For this question, I think it’s useful to first understand the basic shape of the relevant market and its major players.
Right now, I understand 14nm chips and smaller make up ~70% of the global market [1]. While I haven’t found any information about how this will likely change in the next 3 years, I found a prediction about how production capacity will change [2]. According to that, production capacity of <20nm chips will increase from ~35% in 2019 to ~45% in 2023. Bringing these two points together suggests to me that our target, <=28nm chips, will make up between 80% and 90% of the value of the entire market. I haven’t found any specific numbers for China, but I think assuming that the Chinese market will largely reflect the global market here is fine. This implies that the potential for SMIC to earn revenue from this category is extremely large. The total outside view might therefore suggest that SMIC can ~double its revenue share from advanced chips, in line with the global movement.
Next, let’s look at the other players ([3] provides a list of major foundries by revenue). I have only found info for three major players on their revenue share from different chips: TSMC made 74% of its revenue from <=28nm chips in Q2 2021 [4], while UMC made 20% of its revenue from those chips [5]. Tower Semiconductors makes no revenue from that segment [6]. In absolute terms, this means that TSMC is making ~$40bn p.a. off those nodes, UMC is making ~$1bn, and SMIC is currently making ~$0.2bn from that segment. From all of the above it seems that *very* roughly a monthly production capacity of transforming 1 million 200mm-equivalent wafers into <=28nm chips nets revenue of ~$4bn per year (this is corroborated by SMIC’s own figures - see below).
Let’s now turn to SMIC itself. As a starting point, I think that [7] provides a useful overview of SMIC’s capacity from 2020Q1. That source suggests that SMIC had 6-56k capacity for <=28nm nodes. Given that the full 56k capacity also includes 40nm chips, which make up a substantial share of SMIC’s revenue, it’s more likely that the 2020Q1 capacity was more towards the lower end of that range.
Since then, SMIC have added roughly 20k capacity per quarter [8]. From [9], it seems that the major ongoing capacity expansion has come from their Beijing Fab, which produces <=28nm chips, and their Tianjin Fab, which doesn’t. Considering that a huge chunk of SMIC’s revenue comes from 40nm chips, which they only can produce in their Beijing Fab, I think it’s fair to assume that SMIC started out with less than 50k for advanced chips but has slowly ramped up their capacity over the past year. In my best-guess estimate, they are now at ~50k capacity for advanced chips, generating ~$3.5M revenue per 1k capacity, which is ~150% of the average revenue from their capacity. I think it’s also fair to assume that this trend will continue for at least the next year, which would suggest that the revenue share of advanced chips increases by something like 2% per half-year.
From ~mid-2022, most of their capacity increases will likely come from SMIC’s new Shenzhen project [10], which will exclusively produce advanced chips. This suggests that from there, the share of advanced chips will increase more rapidly. Eye-balling again, I’d expect that from then on revenue share of advanced chips will increase by roughly 4% per quarter.
Of course, there are probably dozens of factors which will impact this estimate of three years, ranging from war and catastrophe on one side to further worsening of the chip crunch and a permanent price spike for those chips on the other side. But intuitively, nothing suggests to me that in a market moving this slowly, we should expect other major movements in expectation.
Finally, I need to consider how to approach this new method of recording forecasts. Based on the above, I’m choosing a 2% increase per half-year until 2022H1 and a 4% increase after that, ending at 32% revenue share. The interval is much more difficult though. It’s really hard to cover 80% of expected outcomes. Therefore, this interval should be very wide. For the lower bound, I don’t think it’s reasonable to assume that SMIC’s revenue will decline substantially from that segment. But throughout the period in question, it’s easily imaginable that it doesn’t increase either - primarily because SMIC may not manage to get its new plants usefully operational, or older plants fall short of expected production. I’m therefore using a lower bound which slowly increases from 7% revenue in 2021H2 to 12% in 2024H1.
For the upper bound, I can easily imagine that I am wrong by a factor of 2, especially further down the line. However, it’s rather unlikely that it will increase by much more than that, given that TSMC seems like a natural upper bound with their 75%. I’m therefore putting the upper bound at roughly 1.8x of my central estimate.
For this question, I think it’s useful to first understand the basic shape of the relevant market and its major players.
Right now, I understand 14nm chips and smaller make up ~70% of the global market [1]. While I haven’t found any information about how this will likely change in the next 3 years, I found a prediction about how production capacity will change [2]. According to that, production capacity of <20nm chips will increase from ~35% in 2019 to ~45% in 2023. Bringing these two points together suggests to me that our target, <=28nm chips, will make up between 80% and 90% of the value of the entire market. I haven’t found any specific numbers for China, but I think assuming that the Chinese market will largely reflect the global market here is fine. This implies that the potential for SMIC to earn revenue from this category is extremely large. The total outside view might therefore suggest that SMIC can ~double its revenue share from advanced chips, in line with the global movement.
Next, let’s look at the other players ([3] provides a list of major foundries by revenue). I have only found info for three major players on their revenue share from different chips: TSMC made 74% of its revenue from <=28nm chips in Q2 2021 [4], while UMC made 20% of its revenue from those chips [5]. Tower Semiconductors makes no revenue from that segment [6]. In absolute terms, this means that TSMC is making ~$40bn p.a. off those nodes, UMC is making ~$1bn, and SMIC is currently making ~$0.2bn from that segment. From all of the above it seems that *very* roughly a monthly production capacity of transforming 1 million 200mm-equivalent wafers into <=28nm chips nets revenue of ~$4bn per year (this is corroborated by SMIC’s own figures - see below).
Let’s now turn to SMIC itself. As a starting point, I think that [7] provides a useful overview of SMIC’s capacity from 2020Q1. That source suggests that SMIC had 6-56k capacity for <=28nm nodes. Given that the full 56k capacity also includes 40nm chips, which make up a substantial share of SMIC’s revenue, it’s more likely that the 2020Q1 capacity was more towards the lower end of that range.
Since then, SMIC have added roughly 20k capacity per quarter [8]. From [9], it seems that the major ongoing capacity expansion has come from their Beijing Fab, which produces <=28nm chips, and their Tianjin Fab, which doesn’t. Considering that a huge chunk of SMIC’s revenue comes from 40nm chips, which they only can produce in their Beijing Fab, I think it’s fair to assume that SMIC started out with less than 50k for advanced chips but has slowly ramped up their capacity over the past year. In my best-guess estimate, they are now at ~50k capacity for advanced chips, generating ~$3.5M revenue per 1k capacity, which is ~150% of the average revenue from their capacity. I think it’s also fair to assume that this trend will continue for at least the next year, which would suggest that the revenue share of advanced chips increases by something like 2% per half-year.
From ~mid-2022, most of their capacity increases will likely come from SMIC’s new Shenzhen project [10], which will exclusively produce advanced chips. This suggests that from there, the share of advanced chips will increase more rapidly. Eye-balling again, I’d expect that from then on revenue share of advanced chips will increase by roughly 4% per quarter.
Of course, there are probably dozens of factors which will impact this estimate of three years, ranging from war and catastrophe on one side to further worsening of the chip crunch and a permanent price spike for those chips on the other side. But intuitively, nothing suggests to me that in a market moving this slowly, we should expect other major movements in expectation.
Finally, I need to consider how to approach this new method of recording forecasts. Based on the above, I’m choosing a 2% increase per half-year until 2022H1 and a 4% increase after that, ending at 32% revenue share. The interval is much more difficult though. It’s really hard to cover 80% of expected outcomes. Therefore, this interval should be very wide. For the lower bound, I don’t think it’s reasonable to assume that SMIC’s revenue will decline substantially from that segment. But throughout the period in question, it’s easily imaginable that it doesn’t increase either - primarily because SMIC may not manage to get its new plants usefully operational, or older plants fall short of expected production. I’m therefore using a lower bound which slowly increases from 7% revenue in 2021H2 to 12% in 2024H1.
For the upper bound, I can easily imagine that I am wrong by a factor of 2, especially further down the line. However, it’s rather unlikely that it will increase by much more than that, given that TSMC seems like a natural upper bound with their 75%. I’m therefore putting the upper bound at roughly 1.8x of my central estimate.
[1]: https://www.prnewswire.com/il/news-releases/expert-optimistic-about-china-s-chipmaking-industry-825715340.html
[2]: https://www.eenewsanalog.com/news/sub-28nm-processes-make-half-global-chip-production
[3]: https://www.electronicsweekly.com/news/business/record-q1-foundry-revenues-2021-06/
[4]: https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2021-07/048b7a33a72733c8dd5ab4aa3b6a02ed6167ffbb/2Q21ManagementReport.pdf
[5]: https://www.umc.com/upload/media/08_Investors/Financials/Quarterly_Results/Quarterly_2020-2029_English_pdf/2020/Q4_2020/UMC20Q4_report.pdf
[6]: https://ir.towersemi.com/static-files/4bf2cc6a-a3f3-480d-ab30-05b5ab8e7ab2
[7]: https://www.eetimes.com/smic-advanced-process-technologies-and-govt-funding/
[8]: https://www.smics.com/uploads/610ba2e7/Q2_financials.pdf
[9]: https://www.smics.com/uploads/Q2_2020%20Financials-2.pdf
[10]: https://www.anandtech.com/show/16568/smic-to-build-a-new-fab-in-shenzhen-production-to-start-in-2022