I work in electronics design and manufacturing. STM32 MCUs are impossible to buy at the moment, have been for the last year, and probably will be for the next too (unless you are a Tier 1 like Apple). This has caused an interesting effect; everyone is designing out STM32 parts as fast as possible, and not even considering them for new builds. Instead, they are being replaced by Chinese grown MCUs that are “clones” of STM32s like the gigadevice GD32 or Geehy APM32, as well as the more well known Espressif ESP32 which seems to be relatively available on the market.
And truthfully, the reason STM32s are the number 1 MCU is not the software libraries, which were awful and buggy for the first 5-10 years - they actually made development slower than programming on a register level. The real reason is that they were cheap with lots of peripherals and relatively abundant; ST also pioneered the low cost evaluation kit where you could buy an MCU dev kit for $20 instead of $2000 which was unheard of at the time. This caused the low end of the market to standardise on them. Since hobbyists, universities and engineers playing with their first ARM MCU were all using STM32, you could basically assume that anyone who applied for a job in embedded systems had some experience with them. And because it is what everyone had experience with, that is what they chose for their designs. Other vendors later followed with cheap dev kits, but it was too late.
Unless ST can solve its supply problems soon, I think they will begin to see their market shrink from the low cost end. They probably won’t care about this initially as it would hardly register in their profits, but in the long term the general knowledge base of everyone will be moving away from their devices (and to some extent, already is), giving mindshare and revenue to their competitors like Gigadevice, Nordic and Espressif and allowing them to grow.
I think there are a few things preventing this at the moment. The obvious one is the current international political tensions, on which I will make no comment other than stating that they are present. The next is that they aren’t available at usual distributors like Digi-Key and Mouser, but this is changing slowly (for example, ESP32s are available nowadays, as well as gigadevice flash). Some of the STM32 “clone” competitors are on shaky ground as well, as they use copy+paste code from ST SDKs so they are unlikely to appear unless they fix this. There is a lot of other dubious copyright stuff going on which makes these companies somewhat radioactive (see Allwinner and GPL for example). Finally, there appears to be a different approach culturally to how these companies operate in my experience. In general, the “western” MCU providers seem to have good, comprehensive written documentation. On the other hand, Chinese MCU providers seem to have very poor written documentation (I’m not talking about poor English, I mean more missing critical details, or simply not existing at all). It seems that these companies expect you to contact them and their FAEs in order to get information, or subcontract the development to one of their approved developers. Most western companies would not want to share that level of detail with another western company, let alone a Chinese company that operates with ambiguity in IP. The only shining beacon of hope here is Espressif; early on they hired some European developers who understood these issues, and the ESP32 documentation is now top notch. I don’t think ESP32 devices even remotely have a western competitor for many reasons. I have shipped a lot of devices (many thousands) which have ESP32s and I have never spoken to ESP32 FAEs.
Thanks for the very informative reply. It seems more quality focused Chinese company would still be able to compete and win in the global market? TSMC is Espressif’s foundry though. So technically unless it switches to SMIC it’s not a purely Chinese product. I guess my question is whether or not you see the possibility of a quality competitor arising in China which uses a fully “Chinese made” supply chain that would not be affected by the US led semiconductor containment strategy? Particularly in the 28nm and above sector.
Would a Chinese company be able to *design* complex semiconductors that rival the best on the market using home-grown engineers? I think that Espressif has shown that absolutely, unequivocally yes. Another much smaller but still very interesting company with their own niche is WCH - although their documentation and such is maybe not quite up to that of Espressif (but is still borderline passable), they make good products for good prices and listen to customer feedback (I am not sure who their fab is).
The other question of course is about whether a Chinese company could *manufacture* these complex semiconductors. This is certainly leaving my comfort zone, but I personally think it will be possible for 28nm and above, though maybe not with the current state of world affairs. But I can't really see it happening any time soon for the more leading edge stuff.
I think it is common for people to reduce the semiconductor industry to questions of revenue, profits, forecasts, etc but I think that is missing a key detail. We aren't talking about a factory that makes donuts here - semiconductor devices are arguably the most complex creation of humankind. The attention to detail, quality and methodical nature required in their research and development is like nothing else. When you read something like "TSMC is finding it hard to get 3nm working", it makes you think of a student that is struggling with their homework. It doesn't convey the magnitude of the problems that TSMC engineers and researchers face being held to the (unreasonable!) standard of Moore's Law. I think the idea that any one country can rapidly create even a fraction of all of that underlying knowledge and experience just by spending a lot of money is naive at best.
I once went to a talk from a guy who started a MMIC integration company. Their products were very niche/low volume, unbelievable specifications, eye watering price and extremely export controlled. He was discussing some of their technology (the unclassified stuff anyway) in great detail - and I mean *great* detail. One of the people in the audience asked at the end of the presentation (and I am paraphrasing here) "I don't understand why you would tell us all of this in such great detail; aren't you afraid of your competitors finding out?" His response was to laugh and say (again, paraphrasing) "It's okay, I think you will still find it challenging to replicate my results."
I think that guy truly understood the value of his experience and knowledge, as opposed to the information in an hour of powerpoint slides. I'm sure he would sell you the slides for $20 if you really wanted, but if you want to gain the knowledge behind it you've probably got 20 years of tinkering with MMICs ahead of you no matter how wealthy you are.
Hemm, this is fascinating. It seems the hardest part is the years of experience which requires failures, which is the very thing that the Chinese government won’t allow. Very good point.
I work in electronics design and manufacturing. STM32 MCUs are impossible to buy at the moment, have been for the last year, and probably will be for the next too (unless you are a Tier 1 like Apple). This has caused an interesting effect; everyone is designing out STM32 parts as fast as possible, and not even considering them for new builds. Instead, they are being replaced by Chinese grown MCUs that are “clones” of STM32s like the gigadevice GD32 or Geehy APM32, as well as the more well known Espressif ESP32 which seems to be relatively available on the market.
And truthfully, the reason STM32s are the number 1 MCU is not the software libraries, which were awful and buggy for the first 5-10 years - they actually made development slower than programming on a register level. The real reason is that they were cheap with lots of peripherals and relatively abundant; ST also pioneered the low cost evaluation kit where you could buy an MCU dev kit for $20 instead of $2000 which was unheard of at the time. This caused the low end of the market to standardise on them. Since hobbyists, universities and engineers playing with their first ARM MCU were all using STM32, you could basically assume that anyone who applied for a job in embedded systems had some experience with them. And because it is what everyone had experience with, that is what they chose for their designs. Other vendors later followed with cheap dev kits, but it was too late.
Unless ST can solve its supply problems soon, I think they will begin to see their market shrink from the low cost end. They probably won’t care about this initially as it would hardly register in their profits, but in the long term the general knowledge base of everyone will be moving away from their devices (and to some extent, already is), giving mindshare and revenue to their competitors like Gigadevice, Nordic and Espressif and allowing them to grow.
Thanks for the insightful comment. Do you think the Chinese suppliers will be able to eventually overtake the share of STMicro in this sector?
I think there are a few things preventing this at the moment. The obvious one is the current international political tensions, on which I will make no comment other than stating that they are present. The next is that they aren’t available at usual distributors like Digi-Key and Mouser, but this is changing slowly (for example, ESP32s are available nowadays, as well as gigadevice flash). Some of the STM32 “clone” competitors are on shaky ground as well, as they use copy+paste code from ST SDKs so they are unlikely to appear unless they fix this. There is a lot of other dubious copyright stuff going on which makes these companies somewhat radioactive (see Allwinner and GPL for example). Finally, there appears to be a different approach culturally to how these companies operate in my experience. In general, the “western” MCU providers seem to have good, comprehensive written documentation. On the other hand, Chinese MCU providers seem to have very poor written documentation (I’m not talking about poor English, I mean more missing critical details, or simply not existing at all). It seems that these companies expect you to contact them and their FAEs in order to get information, or subcontract the development to one of their approved developers. Most western companies would not want to share that level of detail with another western company, let alone a Chinese company that operates with ambiguity in IP. The only shining beacon of hope here is Espressif; early on they hired some European developers who understood these issues, and the ESP32 documentation is now top notch. I don’t think ESP32 devices even remotely have a western competitor for many reasons. I have shipped a lot of devices (many thousands) which have ESP32s and I have never spoken to ESP32 FAEs.
Thanks for the very informative reply. It seems more quality focused Chinese company would still be able to compete and win in the global market? TSMC is Espressif’s foundry though. So technically unless it switches to SMIC it’s not a purely Chinese product. I guess my question is whether or not you see the possibility of a quality competitor arising in China which uses a fully “Chinese made” supply chain that would not be affected by the US led semiconductor containment strategy? Particularly in the 28nm and above sector.
I would maybe break it down into a few responses.
Would a Chinese company be able to *design* complex semiconductors that rival the best on the market using home-grown engineers? I think that Espressif has shown that absolutely, unequivocally yes. Another much smaller but still very interesting company with their own niche is WCH - although their documentation and such is maybe not quite up to that of Espressif (but is still borderline passable), they make good products for good prices and listen to customer feedback (I am not sure who their fab is).
The other question of course is about whether a Chinese company could *manufacture* these complex semiconductors. This is certainly leaving my comfort zone, but I personally think it will be possible for 28nm and above, though maybe not with the current state of world affairs. But I can't really see it happening any time soon for the more leading edge stuff.
I think it is common for people to reduce the semiconductor industry to questions of revenue, profits, forecasts, etc but I think that is missing a key detail. We aren't talking about a factory that makes donuts here - semiconductor devices are arguably the most complex creation of humankind. The attention to detail, quality and methodical nature required in their research and development is like nothing else. When you read something like "TSMC is finding it hard to get 3nm working", it makes you think of a student that is struggling with their homework. It doesn't convey the magnitude of the problems that TSMC engineers and researchers face being held to the (unreasonable!) standard of Moore's Law. I think the idea that any one country can rapidly create even a fraction of all of that underlying knowledge and experience just by spending a lot of money is naive at best.
I once went to a talk from a guy who started a MMIC integration company. Their products were very niche/low volume, unbelievable specifications, eye watering price and extremely export controlled. He was discussing some of their technology (the unclassified stuff anyway) in great detail - and I mean *great* detail. One of the people in the audience asked at the end of the presentation (and I am paraphrasing here) "I don't understand why you would tell us all of this in such great detail; aren't you afraid of your competitors finding out?" His response was to laugh and say (again, paraphrasing) "It's okay, I think you will still find it challenging to replicate my results."
I think that guy truly understood the value of his experience and knowledge, as opposed to the information in an hour of powerpoint slides. I'm sure he would sell you the slides for $20 if you really wanted, but if you want to gain the knowledge behind it you've probably got 20 years of tinkering with MMICs ahead of you no matter how wealthy you are.
Thanks for your great insight here - this is a great thread of posts. Appreciated!
Hemm, this is fascinating. It seems the hardest part is the years of experience which requires failures, which is the very thing that the Chinese government won’t allow. Very good point.