Samsung slips further behind TSMC in chipmaking race

[Daniel Nenni]

South Korean giant replaces several executives following supply mishap

SEOUL/TAIPEI -- When Samsung Electronics declared that it would become the world's top chipmaking foundry, it did so with the confidence that one might expect from South Korea's largest company. However, three years later, top rival Taiwan Semiconductor Manufacturing Co. has captured a larger market share, prompting Samsung to replace several executives.

"We will be the world's No. 1 player in the nonmemory sector" by 2030, Vice Chairman Lee Jae-yong, Samsung's de facto chief, said in 2019. Since then, TSMC has added about eight points to its lead.

Samsung slips further behind TSMC in chipmaking race

South Korean giant replaces several executives following supply mishap

asia.nikkei.com

 

[Daniel Nenni]

Definitely worth a read. This is along the lines of my assessment of Samsung foundry losing customers at 5/4nm and no customers at 3nm and Intel is in position to pick up Samsung customers at 18A, absolutely.

 

[benb]

Samsung is basically a memory semiconductor company, with Finfet and now Nanosheets being a "side hustle" in Foundry only.

With DRAM, a new node might scale up slowly until the per chip cost "crosses over" and it becomes less expensive to produce the latest node. So the final stage and gate of development is economic; and there can be no doubt the development is complete, since it is obvious from financial data that the new node is better. Poor financial results from DRAM (and NAND) would mean poor Samsung financial results.

With Foundry, it matters less to Samsung, financial results are buried inside DS, and wafers are owned by customers, which means tangible financial effect only with S.LSI wafers. Now that Foundry is separate from S.LSI, if you are a Samsung executive, you can always blame S.LSI for the results (and you would be correct, mostly). But this raises an issue, how do you know things are going well or not, when every year there is new node, that starts out with low yields? As long as S.LSI is pathetic, you never know, there is no financial red flag.

So my recommendation is, fix S.LSI first. S.LSI sucking every year since 2015, when Exynos was in most S6 phones, including US is a problem. As S.LSI becomes successful, maybe even better than Apple (can you imagine a dream this big?), Foundry will prosper as the early, crucial yield learning each year is successful. And Foundry executives will know when something is wrong, because financial results will clearly show it, and Won don't lie.

 

[Paul2]

Samsung is basically a memory semiconductor company, with Finfet and now Nanosheets being a "side hustle" in Foundry only.

With DRAM, a new node might scale up slowly until the per chip cost "crosses over" and it becomes less expensive to produce the latest node. So the final stage and gate of development is economic; and there can be no doubt the development is complete, since it is obvious from financial data that the new node is better. Poor financial results from DRAM (and NAND) would mean poor Samsung financial results.

With Foundry, it matters less to Samsung, financial results are buried inside DS, and wafers are owned by customers, which means tangible financial effect only with S.LSI wafers. Now that Foundry is separate from S.LSI, if you are a Samsung executive, you can always blame S.LSI for the results (and you would be correct, mostly). But this raises an issue, how do you know things are going well or not, when every year there is new node, that starts out with low yields? As long as S.LSI is pathetic, you never know, there is no financial red flag.

So my recommendation is, fix S.LSI first. S.LSI sucking every year since 2015, when Exynos was in most S6 phones, including US is a problem. As S.LSI becomes successful, maybe even better than Apple (can you imagine a dream this big?), Foundry will prosper as the early, crucial yield learning each year is successful. And Foundry executives will know when something is wrong, because financial results will clearly show it, and Won don't lie.

1. Samsung foundry is more expensive than TSMC
2. Samsung foundry is less advanced than TSMC, but pretends it isn't

They will not get top tier clients because of the point 2, and not get anybody else because of point 1. A strategic failure, and very easy to see. It's very similar how Mr. Lee tried to beat Apple at smartphones, and now he got it — one heck of an expensive smartphone which nobody is buying, but yes he got the folding screen going.