Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/samsung-electronics-losing-billions-of-dollars-on-chips.17517/
        )

    [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 Electronics losing billions of dollars on chips

Daniel Nenni

Admin
Staff member
Samsung Electronics lost 3 trillion won ($2.3 billion) from its memory chip business in the first two months of the year, according to semiconductor industry sources, and losses could be even bigger for the full quarter. The Suwon, Gyeonggi-based chipmaker estimates that the operating loss could balloon to 4 trillion won by the end of the first quarter as the market downturn has continued into 2023.

“Internally, there was a report projecting up to 4 trillion won in operating losses from the memory chip business in the first quarter,” one of the sources said.


It would be the first loss for the division since the fourth quarter of 2008 when the world was reeling from an economic crisis sparked by the bankruptcy of Lehman Brothers.

“Its semiconductor division has a foundry business that makes a profit, but the contract chipmaking segment is not big enough to lessen the massive losses from the memory business. At least 2 trillion won of loss for the Device Solution (DS) division is inevitable,” the source said.

The DS division has been considered the crown jewel of the electronics maker as it generated the most profit until the steep price decline of semiconductors hit the company hard at the end of last year. Other businesses include smartphones, consumer electronics and displays.

Last year, the chip division brought in 23.8 trillion won of operating profit out of the company's 43.4 trillion won. On a quarterly basis, the division’s operating profit came in at 8.5 trillion won in the first three months, 9.9 trillion won in the second and 5.12 trillion won in the third.

Things turn dramatically bleak in the fourth quarter as the division posted a 270 billion won operating profit, down 97 percent on year.

The Samsung Electronics semiconductor division reported operating losses of less than 1 trillion won in quarters in 2008 and 2001.

But this time around, the loss is more daunting as prices of dynamic random-access memory (DRAM) and NAND flash chips have declined sharply.

Average contract prices for 8 gigabyte DDR4, the benchmark DRAM, stood at $1.81 last month, a quarter of the average price four years ago.

Market tracker TrendForce projects that DRAM prices will likely slide 20 percent in the first quarter and another 11 percent in the second quarter, while those of NAND flash chips will go down 10 percent in the first quarter and 3 percent in the second. DRAM prices fell 34 percent in the second half of last year.


af43237b-e587-4c13-9a41-9765b7e4a5f1.jpg


With reduced profitability, Samsung Electronics borrowed 20 trillion won from Samsung Display, a subsidiary, to help in the funding of semiconductor investments.


“The move was inevitable given that its cash cow DS division is expected to suffer a loss for the first time in 15 years and overseas subsidiaries have most of the company's cash reserves,” said Jeff Kim, an analyst at KB Securities.


The electronics maker has said that it will maintain this year a similar level of capital expenditure as last year without any plan to "artificially" cut back production of semiconductors.

Sourcing funds locally was critical since “most of the 53 trillion-won capex planned for this year is committed to domestic operations,” according to Kim.

Even if the memory chip division incurs trillions of won of losses, the overall balance will remain in the black due to its dominance in smartphones and appliances. That is not the case for SK hynix, where over 90 percent of sales come from memory chips.

Analysts forecast an operating loss of 2.7 trillion won in the first quarter for the world’s second largest memory chipmaker, according to market tracker FnGuide. The Icheon, Gyeonggi-based producer logged an operating loss of 1.7 trillion won in the fourth quarter last year compared with a profit of 4.21 trillion won a year ago. Some experts call for the focus on high-end products as they are less affected by the steep decline in prices.

“Now is the time where chipmakers pull resources for the development of high-performing memory products,” said Park Jae-gun, an electronic engineering professor at Hanyang University.
 
Samsung is going scorched earth it seems. It’s in character for them certainly. Does anyone have any insights on the strategic implications of such a strategy?

I would have said brute force strategy but scorched earth works too. It has always been Samsung's strategy even in logic. Even if yield is bad they keep pushing the wafers through at all costs. It has been that way since I can remember. Hopefully the US recognizes how important memory is and gives Micron some CHIPs Act cash otherwise it will be difficult being caught between Samsung and Hynix during a price war.
 
Samsung has no plan to scale back production despite heavy losses
To complete yield learning curve and reach stable high yield, it takes time and wafers to learn through failure identification and continuous improvement. As I remembered few years ago, someone summarized and correlated D0 trend with cumulated wafers manufactured and there was a magic waferout number at ~100k(?) when stable yield achieved. Not sure it works or not now, but definitely for shorter yield learning curve, the volume is critical which could be a compromise between time to higher yield and cost expended(loss).
 
I would have said brute force strategy but scorched earth works too. It has always been Samsung's strategy even in logic. Even if yield is bad they keep pushing the wafers through at all costs. It has been that way since I can remember. Hopefully the US recognizes how important memory is and gives Micron some CHIPs Act cash otherwise it will be difficult being caught between Samsung and Hynix during a price war.
This was my thinking as well regarding micron. Samsung would happily burry Micron if able. I hope they the commerce department is keeping that in mind when distributing any CHIPs act funding, that is what that funding will facilitate in terms of Samsung going after American competitors.
 
Last edited:
To complete yield learning curve and reach stable high yield, it takes time and wafers to learn through failure identification and continuous improvement. As I remembered few years ago, someone summarized and correlated D0 trend with cumulated wafers manufactured and there was a magic waferout number at ~100k(?) when stable yield achieved. Not sure it works or not now, but definitely for shorter yield learning curve, the volume is critical which could be a compromise between time to higher yield and cost expended(loss).
Good point
 
To complete yield learning curve and reach stable high yield, it takes time and wafers to learn through failure identification and continuous improvement. As I remembered few years ago, someone summarized and correlated D0 trend with cumulated wafers manufactured and there was a magic waferout number at ~100k(?) when stable yield achieved. Not sure it works or not now, but definitely for shorter yield learning curve, the volume is critical which could be a compromise between time to higher yield and cost expended(loss).
Massive sampling would be important, especially for catching stochastic issues.
 
Massive sampling would be important, especially for catching stochastic issues.
Yes, there are many challenges/process marginality to be fixed or optimized in advanced nodes. For nanosheet, BPR and 3D package(chiplet), There will be more funs.
 
To complete yield learning curve and reach stable high yield, it takes time and wafers to learn through failure identification and continuous improvement. As I remembered few years ago, someone summarized and correlated D0 trend with cumulated wafers manufactured and there was a magic waferout number at ~100k(?) when stable yield achieved. Not sure it works or not now, but definitely for shorter yield learning curve, the volume is critical which could be a compromise between time to higher yield and cost expended(loss).
100k does not seem that large for DRAM. You get around 2TB yield per wafer, which is about 3 to 4 servers in the cloud. Even with some scaling back, the cloud is building millions of servers per year. And then there are the mobile markets which consume even more, and which have something like 90% commonality for shared learning in production. So, I don't think any of those vendors are at risk of volumes too small to get mature yields.
 
100k does not seem that large for DRAM. You get around 2TB yield per wafer, which is about 3 to 4 servers in the cloud. Even with some scaling back, the cloud is building millions of servers per year. And then there are the mobile markets which consume even more, and which have something like 90% commonality for shared learning in production. So, I don't think any of those vendors are at risk of volumes too small to get mature yields.
You are talking about the volume after ramp stage when the yield is stable. Then the 100k wafer out might be not enough. But for new technology ramping, volume within short time is critical for yield learning.
 
Samsung is going scorched earth it seems. It’s in character for them certainly. Does anyone have any insights on the strategic implications of such a strategy?

Unlike foundry businesses, DRAM's a commodity. They have some variations like power...etc, but you can simply replace them if the price per performance is OK. In other words, you can steal customers via price war.

Meanwhile, Samsung is enjoying a bit of cost superiority and TONS of cash($100B) in hand. Competitors don't have those(their valuts are empty). They're even cutting CAPEX in half. So, in short, Samsung is using scorched earth, burning future fabs of competitors during process.
 
Back
Top