Thanks for the write-up! It's very interesting to peel back the layers on how semis are made. Correct me where I am wrong. The process you described is making one layer of the wafer. Do the # layers increase as nodes get smaller or that is not correlated? Would EUV be used in making all those layers or only the smallest patterns?
Sorry to get around to this late. This is only for 1 layer I believe. Even w/ my obsession I am not 100% sure. I believe that multiple layers are mostly done in deposition and etch more so than Lithography - but the first layer begins with this. This is the foundation every time.
So I have figured out the conclusion to this - yes they do multiple lithography. Usually Lithography > Etch > Lithography > Etch, called LELE, sometimes even up to 3 steps, or LELELE
Thanks for write-up, super helpful as always- may I ask, i get it’s purely illustrative but is there are a reason you’re using net income (so net of d&a) and still subtracting capex in your DCF?
Also referring to your previous posts - if they own 100% of the best incremental technology and we think demand for semis will grow exponentially (due to AI) I would have thought the CAGR implied by the 2025 target is not particularly impressive? Or I am missing something in the bargaining power they have with customers or how the demand for their product correlates with semis demand?
This is a perfect question. I will give it to you that I am probably on the conservative side, but NI has never been above FCF for this company as there is a very rough working capital requirement to grow. They also need to expand capex as well to ramp more units of EUV - but I still think that even if we are being nice and did 100% conversion it wouldn't really move the needle.
And while semi demand will grow exponentially it isn't necessary that units will grow exponentially. There is a chance that units do grow in excess of history, but rather it seems to me that units might just change composition meaningfully - from CPUs to other specialized chips. Nothing is a given right now - and even if we just assume 100% FCF conversion and their high side estimate - it's still hard to make the numbers work.
The 2025 CAGR is not that impressive, but a reminder that once the system gets shipped it can be used for multiple types of chips, and as long as they can fulfill the capacity demands of the 5nm and below chips, I think there is no reason to buy extra units. That is likely what the divergence is. I do believe that there is likely upside surprise - but I would be shocked if it was meaningfully above their best-case scenario of EUV insertion. I could of course be wrong - but will be paying attention for anything that would change my priors.
Super helpful, makes a lot of sense thank you. Another think that I thought was interesting is that the gross margin is obviously good but for a "monopolistic" type of business seems like it could be higher (ie. pharma with patents have higher GM).
In that regard I thought the relationship with Carl Zeiss SMT ("Carl Zeiss SMT GmbH, in which ASML owns an indirect interest of 24.9%, is our single supplier, and we are their single customer, of Optical Columns for lithography systems" from AR19) has an impact on that - maybe they capture some of gross margin as well. Or maybe the investment from TSMC / Intel etc on EUV had an implicit pricing agreement on the back (an explicit one would have been captured in filings I assume).
Fascinating business anyways, thanks for the great write-up.
Thanks for the write-up! It's very interesting to peel back the layers on how semis are made. Correct me where I am wrong. The process you described is making one layer of the wafer. Do the # layers increase as nodes get smaller or that is not correlated? Would EUV be used in making all those layers or only the smallest patterns?
Sorry to get around to this late. This is only for 1 layer I believe. Even w/ my obsession I am not 100% sure. I believe that multiple layers are mostly done in deposition and etch more so than Lithography - but the first layer begins with this. This is the foundation every time.
So I have figured out the conclusion to this - yes they do multiple lithography. Usually Lithography > Etch > Lithography > Etch, called LELE, sometimes even up to 3 steps, or LELELE
Thanks for write-up, super helpful as always- may I ask, i get it’s purely illustrative but is there are a reason you’re using net income (so net of d&a) and still subtracting capex in your DCF?
Also referring to your previous posts - if they own 100% of the best incremental technology and we think demand for semis will grow exponentially (due to AI) I would have thought the CAGR implied by the 2025 target is not particularly impressive? Or I am missing something in the bargaining power they have with customers or how the demand for their product correlates with semis demand?
Many thanks
This is a perfect question. I will give it to you that I am probably on the conservative side, but NI has never been above FCF for this company as there is a very rough working capital requirement to grow. They also need to expand capex as well to ramp more units of EUV - but I still think that even if we are being nice and did 100% conversion it wouldn't really move the needle.
And while semi demand will grow exponentially it isn't necessary that units will grow exponentially. There is a chance that units do grow in excess of history, but rather it seems to me that units might just change composition meaningfully - from CPUs to other specialized chips. Nothing is a given right now - and even if we just assume 100% FCF conversion and their high side estimate - it's still hard to make the numbers work.
The 2025 CAGR is not that impressive, but a reminder that once the system gets shipped it can be used for multiple types of chips, and as long as they can fulfill the capacity demands of the 5nm and below chips, I think there is no reason to buy extra units. That is likely what the divergence is. I do believe that there is likely upside surprise - but I would be shocked if it was meaningfully above their best-case scenario of EUV insertion. I could of course be wrong - but will be paying attention for anything that would change my priors.
Super helpful, makes a lot of sense thank you. Another think that I thought was interesting is that the gross margin is obviously good but for a "monopolistic" type of business seems like it could be higher (ie. pharma with patents have higher GM).
In that regard I thought the relationship with Carl Zeiss SMT ("Carl Zeiss SMT GmbH, in which ASML owns an indirect interest of 24.9%, is our single supplier, and we are their single customer, of Optical Columns for lithography systems" from AR19) has an impact on that - maybe they capture some of gross margin as well. Or maybe the investment from TSMC / Intel etc on EUV had an implicit pricing agreement on the back (an explicit one would have been captured in filings I assume).
Fascinating business anyways, thanks for the great write-up.