One would expect that reverse engineering technologies wouldn’t be impossible.
The importance of these technologies for China is huge.
Moreover, they have a massive budget and probably also own a couple of ASML’s euv lithography machines.
Why is it that they cant remake these machines?
I heard some sources say that it might take them a decade.
In: 338
Followup question by not OP: What the hell is ASML EUV lithography?
To give a real world example of “just because you know how it works doesn’t mean you can replicate it”:
Take a ball of yarn, completely unwind it. Then rewind it perfectly, so it looks like a brand new one.
It’s literally just yarn, but it’s really difficult to do without the right equipment and knowhow.
In manufacturing anything, there’s a lot of proprietary knowledge that makes the difference. A lot of companies produce plastic bricks, but none come close to the quality of LEGO, because they have perfected the exact blends of plastic, analysed their behaviour for decades, have the high quality equipment ro actually make them. Even if you infiltrated the company and stole all their secrets, you would need to find professionals with the knowledge and experience to apply it, then the money to get or build the machinery to use that knowledge. By that time, LEGO already moved on.
It’s even more so for ASML. They’re on the very bleeding edge of the technology. By the time you get around to just reverse engineering their current tech and actually replicate it (which is no small task), they have IMPROVED. Hell, trying to find the experts to reverse engineer the tech could be difficult for the sole reason that they likely already work at ASML. So you’re still massively behind.
There’s also the fact that ASML doesn’t exist in a vacuum. They don’t produce all of their own components. They have equally advanced industrial partners. So now you need to replicate not just their tech, but their supply chain.
It’s essentially easier to just do it from scratch, to build up your own institutional knowledge, the professional workforce who can not just copy, but innovate. But that in itself takes years to decades.
I used to work at ASML. I can tell you the semiconductor industry is in a constant sprint race. Machines, called “tools”, take 3-5 years to design.
For context: these were $30M in the early 2000s when I was there. And they required a $10M laser, constant support equipment, and 24/7 maintenance staff (I was this).
If another company were to buy one, tear it apart and try to copy it, they would be at minimum 5-7 years behind the curve. That is why self-innovation is the only way to win here. ASML, at least at the time, took excellent care of the employees, hired top talent, and worked with AAA vendors.
To give a real world example of “just because you know how it works doesn’t mean you can replicate it”:
Take a ball of yarn, completely unwind it. Then rewind it perfectly, so it looks like a brand new one.
It’s literally just yarn, but it’s really difficult to do without the right equipment and knowhow.
In manufacturing anything, there’s a lot of proprietary knowledge that makes the difference. A lot of companies produce plastic bricks, but none come close to the quality of LEGO, because they have perfected the exact blends of plastic, analysed their behaviour for decades, have the high quality equipment ro actually make them. Even if you infiltrated the company and stole all their secrets, you would need to find professionals with the knowledge and experience to apply it, then the money to get or build the machinery to use that knowledge. By that time, LEGO already moved on.
It’s even more so for ASML. They’re on the very bleeding edge of the technology. By the time you get around to just reverse engineering their current tech and actually replicate it (which is no small task), they have IMPROVED. Hell, trying to find the experts to reverse engineer the tech could be difficult for the sole reason that they likely already work at ASML. So you’re still massively behind.
There’s also the fact that ASML doesn’t exist in a vacuum. They don’t produce all of their own components. They have equally advanced industrial partners. So now you need to replicate not just their tech, but their supply chain.
It’s essentially easier to just do it from scratch, to build up your own institutional knowledge, the professional workforce who can not just copy, but innovate. But that in itself takes years to decades.
ASML doesn’t make the entire EUV machine all on their lonesome, they are a mother of all system integrators really. Different parts and subsystems come from different industry leading suppliers and they put it all together. Its not just one company, it’s an entire industrial ecosystem and ASML spent decades building it to begin with. Eventually competitors will replicate it, but it’s going to be years or even more than a decade behind ASML.
You’re making incredibly high-precision tools. To make those, you need slightly less precise tools and components. To make those…
None of the tooling and components that go into this stuff is off-the-shelf. They’re all bespoke, there’s very few people out there making them, and those people might not be willing or able to sell them to you anyhow.
To replicate ASML’s machines, you need to assemble the whole chain of suppliers for the components that go into those machines, the tooling to make the machines, and people with the skills to make it all work.
You’re making incredibly high-precision tools. To make those, you need slightly less precise tools and components. To make those…
None of the tooling and components that go into this stuff is off-the-shelf. They’re all bespoke, there’s very few people out there making them, and those people might not be willing or able to sell them to you anyhow.
To replicate ASML’s machines, you need to assemble the whole chain of suppliers for the components that go into those machines, the tooling to make the machines, and people with the skills to make it all work.
I used to work at ASML. I can tell you the semiconductor industry is in a constant sprint race. Machines, called “tools”, take 3-5 years to design.
For context: these were $30M in the early 2000s when I was there. And they required a $10M laser, constant support equipment, and 24/7 maintenance staff (I was this).
If another company were to buy one, tear it apart and try to copy it, they would be at minimum 5-7 years behind the curve. That is why self-innovation is the only way to win here. ASML, at least at the time, took excellent care of the employees, hired top talent, and worked with AAA vendors.
It’s not difficult to remake. What is difficult is to secure sufficient parts and suppliers so that you can create enough semicon fab tools to achieve a cost effective productivity that beats existing customers.
EUV requires a great understanding of some pretty high level physics. It generates extreme UV using laser pulsed tin droplet plasma, it’s going to take a while to get that down.
You don’t just magically learn the underlying physics and understand the quirks by taking something apart.
There are lots of fiddly little quirks in real engineering that are often subtle or even completely masked by how the system was built, especially when dealing with processes in the nanometer scale. Slight variations in the tin droplets could change the distribution of the UV. A motor that vibrates a bit different because they picked a different type of bearing could result in everything being garbage.
EUV research started in the 1990s but it wasn’t until 2018 that the first machine came to be. Reverse engineering, building, qualifying, and mass producing machines within a decade would still be a 3x speed up.
You’re also left with, why make the same thing?
If you’re going to spend 10-20 years and $100B on pretty raw R&D, why make today’s system? Why not work to design the system you’re going to need in 10 years? This is also why no one is really competing with ASML for EUV machines, don’t spend billions breaking into an established market when there is a new one just around the corner you could claim all of for the same price
Latest Answers