[I'd like to learn more] About Samsung Foundry ...

[Xebec]

Reading these forums, articles, etc. I learn a LOT about Intel and TSMC, but I haven't quite picked up as much as I'd like on Samsung. I'm curious about a few things:

- What kind of products are really driving the advanced node funding/usage at Samsung Foundry today? (3nm, 2nm)
(DRAM and NAND seem to be larger node only products. Samsung Exynos adoption seems to be shrinking vs. Apple and other Android processors)

- Besides "Not TSMC", what else makes Samsung more vulnerable to Intel foundry than TSMC?

- Is Samsung significantly behind both TSMC and Intel in packaging technology?

- Is it fair to assume that Samsung wouldn't give up on leading edge even if their 'next node' is unprofitable, thanks to being able to use resources from elsewhere in the company, and because of the Samsung-Korean government relationship?

.. Thanks!

P.S. I really liked this article: https://semiwiki.com/semiconductor-...-a-detailed-history-of-samsung-semiconductor/

 

[nghanayem]

Reading these forums, articles, etc. I learn a LOT about Intel and TSMC, but I haven't quite picked up as much as I'd like on Samsung. I'm curious about a few things:

They are certainly less well understood. It is also interesting to me that many folks lump Samsung in as a foundry like TSMC rather than as an IDM like intel. Maybe this changes with the new financials and they are all viewed as foundries?

I also think their TD approach is interesting on paper, as it is the most similar to the progressive innovation approach post I wrote about about around a 1-1.5 years ago.

- What kind of products are really driving the advanced node funding/usage at Samsung Foundry today? (3nm, 2nm)
(DRAM and NAND seem to be larger node only products. Samsung Exynos adoption seems to be shrinking vs. Apple and other Android processors)

My finger in the air guess is 80% of their total wafer capacity (which mind you is far bigger than anyone else's total wafer outs) is for memory. As for the logic side of their business they aren't like intel's IDM. Most of their capacity is on trailing nodes because besides the more premium smartphone SOCs, their IDM also makes power management ICs, flash controllers, cheap smartphones for developing markets, industrial electronics, image sensors, display drivers, and the list goes on. They have a need for things like 28nm, 14LPP, 90nm, 8LPU/RF. This is in stark contrast to someone like intel who traditionally only wanted N and N+1 process for their products. It also helps that I would bet that most of their foundry business is on those lagging nodes since they will have mature yields, variation, PDKs, and rich ecosystems. To say nothing of their only pure foundry fab being their Austin site for 14"nm" and above.

- Besides "Not TSMC", what else makes Samsung more vulnerable to Intel foundry than TSMC?

Having burned up their goodwill would be my guess. If Samsung was competitive on a service and or technology, folks would have no reason to pay intel any mind.

- Is Samsung significantly behind both TSMC and Intel in packaging technology?

Hard to say, but I say yes. With that said, they do have decent hybrid bonding tech for CMOS-image sensors and HBM. But my understanding is these are wafer on wafer technologies, and as a result you can't have known good dies before packaging. For logic MCPs TSMC is the default for the whole fabless industry (NVIDIA's HBM interposer, everyone else's HBM interposers, AMD V-Cache/MI-300). For it's part, intel is probably already the leading consumer/producer of advanced 2.5/3D packages (PVC, MTL, SPR, EMR, Altera FPGAs, Gaudi, and seemingly all future client/DCAI/NEX products).

- Is it fair to assume that Samsung wouldn't give up on leading edge even if their 'next node' is unprofitable, thanks to being able to use resources from elsewhere in the company, and because of the Samsung-Korean government relationship?

.. Thanks!

Hard to say. Given how their mostly trailing edge logic design/foundry buisness is smaller then intel's manufacturing arm, In a vacuum, I doubt the financials are great right now. But as you mention being able to utilize the equipment and floorspace of the memory side of the house drastically improves the total economics. The margin stacking with the rest of Samsung electronics and pricing pressure LSI allows them to put on BRDCOM, QCOM, MTK, etc also make the TVO of being an IDM more complex than just the raw revenue they generate.

P.S. I really liked this article: https://semiwiki.com/semiconductor-...-a-detailed-history-of-samsung-semiconductor/

Agreed it was an excellent read.