Hacker News Reader

Hacker News

Comment by smilekzs | Hacker News Reader

Full thread

smilekzs

Background: This is all about EUVL, a technology for making chips with smaller features.

See:

https://news.ycombinator.com/item?id=11241963

https://news.ycombinator.com/item?id=12841278

View on HN

I don't know if I could stand the tension of working in a company where quarterly shipments were single digit units of product is considered a banner year :-) Makes for a really odd revenue stream. Someone builds a new fab, woo-hoo!, then 2 years of nothing.

It gets better when you consider that Intel et al. depend on them.

Heh, someone has to make the machine that makes a lot of machines.

Though, their future is unsure. How much longer will the chips change and be worth changing?

Take a look at a human brain. Or a bird's brain (they're more densely packed). CPUs are still far from those performance characteristics and their power envelopes.

So there's still room for innovation.

I'm realizing how what I thought was just a refined consumer/prosumer company for photography, is involved in many different fields of deep science. And for a while now.

It's kinda of a side business (cine lenses, too), but "refined consumer / prosumer" would be Fuji o Leica. Not many lens-makers are more "pro oriented" than CZ, maybe Rodenstock and Schneider-Kreuznach.

Zeiss still made a lot of publicity on smartphones recently and digital cameras.

Not only recently: the 2005 Nokia N90 had a Zeiss branded camera."Branded" being the key word.

Other culprits in this depressing marketing gimmick that I can recall without research: Leica, Hasselblad (!), Schneider-Kreuznach (!!). This mostly doesn't affect their actual product lines.

I'm not sure about Leica, but Fujifilm is not in the "refined consumer / prosumer" category either. Neither is Canon, Nikon, or even Olympus for that matter. AFAIK they all make high performance optics that are used in manufacturing, lithography, or machine vision in some capacity or another.

Leica (well, same parent company, I think that the structure is a bit complex) also makes microscopy equipment, but I'm talking about their lens lineup intended for photography. Leica makes very good things that are not very practical or realistically priced. Fuji makes good prosumer cameras and lenses, but they don't have wide professional adoption for ecosystem / features considerations.

Again, I think you are wrong. Fujifilm makes extremely high quality professional lenses that are used on medium and large format cameras. Or at least did fairly recently. They've also announced a new medium format digital system with lenses to match. I have the Fuji "prosumer" X series system, and I can tell you that their lenses are very, very high quality - besting most of the Nikon and Canon offerings with metal construction and in some cases weather sealing.

They stopped making those a few years ago. Their current offerings (the fuji x system) are good for people that like nice cameras but don't need to make a living with them (like me - when I shoot digitally I use a Sony A7r, but I wouldn't have picked that for actual work and could also have gone Fuji, but at the time they had only the 16mp first gen x-trans sensor), most of the lenses are good, but they don't have the coverage of the expansive Nikon or Canon ecosystems, mirrorless tracking autofocus is still lagging behind pro dslr bodies (I don't even use autofocus, but again, I don't generally have to get paid), etc. In the end a 2012 D800e outperforms a brand new X-T2 in image quality in most scenarios, despite the Fuji having better lenses (and you can always natively mount superior Otus or Apo-Sonnar lenses on the Nikon).

In a way all these companies, when they know a bit about material and physics, can diversify through the complexity spectrum. Philips is known for consumer devices but had medicine level machines and electronics spinoffs, probably military too.

as a german CS-Student it's always interesting to see german companies pop up (i have to work somewhere when this is finished). Nearly every time the stereotype is right. They are always building "pro oriented" specialised physical things, now lenses, a few days ago i stumbled across another one that builds really high performing magnets (i think, not into machine engineering) and is located not far away from where i was born. But i have jet to see the next promising Tinder/Slack/Social-Whatever from Germany. Everything is somehow related to machinery.

Each machine costs tens to hundreds of millions of dollars.

The comment was about the variability, not the cost.

I think they view themselves as an engineering services company; cumulative shipments & lifetime value are therefore more important and... meaningful.

I worked there, and it really was very stressful. The whole semicon industry is very cyclical which leads to a lot of layoffs and rehiring. Senior engineers actually advised me to quit as soon as I could.

There are many companies purchasing their machines, however you are right that they lay of a lot of people. Nowadays they mainly use temporary staffing firms for development (except for key people of course). They actually use multiple firms, in case one goes bankrupt.

Interestingly, they also applied this trick to making their machines. For example, they need a power supply, so they design one. Then they give the specifications to multiple manufacturers which build them. If one of the manufacturers goes bankrupt, they always have backups. Also if one of them does not deliver, they have backups. As you might figured out, downtime is very expensive. These manufacturing companies have started selling parts of their machines to other companies and some of them are very successful.

This is quite common in the Eindhoven area in the Netherlands where ASML is located. A lot of companies make some parts for ASML.

I'm not sure how detailed the requirements are when given to the companies but from what I understand, there still is a lot of continuous cooperation between the involved companies.

It's not that bad. ASML ships several tens of units (costing $millions to $tens of millions each) each quarter, but they also have an order backlog of several thousand units. This includes not just the super-high-end EUV machines they are in the process of developing, but also the less-than-super-high-end ("low-end" doesn't really apply to any of their machines).

Oh come on. In 2009, crisis year, the whole local Eindhoven economy was a mess. Nearly all software agencies had people fired, hardware component suppliers went bottom up.

Then some foundry (TSMC if I recall correctly) places an order for six ASML machines. Six! Big news, ASML is hiring and needs hundreds of engineers now please, pronto. Supplies get lots of that need, get back in the black numbers, and half a year later the entire city is back up and running. Everybody's hiring again, bright future hooray everything.

The entire 2008 crisis took less than a year in Eindhoven because one chip foundry ordered six machines from ASML.

The support contracts on these tools runs into huge figures, there can be 100s of engineers keeping a small number in production. Each sale is essentially a huge contract, not a one-time payment.

That is all fine if your product sits on a shelf somewhere. However, if I understand the process correctly, they run physics based simulations of all the optical components in the assembly, order the parts and assemble on-site at the customer. It's like writing a back end where you can commit only once and that commit goes straight to the production server you just bought.

Reply on news.ycombinator.com