Nice summary ! At some point, someone should write an article of the intertwining of Semiconductor manufacturing with Electronic Design Automation... which is the key enablement to use these technologies.
Yes! I worked (as a sysadmin for the TechPubs group) at Cadence Design Systems (San Jose, CA) during 1989-90, and came to admire their product and the engineers who created it.
At the risk of coming across as vain, I recommend the book that Dan Nenni and I wrote called "Fabless: the transformation of the semiconductor industry". If you want it on paper it is on Amazon. If you are happy with a pdf you can downloadd it for free from the Semiwiki website (look for the books section).
Truly excellent article, like one of John Y's at Asianometry. Eventually XUV runs out too, and my impression is that free-electron lasers will be needed to go even tighter. If XUV was a gamble, getting FEL to work economically at scale is like playing Powerball to win. Humanity has been very lucky so far in coming up with lots of similar big wins in innovation in this area just when they are needed.
As you pointed out, managing partnering and collaboration with suppliers and tech majors has been a major factor in ASML's success. The acquisition of Silicon Valley Group (SVG) in 2001 posed a significant challenge in that regard. The SVG technology afforded ASML a singular advantage at a crucial time. However, imagine how uncomfortable it would feel for a California company to suddenly abide by Netherland norms. SVG operated at a Silicon Valley cadence and innovation pattern, attracting and retaining talent that thrived in that culture. Would they stay? For ASML, the folks at SVG must have been a bit strange. The Californians' technical prowess may have been great, but they were probably seen as odd by Dutch standards of design and execution inherent in ASML's culture and processes. This cultural clash most certainly caused dissonance in both companies. However, it appears ASML leadership harmonized differences and brought out the best in both companies without squashing or compromising the traits that had made them stand out. It seldom works out that way.
One minor (and widespread) error is assuming that anything on a 5nm process is actually 5nm (such as the transistor size or spacing). For exaqmple, on TSMC's 5nm process, the transistor pitch (also known as contacted poiy pitch or CPP) is actually 57nm (about half of this is transistor and about half is spacing between transistors). This is still tiny, and much smaller than the 193nm light used in pre-EUV lithography.
Another common error is your comment that a modern iPhone is thousands of times faster than the computers NASA used to get to the moon (a total of about 1 MIPS apparently). That is technically true since tens of trillions is some number of thousands, but it is a bit like saying that the flight from earth to the moon went faster than a snail.
Nice summary ! At some point, someone should write an article of the intertwining of Semiconductor manufacturing with Electronic Design Automation... which is the key enablement to use these technologies.
Yes! I worked (as a sysadmin for the TechPubs group) at Cadence Design Systems (San Jose, CA) during 1989-90, and came to admire their product and the engineers who created it.
At the risk of coming across as vain, I recommend the book that Dan Nenni and I wrote called "Fabless: the transformation of the semiconductor industry". If you want it on paper it is on Amazon. If you are happy with a pdf you can downloadd it for free from the Semiwiki website (look for the books section).
Truly excellent article, like one of John Y's at Asianometry. Eventually XUV runs out too, and my impression is that free-electron lasers will be needed to go even tighter. If XUV was a gamble, getting FEL to work economically at scale is like playing Powerball to win. Humanity has been very lucky so far in coming up with lots of similar big wins in innovation in this area just when they are needed.
Brilliant assessment and storytelling. Thanks.
As you pointed out, managing partnering and collaboration with suppliers and tech majors has been a major factor in ASML's success. The acquisition of Silicon Valley Group (SVG) in 2001 posed a significant challenge in that regard. The SVG technology afforded ASML a singular advantage at a crucial time. However, imagine how uncomfortable it would feel for a California company to suddenly abide by Netherland norms. SVG operated at a Silicon Valley cadence and innovation pattern, attracting and retaining talent that thrived in that culture. Would they stay? For ASML, the folks at SVG must have been a bit strange. The Californians' technical prowess may have been great, but they were probably seen as odd by Dutch standards of design and execution inherent in ASML's culture and processes. This cultural clash most certainly caused dissonance in both companies. However, it appears ASML leadership harmonized differences and brought out the best in both companies without squashing or compromising the traits that had made them stand out. It seldom works out that way.
Brilliant
One minor (and widespread) error is assuming that anything on a 5nm process is actually 5nm (such as the transistor size or spacing). For exaqmple, on TSMC's 5nm process, the transistor pitch (also known as contacted poiy pitch or CPP) is actually 57nm (about half of this is transistor and about half is spacing between transistors). This is still tiny, and much smaller than the 193nm light used in pre-EUV lithography.
Another common error is your comment that a modern iPhone is thousands of times faster than the computers NASA used to get to the moon (a total of about 1 MIPS apparently). That is technically true since tens of trillions is some number of thousands, but it is a bit like saying that the flight from earth to the moon went faster than a snail.
Great description of what a giant company does and how they got where they're at.