I wanted to write to do a follow up on my previous newsletter. I ended up writing a lot more than I wanted and it turned into a mess. So I am going to continue here.
Today, SMIC is a reasonably successful enterprise.
SMIC on Thursday night also said its earnings for the July-September quarter surged 25% on the year to $321 million, while its revenue grew 30.7% to hit a record $1.41 billion. Gross margin also reached a record 33.1%.
To be actually profitable is a good thing. For instance, GlobalFoundries is the third largest foundry and yet they are not actually profitable. The gross margin number is quite good and indicates either that they are using their pricing power, or that they have hit scale.
The bigger concern is that the company seems to be in conflict about their future direction. Four of their board directors left, including two of its prized TSMC hires - Chiang Shang-yi and Liang Mong-song.
Chiang and Liang are legendary (and infamous) actors in the TSMC story. Liang helped pioneer the copper interconnect technology in 2003 that helped TSMC leapfrog IBM. Chiang made a super-critical decision at the 28nm node that helped TSMC leapfrog Samsung.
(Liang is also legendary for leaving TSMC to go to Samsung, sparking a massive lawsuit in Taiwan. Liang helped Samsung catch up to TSMC)
Should SMIC should go for the leading edge? Right now their leading edge is 14nm FinFet - already 5 years behind the leading edge - with another in R&D that the media calls 7nm. I do recall that Nikkei said that Liang has been a really big force within the company pushing them towards more and more advanced nodes.
Personally, I do actually agree with Liang. You want to aggressively seize the technical leading edge, and on the basis of your own natural internal R&D work. The problem is that they need EUV to go further.
What they are already doing at 7nm (more like a 8nm) I bet makes it slow, low-yielding and difficult to commercialize.
A lot of commenters have said that 7nm, 14nm, and 22nm are “enough”. Which is sort of ridiculous because nodes are not all the same and they do not stay the same. The industry is aggressively moving forward. And TSMC is a very aggressive force in the industry. SMIC’s 8nm is still sitting in the lab.
TSMC will aggressively compete away its competitors’ best process nodes until a fabless company won’t think of using any other foundry.
I recently finished a massive series of EUV and Fab related videos (which are sitting in Patreon Early Access right now) and one of the things that I learned is just how much nuance vanishes underneath the simple “nanometer” marketing term. Not just the fact that 7nm does not mean 7nm anymore, but rather how different they are from each other.
But the main idea is that between the 10 nanometer and 7 nanometer generation, current 193 nanometer ArF lithography cannot go any further. That’s it. It seems that Chiang and the SMIC faction that opposed Liang recognized this and is pushing to go the GlobalFoundries route.