Semiconductors and China with a little bit of India
You may have read about semiconductor shortage in the news. What it really is all about is that semiconductors are once again getting their fair dues after a long period of commoditization.
In practical terms though the shortage is due to a huge spike in demand coupled with manufacturing being concentrated in just a handful of companies. Semiconductor manufacturing is highly capital intensive and with the technology advancing rapidly requires continuous capital and R&D expenditures.
The three largest manufacturers are Intel in the US, Taiwan Semiconductor Manufacturing Corporation (TSMC) in Taiwan, and Samsung in South Korea and also in the US (in my hometown Austin). Intel manufactures semiconductors only for its own designs, TSMC does purely contract manufacturing, while Samsung does both.
This arrangement is changing as I write with Intel under its new CEO Pat Gelsinger announcing that it will ship out its own designs to TSMC for manufacturing and also build a new fab that will eventually manufacture chips for other chip design companies. TSMC also has started building a new fab in Arizona in a plan to diversify but more importantly in response to the always present physical threat of occupation of Taiwan by mainland China just 180 kms across the Taiwan Strait.
And now what about China? The largest semiconductor manufacturer in China is the partially state owned (like much else) Semiconductor Manufacturing International Corporation (SMIC). It is anybody’s guess as to what is “International” about SMIC. SMIC’s manufacturing technology is several generations behind the top three I mentioned earlier. But that is plenty good enough to make chips for defense, aerospace, automobile and run of the mill electronics industries. But not good enough to make chips for smartphones, laptops or servers.
Let us now break down the different components in the semiconductor chain. Broadly speaking you have chip design and chip manufacturing.
You can further break down chip design top down into applications (I.e smartphones or servers), architecture (X86, ARM, GPU, ASIC etc), logic design, circuit design, logic verification, circuit simulation, timing analysis, noise analysis, placement and routing, testing. The list is longer but you get the idea.
Chip manufacturing is equally complicated with silicon wafer manufacturing, lithography, etching, packaging and much more.
You can immediately sense the difference between the two, chip manufacturing is all about equipment and chemicals while chip design is all electrical engineering and computer science. The one area that is on the boundary between design and manufacturing is transistor design. You can design your own transistors at your own risk or use the tested and characterized transistor design provided by the chip manufacturer. Almost nobody designs their own transistors.
The complex equipment needed to manufacture chips are themselves manufactured by only a very small number of companies with all of them except one being US companies. The top companies in this sphere are Applied Materials and Lam Research in the US and ASML in the Netherlands. So while the perception that the US is not big in semiconductor manufacturing is true, the reality is that the US dominates the technology behind manufacturing through research in universities and within companies such as Applied Materials. The final story is that without US based semiconductor research there will be no progress in manufacturing in Taiwan or Korea and most definitely not in China. This analysis is largely true but not entirely true.
As for the design side of semiconductors, there are today perhaps thousands of entities that supposedly design chips. Semiconductors are everywhere therefore chip designers are everywhere too. So what is the deal about US domination in chip design? While there are chip designers everywhere, there are chips and there are other chips. There are chips in your toaster, microwave oven and washing machine. There is even a fairly sophisticated chip implanted at the tip of the Apple lightning cable. And then you have really complex chips and these are in your PCs, Macs, smartphones, cloud servers and game consoles. The whole story chapter and verse about these complex semiconductors is all about the US with a little bit of the UK thrown in for good measure, but mostly the US. I know this because I spent a good 30 years on all aspects of semiconductor design. My special expertise during much of my career was in making chips work at high speeds and low power.
The architecture of chips today as for CPUs is either X86 from Intel or ARM. If you are not familiar with the name ARM you should be because ARM architecture is everywhere from your phones to cars to Apple Macs. Apple recently replaced Intel CPUs in their Macs with an in-house designed ARM based CPU. ARM is a UK company they are currently owned by the Japanese conglomerate Softbank. US graphics chip company NVIDIA has entered into an agreement to buy ARM from SoftBank. But this acquisition has run into a few obstacles. So the deal is still pending. ARM’s R&D team is based here in Austin. Then for smartphones you need the CDMA architecture from Qualcomm which is again a US company. GPU architectures are also dominated by the two US companies NVIDIA and AMD (via ATI graphics). You get the idea by now. The US dominates computer architectures. There is not even any runner-up.
The rest of chip design is all software (EDA - Electronic Design Automation) from logic design to layout. Three companies dominate this area, Cadence Design, Synopsys and Mentor Graphics. All US companies. Their software is backed by 50 years of research into combinatorial algorithms, circuit design, simulation, computational algorithms, boolean and sequential logic, electromagnetic theory (Maxwell’s equations) in US universities and corporate labs notably IBM Research and former Bell Labs. So again you get the point. The US dominates all aspects of chip design methodology and software. EDA software is what has led to the simplification of chip design. That is in order to design a chip you need not know even what a transistor is or how it works. All you need to be able to do is to write code in one of the few hardware languages most notably Verilog. My own PhD thesis (1983) at Princeton was titled “Design Implementation and Theory of a VLSI Layout Language.”
Back to China. As everyone knows China is pretty good at manufacturing anything. Chinese workers are very disciplined and dedicated. These traits are also key to successful semiconductor manufacturing. But these are not the traits for innovation. To innovate you have to break rules. The Chinese system does not permit breaking rules. China is much behind in the mathematics, engineering and computer science required to advance the state of semiconductor design technologies. Their universities are not equipped for high quality research and innovation. More importantly the sauce is so much cooked that any new sauce that Chinese universities come up with will take years and years to smell or taste right. No chip design shop including any in China is going to send its design to TSMC or anyone else for manufacturing without the sign-off stamp of approval from Cadence Design or Synopsys software.
So what is happening between the US and China as regards semiconductors? The US commerce department has banned the sale of both US based semiconductor manufacturing equipment and US based semiconductor design software to certain entities in China including Huawei that have any association with China’s defense industry. Apparently this is all perfectly legal under a US law that allows such bans to protect US security interests. Furthermore the US has also banned TSMC or any other manufacturer to manufacture chips for China that uses US equipment. Which essentially means that Huawei and others are dead companies walking. They are rushing to get their inventory stockpiled with current designs as any new design will come under the US ban. The ban on licensing of US based design software to Huawei is essentially a show stopper for Huawei.
SMIC and other Chinese semiconductor manufacturers are also dead in the water. They cannot upgrade their manufacturing equipment to keep pace, albeit many many years behind, with TSMC. There are reports that SMIC is going to equipment graveyards in Japan and elsewhere to pay premium price to buy any junk they have before they are recycled (https://asia.nikkei.com/Business/Electronics/China-hoards-used-chipmaking-machines-to-resist-US-pressure).
Coming under extreme pressure, the Chinese government is throwing billions and billions at anybody who knows anything about anything. And they are getting duped. This is an interesting story by itself. Recently a supposed $20 billion plant in Wuhan (of all places) turned out to be a ghost factory. See the link … https://www.asiafinancial.com/china-gives-up-on-20bn-wuhan-chipmaking-fiasco. Cheating is endemic in China. One has to cheat in order to please the bosses all the way up the chain. Chinese academia is full of plagiarism and cheating. Chinese students coming to the US to study are also a big part of the cheating game. Here are some references …
https://www.nature.com/articles/d41586-021-00733-5
https://www.nature.com/articles/d41586-021-00219-4
https://www.lamag.com/citythinkblog/ucla-cheating/
https://www.campusreform.org/?ID=14951
The very fact that China and its people are resorting to cheating in academia and stealing intellectual property is a clear indication that China is behind the curve in science and technology.
Ok, what about India? All I can do is to repeat the age old adage … ignorance is bliss. Every effort to build any kind of semiconductor plant in India has resulted in utter failure. Now it is all too late. Every now and then some minister or bureaucrat will make some noise and that is about it. The most recent noise comes from the Government of India with an offer of a billion dollar reward for anyone who sets up a semiconductor plant in India (https://www.reuters.com/article/us-india-semiconductor-exclusive-idUSKBN2BN12J). And I am seriously thinking about this myself (ha). See also this link https://www.meity.gov.in/esdm/FAB. To be honest this call for interest in setting up a fab in India is a joke. Anyway this is not how you go about to secure investments in semiconductors. Posting a request for proposal on Twitter is a tell tale sign that nobody knows what they are talking about and it is all about public relations.
Here in the US the drumbeat is on to throw public money at semiconductor manufacturing. See https://fortune.com/2021/04/13/in-call-with-ceos-biden-doubles-down-50-billion-plan-invest-chips/. China is being used as the bogeyman. But in my opinion the key long term strategy should be about investing in long term research in universities and strengthening the already existing research ecosystem in semiconductors as well as futuristic device and computing technologies.
Some articles to read …
https://www.nytimes.com/2021/06/03/us/politics/biden-ban-chinese-firms-trump.html
https://www.washingtonpost.com/technology/2021/03/31/impact-us-campaign-against-huawei/
Much of the stuff in the articles only analyze the business and trade side of the fight between US and China. Believe me when I say that China is facing an insurmountable technology gap that will only grow. To catch up and stay in the race China needs to build a university system on par with the US. But as of now the only two universities in China with any research capability in semiconductors or much of anything else are Tsinghua University (https://www.tsinghua.edu.cn/en/) and Peking University (yes Peking https://english.pku.edu.cn). They are comparable to a 2nd tier US university such as Texas A & M (https://www.tamu.edu) or UC San Diego (https://ucsd.edu). Not exactly Harvard or MIT or Caltech and most definitely not Princeton (ha).
My concluding message is this. If at all there is any point in today’s globalized world for nationalism and self reliance specially in the sphere of semiconductors and other technologies, it must be backed by a good educational system. Everyone has heard of private US universities such as Harvard, Princeton and MIT, but it is the large public state university systems such as University of Texas, University of California, University of Michigan and so on and so on that are primary contributors to the US lead in science and technology. These university systems were started more than a 150 years ago through acts (called the Morill Acts) of the federal government going back the days of Abraham Lincoln.