China vs. USA – 7nm debate

Making a few chips is not enough

• The fuss over Huawei’s new chip has gone all the way to the top but the real issue is not if the chip exists but whether SMIC is able to produce it in massive volumes, economically and with good yields.
• The White House is now looking into the Kirin 9000s chip to ascertain if this chip has been produced via a breach of the sanctions that the US has placed on exports of chips and chip-making equipment into China.
• As usual, the issue is a little bit more complicated than the social media and press headlines would lead us to believe.
• The limitations as they stand today are that China has been blocked from importing extreme ultra-violet equipment (EUV) semiconductor manufacturing equipment for several years.
• EUV is essential to be able to manufacture chips at 5nm and below and to do so with economic yields and volume.
• New restrictions that were placed on China in October 2022 extend those restrictions to include equipment that can manufacture chips below 20nm (deep ultraviolet (DUV)).
• However, because this is a relatively new restriction and nodes below 20nm have been available for many years, China already has large amounts of DUV equipment for these nodes.
• This means if SMIC has been able to figure out how to make 7nm chips using DUV equipment, then it is quite possible that no sanctions or restrictions have been breached.
• At a high level, this would represent progress towards becoming self-sufficient in semiconductor manufacturing, but there is a catch.
• The key to semiconductor manufacturing is not being able to make a few chips with really small lines but being able to manufacture millions of them in high volumes with high yields in order to make an economic return on investment.
• This is what separates the successful companies from the ones that go bankrupt, and in my opinion, is more important than single feats of engineering.
• Using a technique called double patterning, it is possible to manufacture a 7nm chip (or even a 5nm chip) using DUV equipment but, as Intel found out to its great cost, it doesn’t work very well.
• This means while one will get a few chips that work, a lot of the others will not, meaning that the cost to make one working chip will be far higher than is economical.
• This is what Intel tried to do several years ago and it had so many problems with it that it fell several generations behind TSMC and ceded its position as the world’s leading chipmaker.
• The difference between TSMC and Intel is not who can make the smallest chip, but who can make small chips at higher yields and lower cost.
• This is the distinction that no one is applying to SMIC and there is no evidence whatsoever that SMIC will be able to make the Kirin9000s in the kind of volume and yield that make it commercially viable.
• If Huawei is to start regaining share in the Chinese market with the Mate 60 Pro as is widely being speculated, it will need millions of these chips, and I am not convinced that SMIC will be able to make them.
• Even if it could, I am still far from convinced that a Kirin9000s chip that does not support 5G will be able to compete against the likes of Oppo, Vivo, Xiaomi etc. all of whom have access to the best that Qualcomm and MediaTek have to offer.
• At 7nm, the Kirin9000s will be bigger, more expensive, and more power-hungry with lower performance than the flagships of its rivals.
• This means that as long as Chinese consumers act rationally, they would not choose the Mate 60 Pro over those of its competitors and the device will ship in very low volumes.
• In order to get over this hurdle, the Mate 60 Pro will need to be competitive on benchmark tests and demonstrate equivalent performance and battery life compared to its competitors which I think will be difficult.
• Instead, I think the Mate 60 Pro’s main function is geopolitical and that it is being used by China as evidence that it can develop cutting-edge technology without any help from or use of US technology.
• For this use case, the economics of SMIC and Huawei are irrelevant as one only needs to make a few devices.
• Self-sufficiency is a pretty big leap of faith and while this looks to be the case on the surface, I don’t think that this claim stands up very well with further scrutiny.
• The net result is that I don’t see this as very supportive of Chinese self-sufficiency in semiconductors and I continue to think that the latest sanctions will further hinder China although they will take time to have their full effect.
• RFM Research and Alavan Independent continue to expect global technology standards to bifurcate with the Chinese going one way and everyone else going another.
• It looks likely that the USA will do everything it can to ensure that the Chinese standards are not as good as the global versions so that no 3rd party country will elect to use the Chinese version even if it is much cheaper.
• In the long-term, this is bad news for everyone as the most value is generated by large compatible networks (like 4G) rather than two smaller incompatible ones.
• This means lower long-term growth for everyone as well as lower returns on investment over the next 10 to 20 years.