So the fire damaged the production of EUV components at a time when everyone and their dog want to get a fab up and running? How many of these planned fabs rely on EUV? Or did they actually use EUV in that building?
It will also of course hinder any of their would-be competitors from getting their EUV plants up and running.
Also, "affecting almost nothing else" is only true if nobody buys the chips the TSMC and Samsung make with their EUV plants. It's probably more the opposite where almost everyone who buys any sort of product with Samsung and/or TSMC chips in it (ie, pretty close to everyone) will be affected.
– DUV: Deep ultraviolet, a wavelength range in the far ultraviolet. Chip production uses 248 and 193 nanometres.
– EUV: Extreme ultraviolet, the wavelength range between roughly 100 and 10 nanometres. In chip manufacture, used as an abbreviation for EUV lithography (also abbreviated EUVL), that is, lithography with light at a wavelength of 13.5 nanometres. EUV is expected to be used in production before 2020.
EUV is Extreme Ultra Violet and is the highest end manufacturing process for silicon. DUV is Deep Ultra Violet and is used for larger node silicon. Only one company makes EUV machines and that's ASML. I think I read somewhere that ASML make about 50 EUV machines per year for companies such as TSMC, Samsung and Intel. This fire obviously disrupts production of those machines but I'm not sure if they have another factory elsewhere.
Given ASML basically have a 100% market share in EUV, 95% market share in ArF tools, 20% in Dry tools.
I would say anyone that placed an order for EUV kit recently probably placed it with ASML....
(As per my comment on other HN thread, the "problem" is that ASML tech - EUV in particular - is lightyears ahead of any competition. AFAIK no viable competition is expected until the next decade, i.e. 2030s).
Hard to say how big a deal this is from the outside. Could be major.
Anything that delays EUV production is a big deal to everyone doing leading edge chips - there is a massive backlog on demand for EUV capable manufacturing that will be a bottleneck for volume production for everyone. I thought TSMC maybe had most of the existing machines but will need more to keep up with demand - Intel and Samsung will be wanting as many as they can get their hands on too.
The worry with this type of extremely specialist equipment manufacture is that throwing huge amounts of money (in addition to billions already spent) likely won’t resolve the recovery any quicker.
Seems maybe the industry bottleneck is not the ability to manufacture chips - it’s one level up - the single source production of facilities that make the machines that make thr chips
They were planning on having their EUV line for Intel 4 (nee Intel 7nm) come online this year for risk production. This has the potential to put the nail in the coffin for their foundry side.
The production affected seems to be the clamp that holds a wafer, which seems pretty crucial to getting positioning right though the rest of the process. If they aren't able to resolve the production issues, or farm out the work somehow, we're likely to see more constraint of high end chip production for years.
Imagine trying to machine parts on a milling machine without a Kurt vise to hold it in place. Machinists normally work to 100 or 10 micron precision. In EUV chip making, position has to be accurate to 0.001 microns in all dimensions. It can't be done without these clamps.
The market isn’t pricing in any production issues right now, maybe it’s waiting on Jan 19 for Q4 earnings and an update of how the fire affected production.
We've put a lot of focus on asking foundries to diversify to remove risk, it sounds like ASML should be doing the same thing now. Monopolies are dangerous technology wise - and this has the possibility of significantly damaging a number of international companies and states.
Is there a reason - other then (anti-) nationalistic that you think that fixing the critical dependency that EUV presents involves breaking apart any company (including ASML?)
Waiting for a slightly larger market correction on ASML so I can dump a lot of money on their stock. The news of the fire barely affected the price. Bought some time ago and they are such a unique company that I don't think anyone can catch up with them soon.
In this case, the fire is making more investors aware that they're the only supplier in the world for some of their products. I wouldn't hold my breath waiting for a contraction.
Has there been any report on the cause of the fire?
I assume there is nothing untoward.
But I also immediately thought, huh, there's a thriller plot version of this, where a very smart version of say Earth First! determines that the most effective way to retard say global commerce and hence climate impact, with minimal resources,
is to do a lot of homework inspecting networks of dependency, to find the highest pay-off minimal-risk intervention.
Twist: apply a tangle of contemporary tools to walk implicit graphs to mine candidates to evaluate. When you need to scale up, quietly compromise cloudhosting and run some polite little processes for you. You're not in a hurry. You're not destructive. You clean up after yourself. You just borrow some CPU and bandwidth for a while, phone home, and hey patch your host on the way out. Because you're light gray hats. Sort of.
DMs open for book contracts. No agents, publishers only pls thx.
34 comments
[ 2.9 ms ] story [ 57.7 ms ] threadAlso, "affecting almost nothing else" is only true if nobody buys the chips the TSMC and Samsung make with their EUV plants. It's probably more the opposite where almost everyone who buys any sort of product with Samsung and/or TSMC chips in it (ie, pretty close to everyone) will be affected.
– DUV: Deep ultraviolet, a wavelength range in the far ultraviolet. Chip production uses 248 and 193 nanometres.
– EUV: Extreme ultraviolet, the wavelength range between roughly 100 and 10 nanometres. In chip manufacture, used as an abbreviation for EUV lithography (also abbreviated EUVL), that is, lithography with light at a wavelength of 13.5 nanometres. EUV is expected to be used in production before 2020.
https://www.youtube.com/watch?v=f0gMdGrVteI
Two good introduction videos.
I would say anyone that placed an order for EUV kit recently probably placed it with ASML....
(As per my comment on other HN thread, the "problem" is that ASML tech - EUV in particular - is lightyears ahead of any competition. AFAIK no viable competition is expected until the next decade, i.e. 2030s).
Anything that delays EUV production is a big deal to everyone doing leading edge chips - there is a massive backlog on demand for EUV capable manufacturing that will be a bottleneck for volume production for everyone. I thought TSMC maybe had most of the existing machines but will need more to keep up with demand - Intel and Samsung will be wanting as many as they can get their hands on too.
The worry with this type of extremely specialist equipment manufacture is that throwing huge amounts of money (in addition to billions already spent) likely won’t resolve the recovery any quicker.
Seems maybe the industry bottleneck is not the ability to manufacture chips - it’s one level up - the single source production of facilities that make the machines that make thr chips
They were planning on having their EUV line for Intel 4 (nee Intel 7nm) come online this year for risk production. This has the potential to put the nail in the coffin for their foundry side.
Imagine trying to machine parts on a milling machine without a Kurt vise to hold it in place. Machinists normally work to 100 or 10 micron precision. In EUV chip making, position has to be accurate to 0.001 microns in all dimensions. It can't be done without these clamps.
I assume there is nothing untoward.
But I also immediately thought, huh, there's a thriller plot version of this, where a very smart version of say Earth First! determines that the most effective way to retard say global commerce and hence climate impact, with minimal resources,
is to do a lot of homework inspecting networks of dependency, to find the highest pay-off minimal-risk intervention.
Twist: apply a tangle of contemporary tools to walk implicit graphs to mine candidates to evaluate. When you need to scale up, quietly compromise cloudhosting and run some polite little processes for you. You're not in a hurry. You're not destructive. You clean up after yourself. You just borrow some CPU and bandwidth for a while, phone home, and hey patch your host on the way out. Because you're light gray hats. Sort of.
DMs open for book contracts. No agents, publishers only pls thx.