Business Breakdowns
· Tom Walsh
The research into EUV was developed by the Americans. They'd seen their semiconductor industry decimated by the Japanese. They developed through state funding, Department of Energy and DARPA put a lot of money into this in a consortium with industry partners from across the U.S. to fund the development of the technology. But they didn't want to give it to the Japanese, and they didn't have a domestic lithography maker that was going to be capable of taking it forward. So they invited ASML to join the consortium and take that technology on. And that was a really massive step, I think in the history of ASML and of the industry as a whole. And at that time the plan was to introduce it in the mid-2000s. They thought they could get this up and running, they thought they would have to get it up and running to be making semiconductors and continuing Moore's Law from 2004, 2006 onwards. Obviously, it was late, by 2010 it still wasn't there. 2012, ASML managed to persuade Intel, Samsung, TSMC to support them to co-invest in the company, take 23% share between them and ASML. They've got EUR 1.4 billion of R&D funding into ASML. And finally, in 2019, it will take 13, 14 years after it was supposed to be delivered, these first EUV machines came out. And it's all just about being able to make the patterns smaller, make them more efficiently and therefore make them at lower cost, but also to make them better. They're higher quality patterns because you're not having to use lots of tricks to try and get a really wide bandwidth of light down to a very, very tiny design size.
· Matt Reustle, Tom Walsh
[00:00:00] S0: That's incredible in terms of the storyline for how they were trusted with developing this technology. Did the Americans put all their chips in the ASML basket? Were there any other businesses that they trusted to try to develop this? Obviously, none of the companies within Japan. But was there anybody else that they look to or hope would develop this? [00:00:00] S1: So the idea of the consortium that was set up in the late ‘90s was just to research and produce fundamental research into the potential of extreme ultraviolet for that light technology to be available to anyone who wanted. So had there been a domestic U.S. player that was strong enough to carry it on, then they could have done so. The Japanese set up their own consortium and coordinate and collaborated there to try and develop extreme ultraviolet, and they continued to progress. And around 2007, Nikon had a prototype of an EUV machine. But the sheer cost of developing it and the sheer technological challenge meant that ultimately first Canon and then Nikon gave up on it. They had to give up. They couldn't get it to work. They didn't think it would work. And yet only ASML had the funds, the resources, and the support in order to push through and bring this to market. [00:00:00] S0: That's incredible. And where they stand today, is there anyone else relatively in the ballpark of developing a competitive machine?
· Tom Walsh
[00:00:00] S1: The short answer is no. It's taken from the first time they discussed it, ASML through to today is 25 years plus. They've sunk over EUR 10 billion of R&D into the development of it. The components that go into it are not sort of things you buy off the shelf, they're things that had to be created. For this, you need to use a light source that I described. The transition from light bulb to laser. The light source for an EUV machine involves shooting a laser at droplets of tin that are smaller than a dust particle. And you have to strike each dust particle twice. Once to flatten it, once to vaporize it, to turn it into plasma, which is 40 times hotter than the surface of the sun. And you have to do that 50,000 times a second. As these things are shot through a vacuum. It's not the light source you find on the shelf. It's so difficult to make about a decade ago, ASML had to buy in the light source provider, a company called Cymer in San Diego. So no one else could make that light source, never mind the mirrors that are used to focus the light, nor all the other components that go into an EUV machines. I don't see how anyone would catch up. Maybe in 10 years you might want to come up with an approximation of the current EUV machine, but by that point you're onto the multiple generations forward of EUV. So of all the competitive advantages, of all the technological advantages that I've ever come across, I can't think of one that's more significant than ASML's.
· Matt Reustle, Tom Walsh
[00:00:00] S0: There are some barriers to entry to say the least. In terms of the actual value chain and where they fit in, you mentioned the cost of one of these machines. Do you have a sense of what this represents in terms of the overall production cost of a semiconductor and how much this particular step in the process, this machine, how much that represents of the overall pie? [00:00:00] S1: Yes. So the way they tend to look at it is their share of wafer fabrication equipment. So what we know is roughly what it costs to build a new wafer fabrication site and we know what the equipment share of that is. And we can see there's industry stats produced. ASML tends to account for somewhere between 20% and 25% of that. It varies according to the type of semiconductor. So it's much higher in logic chips, which are the chips that Intel make. It's lower in parts of the memory market where they've got a slightly different technology part. [00:00:00] S0: In terms of what they've been able to unlock for customers with introducing this technology, obviously, things have gotten smaller. What have been the outcomes of that in terms of tangible results that customers have seen?
· Tom Walsh
[00:00:00] S1: In simple terms, what they've been able to do is enable the continuation of Moore's Law. So as I think other people have spoken to you about in the past, the nodes as we describe them now, 5-nanometers, 3-nanometes, they don't exactly correlate to the size. Once upon a time, these were actually the size of the smallest dimension on the surface of a chip. That's no longer exactly the case. They're slightly marketing terms. But the fact of the matter is that there is still this process of squeezing more transistors and smaller transistors onto a piece of silicon. The more transistors you have, the more calculations you can perform, the more memory you have on that single chip. So every time they're able to advance the manufacturing process that enables the production of more advanced semiconductors at the same price, which enables your iPhone to get that little bit better, which enables the processing chips made by NVIDIA to do that more complex an algorithm that it enables the things like ChatGPT and artificial intelligence to be done. So in terms of pinpointing single things, there's lots of single things you could point to, but essentially it's the whole of computing progress. Anything that's happening at the leading edge needs advanced manufacturing, needs photolithography to keep progressing, so they can keep squeezing more transistors onto a chip.
· Matt Reustle
[00:00:00] S0: Jumping a bit into the business model itself when a company buys one of these machines, what's the typical life cycle like? Is this something that -- and it's incredibly huge, as you mentioned, in terms of requiring three jumbo aircrafts to get it over. How long is that in operation for a customer?