How important would labor costs be in the value chain, which would make Chinese competition more attractive on costs? Or is quality and specification more important?
Thank you for your comment. While we do not have the company's exact cost structure, we do know that (in order) they are Labor, Depreciation, Supplies, Energy, and Polysilicon. The labor cost does include R&D staff, and in terms of production related labor cost the company has been shifting its cost structure both towards lower cost geographies (Singapore) and implemented automation where possible. Generally speaking I do not see lower labor cost as a big edge in the industry. For one Siltronic has successfully competed with Japanese, Korean, and Taiwanese Players in the past, which should theoretically have had an edge in this area. But more importantly, quality and specification rules. If you are a company like TSMC, selling your finished products at an ASP of around USD 5,000.- per Wafer, making the decision to switch to a completely new supplier, which includes extensive upfront investment in specification and validation, to save 25% on a 100 USD virgin Wafer would be quite risky because your other costs per wafer do not change and if your yield per wafer decreases by just 50bps due to quality issues any money you might have saved on the virgin wafer is lost in wastage. Obviously, this math works out a little differently for memory manufacturers and less leading edge semiconductor fabs, but the trade-off remains.
Impressive analysis, really. I am kind of lost in your cost and revenue scenarios (and it is not your fault!!!) I wonder what kind of visibility the companies in this industry really have. I mean in the last quarter for example MU guided for some results, only to have to preannounce much worse results during the quarter. Granted my example may not be well chosen, but I think even AMD was surprised by lower demand. It seems that the last few years were a bit different from a normal semi cycle. What if we are going to revert to a normal boom and bust semi cycle of recent past?
Thank you for your comment. You are right to point out that visibility, including in the manufacturing of virgin wafers, used to be low historically. I think this is still the case in the semiconductor industry and for some products (e.g. the rather commoditized DRAM and NAND space) more so than others. The point I am trying to make on substrate producers is that, while volumes can still see sharp downturns, the increased use of LTA's and changed capital allocation policy of key players (i.e. do not expand without having high % of new capacity under LTA) have made pricing a lot more predictable and less volatile.
We are clearly currently going through the downswing phase of a semi cycle and are likely closer to the bottom than the top, but I fully expect demand for semiconductors to continue to rise over the long-term and believe that Siltronic is well positioned to benefit from this trend. What I tried to show in Figure 7 is that, at reasonable multiples, the market is already pricing in pretty conservative scenarios. If you are willing to pay 12 times what I reckon to be trough EBIT for a company that is at the leading edge of its industry, with close customer ties, and good growth perspective (especially off that trough), and should management be right about its estimate off flat to slightly up pricing, volumes would need to decrease somewhere in the vicinity of 30% to for the current share price to be justified. A 30% volume decrease would rival the one the industry experienced when the Dotcom bubble burst.
Great analysis, thanks for sharing. I've noticed that Siltronics main competitors have a comparable low valuation. What do you think are the reasons why the capital market doesn't value the importance of this essential part of the semi value chain - especially when the 5 key players form in a way an oligopoly?
Thank you for your comment. Generally speaking I think the market still considers wafers to be commodity products, not accounting for the increased complexity that goes into co-designing and making them, the more extensive use of LTAs and (on a through cycle average basis) tighter market. It is true that demand is still quite cyclical, but what I hoped to bring across is that, even though we are coming off peak demand, current valuations mean you are already seeing reasonable multiples on trough volumes. When comparing these companies using relative valuation measures you should consider the fact that the Japanese peers tend to run their business with a high excess cash position, which will push down EV and lower EV/EBIT. Since this cash is not very productive you might want to adjust for it or use a P/E multiple instead. Also, keep in mind that Shin-Etsu is not a pure play.
The fact that the industry has an oligopolistic structure does not necessarily tell you a lot about the returns earned by companies in it. There are many examples for both oligopolies with strong competition and low excess returns (e.g. air travel) and those with low competition where profits are bountiful (e.g. card payment networks). That being said, the increased consolidation of the sector, mirroring the one in the semiconductor fabrication industry, can tell you something about benefits of scale in an increasingly technologically challenging business.
Why has this investment not worked?
Very nice analysis.
How important would labor costs be in the value chain, which would make Chinese competition more attractive on costs? Or is quality and specification more important?
Hi Emerging Value,
Thank you for your comment. While we do not have the company's exact cost structure, we do know that (in order) they are Labor, Depreciation, Supplies, Energy, and Polysilicon. The labor cost does include R&D staff, and in terms of production related labor cost the company has been shifting its cost structure both towards lower cost geographies (Singapore) and implemented automation where possible. Generally speaking I do not see lower labor cost as a big edge in the industry. For one Siltronic has successfully competed with Japanese, Korean, and Taiwanese Players in the past, which should theoretically have had an edge in this area. But more importantly, quality and specification rules. If you are a company like TSMC, selling your finished products at an ASP of around USD 5,000.- per Wafer, making the decision to switch to a completely new supplier, which includes extensive upfront investment in specification and validation, to save 25% on a 100 USD virgin Wafer would be quite risky because your other costs per wafer do not change and if your yield per wafer decreases by just 50bps due to quality issues any money you might have saved on the virgin wafer is lost in wastage. Obviously, this math works out a little differently for memory manufacturers and less leading edge semiconductor fabs, but the trade-off remains.
Best,
Tachy
very insightful
Impressive analysis, really. I am kind of lost in your cost and revenue scenarios (and it is not your fault!!!) I wonder what kind of visibility the companies in this industry really have. I mean in the last quarter for example MU guided for some results, only to have to preannounce much worse results during the quarter. Granted my example may not be well chosen, but I think even AMD was surprised by lower demand. It seems that the last few years were a bit different from a normal semi cycle. What if we are going to revert to a normal boom and bust semi cycle of recent past?
Hi Stefano,
Thank you for your comment. You are right to point out that visibility, including in the manufacturing of virgin wafers, used to be low historically. I think this is still the case in the semiconductor industry and for some products (e.g. the rather commoditized DRAM and NAND space) more so than others. The point I am trying to make on substrate producers is that, while volumes can still see sharp downturns, the increased use of LTA's and changed capital allocation policy of key players (i.e. do not expand without having high % of new capacity under LTA) have made pricing a lot more predictable and less volatile.
We are clearly currently going through the downswing phase of a semi cycle and are likely closer to the bottom than the top, but I fully expect demand for semiconductors to continue to rise over the long-term and believe that Siltronic is well positioned to benefit from this trend. What I tried to show in Figure 7 is that, at reasonable multiples, the market is already pricing in pretty conservative scenarios. If you are willing to pay 12 times what I reckon to be trough EBIT for a company that is at the leading edge of its industry, with close customer ties, and good growth perspective (especially off that trough), and should management be right about its estimate off flat to slightly up pricing, volumes would need to decrease somewhere in the vicinity of 30% to for the current share price to be justified. A 30% volume decrease would rival the one the industry experienced when the Dotcom bubble burst.
Hope that helps.
Best,
Tachy
Great analysis, thanks for sharing. I've noticed that Siltronics main competitors have a comparable low valuation. What do you think are the reasons why the capital market doesn't value the importance of this essential part of the semi value chain - especially when the 5 key players form in a way an oligopoly?
Hi Max,
Thank you for your comment. Generally speaking I think the market still considers wafers to be commodity products, not accounting for the increased complexity that goes into co-designing and making them, the more extensive use of LTAs and (on a through cycle average basis) tighter market. It is true that demand is still quite cyclical, but what I hoped to bring across is that, even though we are coming off peak demand, current valuations mean you are already seeing reasonable multiples on trough volumes. When comparing these companies using relative valuation measures you should consider the fact that the Japanese peers tend to run their business with a high excess cash position, which will push down EV and lower EV/EBIT. Since this cash is not very productive you might want to adjust for it or use a P/E multiple instead. Also, keep in mind that Shin-Etsu is not a pure play.
The fact that the industry has an oligopolistic structure does not necessarily tell you a lot about the returns earned by companies in it. There are many examples for both oligopolies with strong competition and low excess returns (e.g. air travel) and those with low competition where profits are bountiful (e.g. card payment networks). That being said, the increased consolidation of the sector, mirroring the one in the semiconductor fabrication industry, can tell you something about benefits of scale in an increasingly technologically challenging business.
Best,
Tachy