Behind Samsung's $116 Billion Bid for Chip Supremacy (bloomberg.com) 16
Technology giants are increasingly designing their own semiconductors to optimize everything from artificial intelligence tasks to server performance and mobile battery life. Google has the Tensor Processing Unit, Apple has the A13 Bionic and Amazon.com has the Graviton2. What the titans all lack, however, is a factory to build the new chips they are dreaming up. Enter Samsung Electronics, which is planning a decade-long, $116 billion push for their business. From a report: The South Korean company is investing heavily in the next step in miniaturizing semiconductors, a process called extreme ultraviolet lithography (EUV). It's by far the priciest manufacturing upgrade Samsung has ever attempted, a risky bid to move beyond its established business of cranking out commoditized silicon and to leapfrog the incumbent leaders in the $250 billion foundry and logic-chip industry. "A new market is opening up," Yoon Jong Shik, executive vice president of Samsung's foundry business, said at a forum recently held in Seoul. "Companies like Amazon, Google and Alibaba, which lack experience in silicon design, are seeking to make chips with their own concept ideas in order to boost their services. I think this would bring a significant breakthrough for our non-memory chip business."
Samsung is a relative underdog in this growing field. The foundry business -- as the manufacturing of chips for companies like Google and Qualcomm Inc. is known -- is dominated by Taiwan Semiconductor Manufacturing Co. with more than half the market, according to TrendForce data that puts Samsung at 18%. TSMC also took over Apple's A-series processor manufacturing from Samsung, which was the original production partner. Samsung plans to spend about $10 billion per year on equipment, research and development over the next decade, but TSMC is even more ambitious with capital expenditure of around $14 billion for this year and next.
Samsung is a relative underdog in this growing field. The foundry business -- as the manufacturing of chips for companies like Google and Qualcomm Inc. is known -- is dominated by Taiwan Semiconductor Manufacturing Co. with more than half the market, according to TrendForce data that puts Samsung at 18%. TSMC also took over Apple's A-series processor manufacturing from Samsung, which was the original production partner. Samsung plans to spend about $10 billion per year on equipment, research and development over the next decade, but TSMC is even more ambitious with capital expenditure of around $14 billion for this year and next.
Re: (Score:2)
Cute, but stupid. Samsung's fabs don't spontanesouly combust like some of their phones with bad batteries.
Re: (Score:2)
It wasn't just batteries. They also made washing machines that burst into flames. And wasn't there some third product line of theirs catching fire around the same time period?
Re: (Score:2)
Huh, hadn't heard of that one. Still nothing to do with their fabs though. Interestingly enough, some of TSMC's fabs DO catch on fire:
https://www.eetimes.com/fire-r... [eetimes.com]
Well one of them did, anyway.
Itâ(TM)s a Trap! (Score:2)
Funny how this coincides with the shuttering of Samsungâ(TM)s SoC R&D facility.
I guess if you canâ(TM)t design a decent SoC yourself, then just look over the shoulders of those who can...
Re: (Score:3)
They'll just go back to bog-standard ARM-derived designs. Everyone except Apple has gone that route. Huawei and Qualcomm shipped designs based on A76 last round (Kirin 980 and Snapdragon 855, respectively). They'll move to A77 next. Samsung will probably make their next Exynos chip based on A77 as well, unless they expect the 990 to carry them through 2020.
Re: (Score:2)
They'll just go back to bog-standard ARM-derived designs. Everyone except Apple has gone that route. Huawei and Qualcomm shipped designs based on A76 last round (Kirin 980 and Snapdragon 855, respectively). They'll move to A77 next. Samsung will probably make their next Exynos chip based on A77 as well, unless they expect the 990 to carry them through 2020.
Hmmm.
If that's true (and I have no reason to believe it would be a lie), maybe that's why Apple's Ax SoCs continue to smoke the competition.
Re: Itâ(TM)s a Trap! (Score:1)
Re:Itâ(TM)s a Trap! (Score:4, Informative)
Re: (Score:2)
SoC design and lithographic fabrication are entirely different. The latter is putting blocks in place, the former is deciding how to arrange the blocks. Lots of companies design ICs and SoCs but have no fab capability (Nvidia, AMD, Apple, among others). Lots of companies can fab but have no design capability (TSMC, Global Foundaries, UMC). They just accept orders from the previous type of company and manufacture it for them. Samsung (along with Intel) is one of the few who do both.
Apple has an "Architecture-class" ARM license.
They do a lot more than "push blocks around".
They just don't want to mess-around with having to have a Fab.
Re: (Score:1)
Re: (Score:2)
Exactly the same as apple does; they copy everyone. But Samsung can make their own chips.
Wrong.
Else, why would Apple's SoCs trounce the competition (all of which also use ARM), year after year? If they were "copying everyone", wouldn't "everyone" be trouncing them, performance-wise?
Apple just chooses not to have the headache of building and maintaining a Fab line. So they subcontract that out to TSMC (mostly).
Got it?
Re: Itâ(TM)s a Trap! (Score:1)
This article is not for the technically-aware (Score:5, Insightful)
TSMC has an EUV node already in full production (7nm+). It's funny how this article has to go out of its way to tell you about EUV when it's been an industry buzzword for years now. Samsung won't be "first to market" with anything. TSMC is too far ahead. It's also funny how the writer thought that A13, Graviton2, and "Tensor Processing Unit" even belong in the same sentence. It's like the writer either knows nothing of the semiconductor industry, or assumes the reader is similarly-unaware.
Re: (Score:2)
It's like the writer either knows nothing of the semiconductor industry, or assumes the reader is similarly-unaware.
One leads to the other. See the comments in the "fox news" story for evidence that its also endemic on slashdot.