USA vs. China – All about Economics

New node is less economic than the old one.

• Analysis of the Huawei Kirin 9030 that powers the Mate 80 Pro Max, reveals a new iteration of SMIC’s 7nm process that takes it closer to 5nm but with even lower yields, making it even more expensive and uneconomic to produce.
• The Mate 80 Pro Max was launched a few weeks ago, and as is now customary, analysts have taken it to pieces and had a look at the components.
• Here, TechInsights has identified that the SoC (Kirin 9030) has been made using a slightly more advanced process than the 7nm N+2 that has been around for some time.
• SMIC and Huawei have long claimed that they would be able to get to 5nm, and while this new 7nm N+3 gets them some of the way there, the 9030 can not be described as a 5nm chip.
• The Kirin 9030 is described as a “scaled extension” of the 7nm process, which means it is the same process using the same equipment and has slightly tighter spacing between the transistors.
• This means that the chip will be a bit smaller than ones made on n+2 and hopefully will consume less power.
• However, the problem is that Huawei and SMIC are pushing the production technology to which they have access to the absolute limit, meaning smaller and smaller improvements for ever-increasing amounts of effort.
• The other problem is that because the process is more intricate and sits closer to the limit of physical possibility, the yield will be even lower, meaning that the chips are more expensive to produce.
• This is nothing like the leap that was taken when TSMC went to 5nm from 7nm, and to be fair to the tech press, no one is painting it as such a leap, which is a welcome surprise.
• SMIC and Huawei are known to have executed ingenious and notable innovations in the multipatterning technique, and are probably now better at it than TSMC ever was, but they remain hampered by the laws of physics.
• Without EUV, they will be unable to execute the jump to 5nm that triggered the leap in both performance and battery life first experienced by the Apple M series of chips and now widely available across the industry.
• Furthermore, with the ever more intricate process, the yields are going to be lower when making these chips much more expensive to produce, even when taking into account the $400m cost per EUV machine.
• This means that while this process will allow Chinese companies to use more advanced silicon in their products without fear of international interference, competing overseas will remain a big problem.
• This is because a product using 7nm n+3 will be larger, more expensive and more power hungry than the same product made using 3nm or 2nm from TSMC et al.
• This means no rational buyer will chose a Chinese chip made on 7nm n+3 when a 2nm or 3nm is on offer at a similar or lower price.
• I also think that there is a fundamental limit to the amount that the Chinese state can spend on subsidising uncompetitive products, given its demographic and debt problems.
• This is why I don’t worry too much about market share loss in China for international companies, because what they lose at home, they are more than likely to make up overseas, where China will lose out.
• This will be crucial to the ideological struggle that is being played out between China and the West, and at the moment, I continue to think it is the West that has the advantage.