6 ms·

Seems like quite a risk to only have one company provide these machines. Semiconductor companies seem to follow the same trend, the major contributing factor s

Seems like quite a risk to only have one company provide these machines. Semiconductor companies seem to follow the same trend, the major contributing factor seems to be the patent portfolio needed to secure cross-licensing deals. It looks like patents are actively preventing new companies in the semiconductor business.

hughrr 5y ago
I think the extreme cost and risk to enter the business is the key issue.
- cadence- 5y ago
  Sounds like something governments could help with.
  - hughrr 5y ago
    Indeed. Only time that happened was in East Germany and it didn’t work out too good.
    - SkyMarshal 5y ago
      That wasn't the only time.
    - Bootvis 5y ago
      It’s quite a bit more nuanced. I recommend reading some work by Mariana Mazzucato to get some perspective. https://www.google.com/search?q=mariana%20mazzucato https://www.google.com/search?q=mariana%20mazzucato
      - cinntaile 5y ago
        Which book would you recommend?
        SkyMarshal 5y ago
        Start with "The Value of Everything" [1], then "The Entrepreneurial State". [1]:https://news.ycombinator.com/item?id=23890312 https://news.ycombinator.com/item?id=23890312
        Bootvis 5y ago
        I saw her present her book “The Entrepreneurial State” and read it. It made me look at public and private R&D investments in a new way.
      - hughrr 5y ago
        It's really not. The cost of production was too high compared to other markets which reduced demand. Which is exactly what happens with all government projects.
    - utor 5y ago
      Interesting! Did you have any source about East German semiconductor history?
  - arp242 5y ago
    There are a number of companies that can produce lithography machines, just not on the same level as ASML (at the moment anyway). Even if we ignore any patent/IP and upfront investment issues, designing and actually constructing machines which are on par with ASML's is not an easy task and will take years of research, assuming the project succeeds in the first place. What ASML is doing now is the culmination of years of research and investment and some pretty specific know-how. It's kind of like trying to replicate NASA's James Webb telescope project: many of the basic principles aren't very hard, but actually building the thing is.
    - sharken 5y ago
      I agree it's a tall order to compete with ASML, patents notwithstanding. So yes patents are not everything, even though cross-licensing is beneficial, rather than invent it yourself. Actually the James Webb telescope, which has a scheduled launch this year, started out in 1996. Given the 25 years it has taken so far and a huge amount of money, then it's not too far fetched to imagine someone replicating a space telescope. https://en.m.wikipedia.org/wiki/James_Webb_Space_Telescope https://en.m.wikipedia.org/wiki/James_Webb_Space_Telescope And now it seems that the name might be changed before launch. https://www.nature.com/articles/d41586-021-02010-x https://www.nature.com/articles/d41586-021-02010-x
Sebb767 5y ago
That, and upfront investment costs. Very few companies have a few spare billion to pour into setting up the infrastructure, plus deal with resource deals, regulations, politics etc.. I'm pretty sure that a lot of countries would prefer to be a bit more independent on the semiconductor side, but getting up and running is hard. > It looks like patents are actively preventing new companies in the semiconductor business. As far as I know, the really deep technology on the CPUs is actually not patented - it would be nearly impossible to reverse engineer or proof an infringement, but you'd need to reveal your methodology to file the patent in the first place.
- jbverschoor 5y ago
  Yeah they have trade secrets. Patents are nice and all, but good luck sueing in China. Even if you win, it’ll be a one-time fee
Someone 5y ago
https://en.wikipedia.org/wiki/Moore%27s_second_law https://en.wikipedia.org/wiki/Moore%27s_second_law: “the cost of a semiconductor chip fabrication plant doubles every four years” I can’t find it, but I’ve seen people extrapolate that to compute when there would be room for only one fab in the world. Seems we’re there for one part of a fab.
- Beldin 5y ago
  Indeed. I've heard that there was a lead time of about 12 years on the EUV machines at ASML. I had a tour of one off their facilities once (generic recruitment thingy). Back in early 2000s they were working to correct for errors where the margin of error on their error detector was greater than the error that needed correcting. (The wave length used was about an order of magnitude greater than the lirhography result; I have no recollection of how they claimed to handle that.) They are pushing the boundaries of what's possible, using incredibly long lead times. I don't see how any start-up can compete with that, even in the absence of patents.
  - Throwawayaerlei 5y ago
    > The wave length used was about an order of magnitude greater than the lirhography result; I have no recollection of how they claimed to handle that. https://en.wikipedia.org/wiki/Multiple_patterning https://en.wikipedia.org/wiki/Multiple_patterning and it's pretty insane.
- jokoon 5y ago
  This law is really cancelled by Wirth's law
jbay808 5y ago
Canon and Nikon make similar machines too, but they've lost market share ever since ASML released their dual scan system. And at the moment EUV doesn't appear to be a mature enough market to support competitors.

hughrr 5y ago

I think the extreme cost and risk to enter the business is the key issue.

cadence- 5y ago
Sounds like something governments could help with.
- hughrr 5y ago
  Indeed. Only time that happened was in East Germany and it didn’t work out too good.
  - SkyMarshal 5y ago
    That wasn't the only time.
  - Bootvis 5y ago
    It’s quite a bit more nuanced. I recommend reading some work by Mariana Mazzucato to get some perspective. https://www.google.com/search?q=mariana%20mazzucato https://www.google.com/search?q=mariana%20mazzucato
    - cinntaile 5y ago
      Which book would you recommend?
      - SkyMarshal 5y ago
        Start with "The Value of Everything" [1], then "The Entrepreneurial State". [1]:https://news.ycombinator.com/item?id=23890312 https://news.ycombinator.com/item?id=23890312
      - Bootvis 5y ago
        I saw her present her book “The Entrepreneurial State” and read it. It made me look at public and private R&D investments in a new way.
    - hughrr 5y ago
      It's really not. The cost of production was too high compared to other markets which reduced demand. Which is exactly what happens with all government projects.
  - utor 5y ago
    Interesting! Did you have any source about East German semiconductor history?
- arp242 5y ago
  There are a number of companies that can produce lithography machines, just not on the same level as ASML (at the moment anyway). Even if we ignore any patent/IP and upfront investment issues, designing and actually constructing machines which are on par with ASML's is not an easy task and will take years of research, assuming the project succeeds in the first place. What ASML is doing now is the culmination of years of research and investment and some pretty specific know-how. It's kind of like trying to replicate NASA's James Webb telescope project: many of the basic principles aren't very hard, but actually building the thing is.
  - sharken 5y ago
    I agree it's a tall order to compete with ASML, patents notwithstanding. So yes patents are not everything, even though cross-licensing is beneficial, rather than invent it yourself. Actually the James Webb telescope, which has a scheduled launch this year, started out in 1996. Given the 25 years it has taken so far and a huge amount of money, then it's not too far fetched to imagine someone replicating a space telescope. https://en.m.wikipedia.org/wiki/James_Webb_Space_Telescope https://en.m.wikipedia.org/wiki/James_Webb_Space_Telescope And now it seems that the name might be changed before launch. https://www.nature.com/articles/d41586-021-02010-x https://www.nature.com/articles/d41586-021-02010-x

Sebb767 5y ago

That, and upfront investment costs. Very few companies have a few spare billion to pour into setting up the infrastructure, plus deal with resource deals, regulations, politics etc.. I'm pretty sure that a lot of countries would prefer to be a bit more independent on the semiconductor side, but getting up and running is hard. > It looks like patents are actively preventing new companies in the semiconductor business. As far as I know, the really deep technology on the CPUs is actually not patented - it would be nearly impossible to reverse engineer or proof an infringement, but you'd need to reveal your methodology to file the patent in the first place.

jbverschoor 5y ago
Yeah they have trade secrets. Patents are nice and all, but good luck sueing in China. Even if you win, it’ll be a one-time fee

Someone 5y ago

https://en.wikipedia.org/wiki/Moore%27s_second_law https://en.wikipedia.org/wiki/Moore%27s_second_law: “the cost of a semiconductor chip fabrication plant doubles every four years” I can’t find it, but I’ve seen people extrapolate that to compute when there would be room for only one fab in the world. Seems we’re there for one part of a fab.

Beldin 5y ago
Indeed. I've heard that there was a lead time of about 12 years on the EUV machines at ASML. I had a tour of one off their facilities once (generic recruitment thingy). Back in early 2000s they were working to correct for errors where the margin of error on their error detector was greater than the error that needed correcting. (The wave length used was about an order of magnitude greater than the lirhography result; I have no recollection of how they claimed to handle that.) They are pushing the boundaries of what's possible, using incredibly long lead times. I don't see how any start-up can compete with that, even in the absence of patents.
- Throwawayaerlei 5y ago
  > The wave length used was about an order of magnitude greater than the lirhography result; I have no recollection of how they claimed to handle that. https://en.wikipedia.org/wiki/Multiple_patterning https://en.wikipedia.org/wiki/Multiple_patterning and it's pretty insane.
jokoon 5y ago
This law is really cancelled by Wirth's law

jbay808 5y ago

Canon and Nikon make similar machines too, but they've lost market share ever since ASML released their dual scan system. And at the moment EUV doesn't appear to be a mature enough market to support competitors.