We got to a .8 micrometer error margin, which is "an enormous valley" by ASML standards but still fine for the top ~5 layers of normal semicon designs, where there's typically way fewer lanes, placed more widely apart. Again, I don't know much about chip design so I probably got all the details wrong.
Basically, yes there's a market. It's not at all as big as ASML's but it doesn't need to be. There's also a market of refurbishing old ASML/Canon/Nikon machines and repurposing for different parts of the production process (either chips or other semicon stuff like LEDs), which is what a new litho startup would compete with I think.
Note that all my knowledge about this is 5 years outdated, I now run a chat api startup :-)
To me, the semiconductor fabrication world is like an opaque industry where you need a significant amount of capital to do anything useful. Very happy to learn that startups are getting a chance too. Super curious to know if the company made it.
I know it's easy to say "we did great but the HQ fucked it up" so I don't want to make that claim. I honestly don't know where it went wrong (and I was way less interested in sales and marketing and entrepreneurship then, so I didn't ask).
I learned everything I know about lean and agile and just shipping from our R&D lead though. He was ex-ASML and they definitely know how to ship.
I know quite some people at ASML and I worked a bit in lithography in the past. Yes, it's extremely hard to make a machine that can compete with ASML's. But given enough smart people, I'm convinced that it's totally possible. Making a machine that can compete with what ASML could do, say, 20 years ago is a matter of 10 people and 2 years (I did some contracting at a company that did just that). If you start there and then just keep adding the right kind of talent, then I bet in 10-15 years you can get impressively close.
Don't underestimate the enormous amount of technical and engineering debt ASML has accumulated through decades of mad deadline pushing. Nobody has time to invest in eg automated tests or anything that lets you guard code quality because all the good engineers are needed on the critical path to make the next deadline, and there's always a next deadline and it's always 3 months sooner than in any way reasonable. This is how they became a world monopoly, effectively. It's been a good strategy. They ship so early that the machines they sell simply don't work yet, so they put two engineers inside the box with it who (together with 50 more engineers back home) make the thing actually work while the customer sets it up in their fab.
As a result, their codebase is a steaming pile of shit. Copy & paste galore. An example: the code is so bad that you can't just fix bugs. If you find an off-by-one error in the code and you want to fix it, you first write a "risk assessment document" where you describe what the bug does and whether any code could potentially rely on the presence of the bug. Then multiple layers of risk assessment officers assess whether you get a go. Then, if you get a go, you can go and fix your off-by-one error in similar-but-not-the-same C files across 7 product families.
To be honest I don't know if the same holds for ASMLs mechanical engineering, physics, electro etc departments, but their software dept (a large percentage of their entire R&D) is easily 10x less productive than similarly sized and skilled software teams elsewhere. It's really mind boggling.
I'm convinced that a good startup with a good mindset, plus sufficiently good funding, can totally build a competitor. China just has to fund 20 litho startups so that 19 can mess it up and 1 makes it big.