Wonder what the applications of the mid tier wafers (32-90 nm) will be. MCUs, power, analog, etc. can be done with super cheap KrF tools at 130+ nm. Once you get to <90 nm you have to use dry ArF, a step up in cost. From that list I see lots of automotive chip companies like CanSEMI and CR Micro.
Somewhat advanced MCUs for cars. Like in car entertainment systems. SoCs for TVs and lower end set top boxes. But I think a lot of these applications will move to 28nm in due time. TVs are moving towards higher resolutions and the H.265 and AV1 codecs are becoming more prevalent.
The way I see it, China's semiconductor industry is surviving. But it's not enough. SMIC is meant to be China's answer to TSMC, are they really gonna to be happy with mass production of 28nm/14nm chips and extremely limited production of 7nm, even if it's all done with entirely domestic equipment?
No. But it is a matter of what you can do right now. Also, outside of mobile the current advantage of 5nm and better is kind of limited. With good chip design you can do well enough with 7nm. For example Intel has been able to remain competitive despite a two node disadvantage by having asymmetrical cores on their CPUs similar to ARM big.LITTLE.
And it seems that America purposely doesn't target the low end chips for this reason, they said it themselves, "I don't care about the chips that they use in the airbags of cars". The sanctions are meant to slow down the high end chip development and it has succeeded.
The low end chips often have much lower margins. But there are exceptions. For example Texas Instruments has consistently high margins despite making chips with comparatively obsolete technology. It depends on the market segment you are in.
By the time SMIC can get access to EUV in 2025-2030, who knows what node size TSMC and Intel are up to, or if they're even planning to shift to an entirely new architecture and/or material other than silicon to get them below or faster than 1nm.
No one knows what will happen after 1nm. This is the first time the sector has not had a clear roadmap that far ahead. So who knows.
And keep in mind that even by some miracle that the EUV is in mass production by 2025 and SMIC gets like a dozen of them, it could take 1-2 years for them to figure out how to get mass production of sub 5nm chips, Intel has EUV for years and hasn't figured it out 5nm.
Intel dabbled in EUV initially but then they gave up on it. They focused on highly complex multiple-patterning schemes with DUV that no one else uses. And they failed hard at getting this to work. They wanted to skimp out on tool expenses and it bit them. Globalfoundries is stuck at 14nm for much of the same reasons. The extremely high cost of EUV tools means they are out of reach for a lot of businesses.
And this is important years, A.I is developing so fast, a reduction in computing power now could meant a massive lack of progress in A.I development compared to America.
I think AI is overrated. I am much more concerned with the introduction of GAA transistors which are claimed to have 50% less power consumption and be radiation resistant. Learning applications can be done in the back end in large facilities. And any AI applications in the field will use static databases and algorithms with much simpler to implement hardware.
Hell it seems like foresight is a major issue with the semiconductor industry. Huawei 2018 and the EUV ban should have been a wake up call for the entire industry but most firms still went ahead with American chips and equipment.
They have no viable alternative so what do they want them to do. I do think the industry should have made a fund to work on Chinese tools but they have limited capital so the government has to do this.
Even the massive subsidies and loans that the government gave the industry went into Fabs and design firms instead of the glaring weakness of semiconductor equipment sector.
The government invested quite a lot on the tools sector so this is not true. You could argue they should have invested even more but that is besides the point.
The EDA and most recent equipment ban seemed like a massive shock to the industry but really shouldn't have been, seeing how bad the relationship between Washington and Beijing is.
You are taking this Western news release way too seriously. China has been heavily investing in EDA for a couple of years already.
14-28 nm is already extremely capable for even logic at 20+ nm as shown by IBM Power8
POWER8 uses 22nm SOI process. It is not quite the same as bulk silicon.