Intel - There's been very positive news regarding the Intel 4 process node lately although a recent report stating that Meteor Lake may be delayed until late 2023 adds a bit of concern. From the outside, Pat seems to have Intel firing at all cylinders and the news that they've hired former TSMC executives (Suk Lee and Michael Chang) adds to that narrative. Other than Qualcomm, it will be fascinating to see who else will line up for Intel's foundry services.

4. SMIC - Given the US government's efforts to block ASML from selling DUV/EUV machines to China in an effort to stem their semi manufacturing abilities, I doubt SMIC will be capable of producing a leading-edge node anytime soon.

5. Europe - TSMC has stated that there aren't enough (TSMC) customers in Europe to justify building a fab although talks are still underway to bring TSMC into Europe. Intel, on the other hand, is building a fab in Germany and opening one in Ireland. Given European efforts to bring both TSMC, Intel and Samsung into their continent...would there be enough money left for a European competitor?!

NOTES:
1. Japan - Japan supposedly has a "bilateral chip technology partnership" with the US that will both countries to manufacture 2nm chips as early as 2025 according to a report from Nikkei. There has been very few details on this partnership.

2. IBM - IBM previously unveiled their 2nm chip technology in 2021. IBM has stated that they will be partnering with Intel in developing the next-generation logic and packaging technologies.
 
IBM has stated that they will be partnering with Intel in developing the next-generation logic and packaging technologies.
History doesn't repeat itself although it sure rhymes, but this time around I'd eat a dirty shoe if IBM is able to bail Intel out with technological expertise like they did in the 80s. The only way to turn around that kind of institutional inertia is with a Stalinist purge of middle/upper management.
 
Prime's list of Samsung problems is a good start, I would add:
-Stacked capacitor DRAM is out of scaling juice and no one, not just Samsung, knows what to do
-S.LSI issue; Samsung Foundry and S.LSI rise and fall together historically; lack of a strong S.LSI Exynos product
-Ongoing legal issues with defacto CEO JY Lee; question of who will be in charge
-NAND fab in Xian issues; Zero-COVID through 2027

Speculations about what may happen next:
-Convert some Foundry fabs to Memory; survival mode for Foundry
-Get the heck out of China ASAP (Sell or close Xian)

Long run:
-Demonstrate a revival of Foundry competitiveness with an advanced, 3GAA Exynos product in S phones
-Make the leap from stacked capacitors to something else
 
History doesn't repeat itself although it sure rhymes, but this time around I'd eat a dirty shoe if IBM is able to bail Intel out with technological expertise like they did in the 80s. The only way to turn around that kind of institutional inertia is with a Stalinist purge of middle/upper management.
How did IBM "bail out" Intel with fabrication expertise in the 1980s?
 
How did IBM "bail out" Intel with fabrication expertise in the 1980s?
It's the best kept secret of Intel's famous pivot to processors from memory, IBM took half of Intel's stock in return for teaching them to design and verify logic chips. Of course the absolute geniuses at IBM held on to the stock for a few years, did not see how CMOS or personal computers could overtake their lucrative BJTs and mainframes, and sold it back riiight before Intel went exponential in the 90s. The manager at one of IBM's microelectronic spinoffs told me the story like it was a hilarious goof, but it was really the beginning of the end for IBM.
 
It's the best kept secret of Intel's famous pivot to processors from memory, IBM took half of Intel's stock in return for teaching them to design and verify logic chips. Of course the absolute geniuses at IBM held on to the stock for a few years, did not see how CMOS or personal computers could overtake their lucrative BJTs and mainframes, and sold it back riiight before Intel went exponential in the 90s. The manager at one of IBM's microelectronic spinoffs told me the story like it was a hilarious goof, but it was really the beginning of the end for IBM.

IBM did own 20% of Intel stocks at one time until it sold all the remaining shares in August 1987.

 
Back
Top