Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/will-samsung-soon-lose-to-intel.22340/
        )

    [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] => 2021770
            [XFI] => 1050270
        )

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

Will Samsung soon lose to Intel?

F

Fred Chen

Moderator
Some recent speculation and rumors suggest Samsung is in real trouble, even scrapping 1.4nm: https://www.trendforce.com/news/202...-1-4nm-node-to-prioritize-2nm3nm-yield-gains/

As Samsung is eager to stem losses in its foundry business, the latest market rumor indicates that the company might plan to cancel its 1.4nm node, according to Techspot and Wccftech.

Citing remarks from tech tipster @Jukanlosreve, Wccftech suggests that the move could be made due to foundry struggles or a shift to focus on improving wafer yields first. According to the report, Samsung plans to start Exynos 2600 production in May, aiming to improve 2nm GAA yields to launch its Galaxy S26 SoC and attract more customers.

Though no official reason has been given, it may be a strategic move for Samsung to strengthen its 3nm and 2nm GAA nodes, which have yet to reach stable mass production yields, as noted by Wccftech.

According to Techspot, Samsung originally aimed to begin 1.4nm mass production by 2027, alongside specialized nodes like SF2A (automotive) and SF2Z (backside power delivery). However, the plan may now reportedly be under reconsideration.

The Techspot report further notes that Samsung Foundry’s SF3 yields remain low, delaying Exynos 2500 and prompting cuts to 5nm and 7nm production due to weak demand.

Before the latest rumors, Sedaily reported that Samsung’s foundry division planned to cut 2025 facility investments by over half due to ongoing struggles.

The budget is set at KRW 5 trillion ($3.5 billion), down sharply from KRW 10 trillion in 2024, as per Sedaily.

Beyond speculation about Samsung scrapping 1.4nm, a Korea Economic Daily report suggests the company is reviewing its chip design and foundry businesses, including the possibility of halting investments in its Pyeongtaek (Korea) and Taylor (U.S.) plants.

While this seems unlikely, it still raises the question of whether at this point, Intel Foundry is ready to take business from Samsung Foundry.
 
Aren't they Behind Intel now already ?
I don't think Samsung were able to yield a big die after 8nm Node all nodes after that had either Yield or PPA issues meanwhile Intel is yielding big dies on Intel 3 which is something in between N5 and N3 and 18A is going to enter HVM Soon
 
Per Samsung's website, 8nm was the last node before EUV FWIW:

1742301115123.png
 
The 26th Semiconductor Exhibition 2024

The logo of Samsung Electronics is seen at its booth during The 26th Semiconductor Exhibition 2024 in Seoul, South Korea, October 23, 2024. REUTERS/Kim Hong-Ji/File Photo

SEOUL, March 18 (Reuters) - Samsung Electronics is set for a tough annual general meeting on Wednesday with shareholders frustrated by its failure to ride a boom in artificial intelligence that made it one of the worst-performing tech stocks last year.

Co-CEO Han Jong-hee and the head of its chip division Jun Young-hyun will be among executives attending the meeting which is scheduled to start at 9 a.m. (0000 GMT). Other issues in focus include strategies to mitigate the impact of U.S. tariffs and what will drive new growth.

In internal meetings, Samsung has acknowledged it has lost its technological edge. This is particularly true in semiconductors where it lags SK Hynix in high bandwidth memory (HBM) chips that Nvidia and others rely on for AI graphic processing units.

"Our technological edge has been compromised across all our businesses," according to a transcript of a message from Chairman Jay Y. Lee given to an internal executive seminar that was seen by Reuters.

"It's hard to see that efforts are being made to drive big innovation or tackle new challenges. There are only efforts to maintain a status quo rather than shaking things up."

Shares in Samsung tumbled by nearly a third last year while those of SK Hynix climbed 26%. In recent years, Samsung has also lost market share to TSMC in contract chip manufacturing and to Apple and Chinese rivals in smartphones.

In January, Samsung warned of sluggish sales of its AI chips in the current quarter due to U.S. export restrictions to China, which has become its most important market. That also means it faces bigger headwinds than rivals from potential U.S. tariffs on China.

The tech giant launched a share buyback plan worth 10 trillion won ($7.2 billion) in November after its stock plunged to more than four-year lows. Its shares have gained 7% since then.

Samsung is South Korea's most valuable company, with its market capitalisation of $235 billion accounting for 16% of the total value of the country's main bourse. Nearly 40% of investors in South Korean stocks own Samsung shares, according to market data.


 
Samsung did well at 7/6 nm but their 5/4 nm did not yield and was late to the market. They have 4nm business but not with the top semiconductor companies. Samsung 3nm was a non-starter so Samsung is now pushing their 2nm offering.

Intel Foundry still had PDK issues at 4/3 so they did not get much outside business.

Due to the above, TSMC N3 won 99% of the 3nm business and that is quite a bit of momentum going into 2nm.

Bottom Line: The industry wants Intel to succeed so I believe Intel 18A will get a fair shot at design wins with the top semiconductor companies. We will know more at the Intel Foundry event next month. From what I am told there will be customer announcements etc... It will also be a coming out party for Lip-Bu so I think it will be big. The prior week is the TSMC event. Exciting times in the semiconductor industry, absolutely.
 
For my education - is this conversation relevant to Samsung logic only or does it have impacts to the Samsung memory business?

I assume there's some tech development from advanced logic nodes that can benefit memory - (DRAM, NAND, etc.), even if they're somewhat different technology tracks?
 
siliconbruh999 said:
Aren't they Behind Intel now already ?
I don't think Samsung were able to yield a big die after 8nm Node all nodes after that had either Yield or PPA issues meanwhile Intel is yielding big dies on Intel 3 which is something in between N5 and N3 and 18A is going to enter HVM Soon
Click to expand...
How are they behind Intel? I do think Intel is poised to take market from Samsung if the execute and i hear it has been years since Samsung had a manufacturable foundry tech.
 
MKWVentures said:
How are they behind Intel? I do think Intel is poised to take market from Samsung if the execute and i hear it has been years since Samsung had a manufacturable foundry tech.
Click to expand...
In terms of logic manufacturing their 5nm/4nm process lags behind I4/I3 in PPA and ofc yield have you seen any 400-500mm2+ die from Samsung 5/4nm cause Intel has them with their I3 and Intel doesn't yield such big dies without good DD.
 
siliconbruh999 said:
In terms of logic manufacturing their 5nm/4nm process lags behind I4/I3 in PPA and ofc yield have you seen any 400-500mm2+ die from Samsung 5/4nm cause Intel has them with their I3 and Intel doesn't yield such big dies without good DD.
Click to expand...
For comparison - Apple M3 Max launched to retail Oct/Nov 2023; and Emerald Rapids (Intel 3) in Dec 2023. (I was curious how long ago the other "3nm" processes produced large dice for retail purchase).

EDIT: 3am response brain struck again. Granite Rapids launched in Dec 2024 on Intel 3;, Sierra Forest also launched June 2024 on Intel 3.
 
Last edited:
Xebec said:
For comparison - Apple M3 Max launched to retail Oct/Nov 2023; and Emerald Rapids (Intel 3) in Dec 2023. (I was curious how long ago the other "3nm" processes produced large dice for retail purchase).
Click to expand...
It is Granite Rapids/Sierra Forest on on Intel 3 found the exact area no public data available for area for M3 max but from my rough measurements it is approx 500mm2 based on this image and M3 being 146mm2.
1742382371203.jpeg

1742380984443.jpeg


1742380963815.jpeg
 
Xebec said:
For comparison - Apple M3 Max launched to retail Oct/Nov 2023; and Emerald Rapids (Intel 3) in Dec 2023. (I was curious how long ago the other "3nm" processes produced large dice for retail purchase).
Click to expand...
emerald rapids is not Intel 3, it is Intel 7.

What do you think the yields are on Intel 3 today (Sierra Forest/granite Rapids)?
@siliconbruh999
 
siliconbruh999 said:
I don't know what the yields are but they are producing such large dies in volume and they cut the price of Xeons recently that all point to yield being good otherwise both of these should not be possible imo.
Click to expand...
I didnt know they cut the price on Granite rapids/Sierra Forest. thanks.

Intel 7 yields are good for sure
 
You must log in or register to reply here.
Back
Top