My notes from SMIC earnings call. Took me a long time to listen through this because I had to stop the playback multiple times to make sure I get the gist down. I think China's Semiconductor industry is lucky to have SMIC, because it is really tasked with a lot of responsibility. Many times, they stated that they have 10 platforms at each node in order to capture all the products and demands. Which makes sense given the need to have everything built locally. Anyhow, the good news is that both Jingcheng & Shenzhen have started mass production & SMIC is quickly building up volume to a breakeven point right now, especially with a lot of 28/40/55nm capacity
Lingang is building production line too and Tianjin is getting roofed now. Everything is coming together. Just needs more time. Demand for SMIC services in more advanced nodes (and even nodes that many westerners don't consider advanced like 55/65nm) are very high. This is not a problem that foreign fabs have to "deal with". SMIC really needs to ramp up quickly so that TSMC & UMC don't come in with their additional 28nm capacity and take everything
Now I need to listen to Huahong's earnings call. another long one!
I think anyone that thinks of themselves as a semi industry expert and don't listen to SMIC earnings call are just kidding themselves.
While we often talk about reaching the limit of shrinking transistors for your advanced process, many other process already got there.
For example, SMIC started at 38nm in NAND flash 9 years ago and then was able to produce 24nm NAND a few years after that. IIRC, state of the art is like 19nm? but seems like customers are not asking SMIC to go below 24nm, so they are basically at the end point there
With NOR flash, it announced that it was able to do 40nm at earnings call, It was already able 40nm embedded flash 3 years ago. When Gigadevice came out with its 40nm low power embedded flash, it had to be fabbed at TSMC
Now, all of this can be fabbed at SMIC. Given the de-risking pressures, it will be very motivated to validate SMIC for same chip. SMIC just needs to make sure its chips are auto grade.
For a lot of these chips, it seems like 28nm and 40nm process is the end point. You are not getting much performance gain beyond that.
SMIC itself mentioned BCD, LCD driver & MCUs as processes that once you get to 40/28nm, there is probably no reason to go beyond that. As such, while SMIC still has work to do in making sure its 28nm product line covers all the platforms, we are probably just a couple of quarters from that. By this point next year, I will be very annoyed if SMIC doesn't have full product line at these nodes to compete against TSMC & UMC in its own backyard. It should be adding at least 40k 12-inch wpm of 28/40nm process by then. SMIC itself even mentions that 10k wpm is not enough to break even, so Jingcheng & Shenzhen both probably need to add at least 20k wpm.
Its fully booked for 55nm and above. There is clearly demand for them to rapidly add more. And the more they add and more specialty platforms they can add, the more domestic supply chain can de-risk and capture larger market share. During this time, Chinese govt can help them by slowing down the approval process for TSMC & UMC expansion. Longer term, I have no issue with TSMC/UMC expanding here. Looking at how many chips China still imports, they need more fabs to produce in China. But in shorter term, delaying those two's expansion gives SMIC a chance to really expand and be on the right path to be world's #2 behind TSMC.