Ex-ASML worker accused of stealing chipmaking secrets for China is Huawei to a new job A former ASML employee accused of stealing trade secrets about advanced chipmaking equipment may now be employed by Huawei. Citing several unnamed sources in and around ASML, Dutch newspaper NRC reports that following the employee's departure they have worked at Huawei. The alleged theft was revealed earlier this year as part …
Not saying what the person is accused of is right, but I am saying I have zero sympathy for ASML. I once interviewed for them, and literally one of the interviewers says to me, "I don't know how well you'll fit in here because you're not Chinese." Then they just left me sitting in a conference room for like an hour before I finally left on my own. So, ASML can eat shit and die for all I care.
This isn't even the first time this has happened to them. There was the whole Xtal case some years back where a bunch of employees left, taking a bunch of proprietary info with them, to form their own competing company. If this sort of thing keeps happening, at some point you have to wonder if maybe ASML is asleep at the switch. If it happens one time... shit happens, learn from your mistakes and move on. If it happens a second, or even third, time... now you have to wonder if anyone bothered to learn from mistakes made the first time.
Unless you were applying to join a very specific Customer Support group catering exclusively to Chinese customers and consisting exclusively of Chinese employees, that seems like an EXTREMELY weird and out of place statement. Only a tiny fraction of the people I've interacted with at ASML (from engineering to CS and from Logistics to management) are Chinese. So I don't get why that would even be a concern. I'd certainly have wagered an email to ASML HR because they'd certainly have a few words for that particular manager. (No I don't work for ASML, though I could probably get hired there tomorrow if I wanted to and I'm not Chinese either)
ASML's EUV scanners are some of the most complex machines ever built. ASML depends on dozens of subcontractors to make unique parts to mind blowingly exact specs, stuff like mirrors that reflect a sufficient amount of EUV light to be useful that required many years of research to develop. And those subcontractors have their own array of sub-sub-contractors, there are thousands of companies that contribute in some way to manufacturing a single EUV scanner.
I've said it before but I'll say it again. Even if China had the complete set of blueprints for ASML's EUV scanners AND had a working model to disassemble and reverse engineer they probably couldn't replicate it in less than a decade. It took a quarter century of research to make it happen the first time, with much of the western world contributing. Even with that huge head start China would still have to figure out how to make all the parts that go into it. A blueprint showing where part 'X' goes does you no good if you don't have the blueprints/specs for part 'X' and cannot buy one anywhere since only one company makes them and they will only sell them to ASML.
This is like if someone on NASA's Apollo project had defected to the USSR after Apollo 11's successful landing. He would have been just a tiny cog in a huge machine who might understand some portion of the Saturn V rocket or the lunar module or whatever but would hardly have made in a difference in Russia's space program.
China is much more likely to figure out a DIFFERENT way of achieving lithography beyond DUV (or a way that doesn't use lithography like if Canon's nanoimprint stuff can be made to fully fill that role) than they are to clone ASML's current tech.
AND had a working model to disassemble and reverse engineer
This is what happened to somebody I know. They had designed , and manufacture a bit of kit to be used in the mining industry.
About a year later they kept getting enquiries why part x was continually failing. Odd because they had never sold that kit to the companies making the enquiries. It seemed that a Chinese competitor somehow acquired a 2nd hand model, and reverse engineered it, including the wear on part x which was causing it to fail.
It still saves a lot of R&D effort in not going down the wrong paths and knowing which bits need to be built to what tolerance
You can't run our source code without having our proprietry interface cards - but we still aren't open sourcing our algorithms !
It still saves a lot of R&D effort in not going down the wrong paths
Perhaps it does, perhaps it doesn't.
For example, ASML's EUV scanners create their EUV light using a pulsed X ray laser to vaporize tin droplets at a rate of about 50 droplets per second. Free electron lasers became able to generate continuous EUV light at the same/similar frequencies and at the (rather high) required power levels in the middle of the previous decade, which was a few years before ASML's first EUV scanners were delivered, but ASML was far enough along their existing path they didn't want to change course. They'd worked out all the issues with the tin droplet method, but had no idea what obstacles they might run into using free electron lasers.
Using free electron lasers would be vastly cheaper since the particle accelerator (yes, like CERN's, but not as big) it requires uses far less power and one accelerator could be shared by all the EUV scanners on a foundry "campus" and power is one of the largest if not the dominant cost in modern chip production (the upcoming next generation "high NA" EUV scanners that will be delivered to fabs starting next year require something like 4.4 MW) so that would probably be the route China would want to take all things being equal. But if they had ASML's blueprints they might follow the tin droplet plan because so much else would have to be changed going from tin droplets to free electron lasers that they'd essentially be starting from scratch anyway. So going the free electron laser route would mean they'd have to do all their own research and follow all their own blind alleys just like ASML did, but if they trod the same path as ASML they would end up with a solution that's much more expensive to operate.
One of the problems that ASML would have in using a single FEL for delivering EUV light from a single source to multiple scanners is that it would involve even more mirrors at an even higher power level (undesirable from an efficiency, longevity and wavefront uniformity standpoint) or have a beam splitter to split the high energy electron beam to different systems and a separate exciter section for each system (complex, requires more space). The engineering issues with using a FEL were such that even if they weren't as close to figuring out Laser Produced Plasma light sources they'd still have stuck to using those. Not to mention the human safety risks involved in running an extremely high power relativistic particle beam system in a production environment. FELs were seriously considered by ASML, but rejected based on a multitude of reasons, not only because the LPP source was the "known" option (I've seen up close just how much was still unknown at the time about the LPP source and it's function. It seemed to me at the time if it worked most of the time it was through luck and deep dark magic, not knowledge.
I think everyone assumes it would use a beam splitter and separate exciters. The human safety concerns are real, but there are already more than a few ways a person not following all the safety procedures could die in a modern fab.
One downside would be financial pressure to put more eggs in one basket, as the cost of setting up a new site goes up due to the additional land and construction cost to install an accelerator. So you'd much rather continually expand a campus that already has an accelerator, and if something happens there you'd knock a lot more production offline at once.
Thinking back about my talks with experts I now recall that the biggest issue with a "one source to rule them all" was that if that source goes down, you suddenly have a large multitude of systems down. Given the cost per system of an EUV system they basically need near constant uptime to be economical. The cost of downtime of let's say 10 systems simultaneously would be so tremendous the source would basically have to be guaranteed to such a ridiculous level of uptime it became nearly impossible and this made customers extremely skittish of this approach. A separate LPP source per system probably results in more downtime per system but that downtime is completely independent of every other scanner.
CERN runs for months at a time, and they only shut it down when they want to upgrade part of the system. That's a far far more complex accelerator than the fabs would require.
Anyway supposedly the big attraction of this type of setup would be that it costs less to operate, so as long as the downtime was planned maintenance rather than "oops the particle accelerator went offline again go send Gary to see what broke this time" they'd be able to tolerate it. They might not like it, but if it produces wafers for less cost (both because it uses much less power and because it could run more wafers per hour since the light output is stronger) it would still be an improvement over the current scanners.
The unintended consequence of the sanctions is highly likely to be that China develops BETTER ways of doing what ASML does and sells it, putting ASML out of business
This kind of thing has happened before as the result of trade sanctions. If you're merely copying, you'll always lag behind but if you have to innovate, you may hit upon better methods
or in other words, ASML has proven EUV can be done, but knowing it's possible is all it takes for someone else to replicate the results even if they do so using different methods - which is why dolphins and sharks look pretty similar despite the internal differences
I wouldn't put a time scale on China being able to reverse engineer, they have a lot of clever engineers, case in point, 1945 UK supplied the Soviets with Rolls-Royce Nene engine and blueprints for civilian use, within a year they had a copy in the Mig-15 wiping the floor with western jets for speed and manoeuvrability.
If they had the knowledge and ability at the required level for producing multi-layer mirror optics, stacked multi-axis nano-meter accurate positioning systems and laser optical systems, then maybe they'd be able to copy things that fast. For copying the jet engine the Soviets had basically all knowledge already there, they just needed a push to figure out how to combine everything together to make a more efficient jet engine. For something like an ASML EUV photolitho system, China currently doesn't even have the institutional knowledge and production capacity for the disparate parts, let alone combining them into a functional system.
Yes, they'd benefit from knowing which direction to go a bit, but it'll still take them probably over a decade to get there.
What about reliability or any other quality factors or range ?
Remember the foxbat which could go mach 3, but only for a bit and it destroyed its engines..
Thats the problem with Russian/Soviet equipment the full picture isnt really that great, aka look at Ukraine.
Soviet/Russian doctrine was that the Foxbat only NEEDED to do that sprint once, so that's what it was designed to do
Similarly the other equipment is essentially disposable because the calculation is that a wartime lifespan of most kit is 2 weeks. so why design it to go much further than that?
Wiping ?
Reminds me of the concordski, the engines in the beginning when pax flights began only lasted 50 hours and had to be replaced. THis is why it was shutdown, nobody can run an airline when the engines barely last a few weeks.
Yup China like Russia is all about appearance and not about reality.
I agree with you about the complexity. There is no one around who really is understanding the machine as a whole, but.....
Your argument is written time and time again, and yet there is news that the Chinese have their laser to go under 7nm. This is not invented from the stars, but done by (partially) stealing IP from other companies, like ASML. (Let us see if their laser also works with tin droplets)
I do not agree with the sanctions that Europe, the US and Japan have imposed on China not sell certain equipment to China. But the real problem is that the Chinese don't keep any agreements. They have a wrong way of doing business, and we can't cope with that. This is where the *real* problem is. We should build and practice a worldwide punishment system if they do that.
From a technical point I do not agree with you at all: This is Western arrogance (I am sorry) and it is just a matter of time. The only real argument is time indeed, since you are going back in time during reverse engineering while ASML is going forward. But don't forget that ASML does scrape our globe for capable technicians, and in China they have lots of people available. With some math one can imagine that the amount of brilliant capable people does finish school each year, the same amount as working for ASML. The laser they have now proves this. So it is a matter of priority for them to shorten this time. And now with these sanctions we have given them *motivation* wow....
China is using ASML's improved DUV machines to do 7nm (and it is still disputed if that chip is really 7nm we'll have to wait to find out)
TSMC did 7nm using DUV only with previous generation DUV machines, so it was always known it would be possible. The improved ones might be able to extend a little beyond 5nm but that's the limit without EUV.
"They have a wrong way of doing business, and we can't cope with that"
I _very_ strongly urge you to look at the history of American government sanctioned intellectual property theft and ignoring patents/copyrights issued in other countries
The USA only started paying lip service to copyright and patents when it became to their advantage to do so
Some starting points:
https://www.history.com/news/industrial-revolution-spies-europe
https://www.realclearmarkets.com/articles/2018/07/30/ip_theft_is_what_once_helped_make_america_great_103367.html
https://apnews.com/general-news-b40414d22f2248428ce11ff36b88dc53
I foresee China cloning ASML's DUV machines in a couple of years, but EUV will remain out of reach for a long while.
Anyway, DUV will probably be good enough for them for the near future, especially if they can improve it further. Western companies have no incentive to do so since EUV can be purchased here freely.
If by "purchased freely" you mean "buying a thing about the size of a bus that costs many millions of dollars and requires a dedicated team of ASML employees to keep it running" then you may have a point. These technologies, though, aren't just the product of one company (as has been pointed out elsewhere), they're a collaborative effort that involves a lot of different companies sited all over the world. In fact from what I've read instead of being a company like ASML producing a machine that everyone then rushes to buy its more like the industry decides which technologies it needs to use for such and such a generation of fabrication and a company like ASML gets the lithography job, just one piece of the overall puzzle.
Building silicon is an international activity. Trying to pretend its something that belongs to, and so controlled, by one country is futile. Thirty years ago, probably. Twenty years ago, maybe. Today, not a chance.
The Americans have definitely forgotten the commercial espionage which kickstarted their industrial revolution. Quite a bit of "American technology" started out elsewhere before being stolen or acquired under less than honest circumstances (Jets, Magnetrons, Niuclear weapons being recent examples)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
