Hacker News Reader

Comment by coolspot | Hacker News Reader

1B is like three last-gen ASML machines, good start I guess, but chip manufacturing can easily absorb 10x of that.

As I understand it, there are no plans to turn Europe into a chip leader, they just want to protect automotive supply chain.

View on HN

Complete outsider doubt: every time I hear about chip manufacturing, somebody talks about ASML machines. Considering that many of the chips we use are built by ASML machines, why isn't Netherlands a top chip manufacturer?

The skills and risks in different parts of the chain are different. Having access to excellent photo-lithography machines is not enough to guarantee having an excellent foundry or process - that much should be evident from ASML supplying TSMC and Intel, and observing Intel's struggles with 10nm.

The machines ASML produces allow you make ultra precise and tiny patterns of material. You still need to choose what those materials are, figure out how you're going to create the material (as in what dopants are you adding in what quantities), what your geometry actually is, and on and on and on.

Philips is fractured and a shadow of its former self. They used to both make the machines and the chips now its ASML and NXP

And ASMI too.

Signify is another subsidiary, the former lightning division from Philips.

Philips itself is doing very well in medical equipment. A sensible, sustainable and profitable pivot: in an age where every Chinese factory can churn out toasters or radio's with near zero margins, it makes little sense to remain worlds' leader in toasters or radio's.

Can't fault that argument. I mean Philips made and invented some nice products - CD's, LED lighting, my wake-up light - but when it comes to production they just can't keep up, and I'm confident the profit margin on electronics like that has completely tanked. Especially when their designs and inventions quickly get copied shamelessly.

Corporate espionage is still a big thing; my girlfriend told me a story about a company presenting their new device at a trade show (iirc it was a jewelry casting machine or something). The Chinese competitor that had a stand not far from them had their stand empty, but on the third day there was a replica of that same casting machine at half the price.

Medical equipment is a lot more specialized and probably harder to duplicate, and maybe more importantly, not a race to the bottom.

> in an age where every Chinese factory can churn out toasters or radio's with near zero margins, it makes little sense to remain worlds' leader in toasters or radio's.

Surprise for you, Chinese factory can't run with near zero margins.

Electronics industry been on the downsizing trend in China for the last 10 years.

It now costs less to hire labour in flyover states in USA, than South China.

Chinese factories themselves are rushing to relocate to South Asia.

> Surprise for you, Chinese factory can't run with near zero margins.

I am aware of that. But when they can "outsource" most R&D (read: copy or steal IP) a lot of businesses have a very hard time competing against those margins.

For example, when all the costs you make is plastics (raw materials), amortisation on your machines, wages and electricity, and some Ali-express fees, you can compete with a company like Lego easily. Because Lego has those same costs, but also marketing, R&D, legal, QA, distribution and so on.

I'm not saying that any of those are "wrong" per sé. Maybe Lego is in the wrong business or on a dead end, and maybe that is not bad: IDK. But I do see why many companies are pulling out of this race-to-the-bottom and either becoming "almost entirely marketing" (Nike, Adidas, etc.) or pivoting to niches with other easier to defend moats, like Philips.

Also what kind of protections do low level employees have in China? If an employee is harmed on the job, does the company have many repercussions? Can employees organize and negotiate as a group? What about healthcare costs? Or taxes? If the factory owner is in with the local government and/or the party, do they have to pay much in the way of taxes? What about public accountability? Are companies publicly listed and subject to financial disclosure rules? It’s my second and third-hand understanding that all of these factors contribute to an exceedingly low labor cost in China. I think we enjoy cheap stuff at the cost of supporting a pretty dismal situation for many ordinary Chinese.

I cannot evaluate if what you say is true. But it does not really matter when both the European Philips and the Chinese brand Pilhips, or both the Lego Harry Potter™ and "Justice Magician Bricks" are made under the same conditions. Both reap the same "benefits".

This unfair advantage would only matter if the European Philips or Lego keeps all its production (of the entire chain) in Europe, which they do not.

Did you hear about integrated circuits in school?

Did anyone offer you a course on how to design ICs in university?

I heard that Singapore NTU has 100+ graduates with PCB and IC design skills each year. If I was a chip-foundry-to-be, I'd build my factory where the employees are.

> Did anyone offer you a course on how to design ICs in university?

Yes, standard courses in Computer Engineering, at least here at Vienna's Institute of Technology, which with comp. eng. and masters in similar fields (just a slight focus shift to, e.g., the modelling/verification side) there's about the same order of magnitudes of students graduating per year too.

Chip foundries do not go where their employees graduate, but where they get gov incentives for building their fabs.

Processor design was a required class during my undergraduate, and from what I’ve heard several other universities have very similar classes.

It sounds like a chicken / egg problem. We had some in school, as part of the software engineering curriculum. The electrical engineering department probably touched on it as well.

I went to university in Sweden and they promoted a VLSI track

Given the amount of money ASML is printing, they could easily build a decent fab that will over time compete directly with eg TSMC. But I doubt they have the balls to compete against their customers, which is a shame.

(by "easily", of course, I mean "super hard but possible"). If there's any European company that knows how to source and train people who can operate fabs, it's ASML.

Doesn't have to be ASML itself, they can branch off and start a company near it though and develop close ties to it and the local university.

> But I doubt they have the balls to compete against their customers, which is a shame.

Regulators will block this.

Why? The EU very much wants to be able to make their own chips. The world is clearly moving towards each bloc wanting to be more self-sufficient.

There are other semiconductor plants in the EU.

There's no especial reason in a globalized economy for the different parts in a tooling chain to be right next to each other. Germany already makes specialized machine tooling that gets sent all over the world to be actually used, this is just a more specialized example.

The Netherlands is a very dense country with an agricultural economy, and fabs take up a lot of land and dirty a lot of water; but nonetheless, there is already a fab there, belonging to NXP (formerly Philips) https://www.nxp.com/company/about-nxp/worldwide-locations/ne...

different skillset. making equipment and running a profitable fab is like Apple and Orange.

Intel with its years of experiences in semis industry and yet Intel is not in the semis equipment business and Intel was in the very early stage of semis industry.

It's a dirty, dangerous industry. I imagine that running it in the neatherlands would be more expensive. That's one of the reasons it went abroad in the first place.

Just a supplier of certain machines for the process doesn't make you a successful foundry. I.e. ASML heavily relies on Laser technology by Trumpf Werkzeugmaschinen (a German company) - which besides Laser technology is mostly in the sheet metal machine business. Focus on what you can do best, they say.

They probably don’t have the manpower. The Asian countries produce hardware engineers at 10x that of the west, and with a much larger population, have a vast talent pool.

Isn't ASML a dutch company? How is Europe not a leader? Unless you mean in wafer production/fabs or being on the forefront of shrinking process-nodes.

Europe is behind in fabs and has historically underfunded the industry - understandable, as it had worked fine for a long time, with globalization, long supply chains and JIT inventories. Then came nativism, "Tariff Man" and a global pandemic, and Europe had to rethink its long-term strategy.

ASML is a supplier for one part of the overall process yes, but their core innovation is also somewhat limited. Their EUV technology is actually based on a license from the US DOE (from 1999: https://www.eetimes.com/u-s-gives-ok-to-asml-on-euv-effort/), as part of a consortium of companies that were allowed to benefit from early US R&D into EUV. This is also why the US is able to require ASML to not export EUV machines to China. In my opinion from limited knowledge, being a leader means also being able to change the industry in significant ways. You could argue ASML has contributed to that but I think their dependence on US research indicates they aren’t independently set up to be a leader in this space.

>U.S. Undersecretary of Energy Ernest Moniz said, “if the EUV technology proves viable, ASML has agreed to build a factory in the U.S., similar to its Netherlands facility, as well as to establish an American research and development center. The factory will supply 100 percent of all ASML's sales in the United States.”

Did this actually happen? Does Intel buy ASML machines?

I am not actually sure, but ASML does have a US R&D and manufacturing presence. Their largest site is in CT (https://www.asml.com/en/company/about-asml/locations/wilton-...). As for Intel, they bought a 15% stake in ASML in 2012 (https://www.reuters.com/article/us-intel-asml-idUSBRE86819B2...) and ASML is have suspected to have sold machines to Intel subsequently (https://www.kitguru.net/components/graphic-cards/anton-shilo...). However, Intel may have not needed as many machines because they were able to avoid depending on EUV for some nodes (https://www.reuters.com/article/asml-intel-idUSL8N1AY2H4).

How can a patent from 1999 still provide any leverage? Isn't the lifetime of US patents (or patents in general in most countries) limited to 20 years, so this patent should have expired in 2019?

I can imagine the leverage is 20 years of refining their technology. Sure, other companies could have invested in developing the technology in the last 5 years, and start selling or using it in 2019, but I assume it takes a lot of refinement to have this work nicely.

I'm no expert, but I'd say that patent is only one of many patented inventions needed to turn it into a working product. That is, say the patent expired, anyone could make an EUV laser, but that in itself does not produce nanometer scale transistors.

> Isn't ASML a dutch company? How is Europe not a leader?

The same reasons why the US doesn't have much presence any more:

1) Environment protection and emission laws: just look at the Silicon Valley, half of it is a Superfund site from decades of silicon manufacturing.

2) Lack of skilled and experienced (!) engineers

3) For Europe: lack of access to the many billions of dollars that a fab build requires

4) Cost: No matter what you do, simply alone due to wages, real estate and raw material cost, US and European-made components will always be more expensive than Asia-made. The military can afford it to specify "homemade" as requirement, but everyone else will be going with whoever is the cheapest option. Yes we have a bit of a supply stretch at the moment ("thanks" to shitcoin miners and fuck-ups by automotive), but this will even out sooner or later and the prices return to their old normal level, which means a domestic production won't be cost-effective any more, which in turn means it's hard to get funding for such a project from banks.

The only way out is massive government investments and subsidies.

Your phrasing "fuck-ups by automotive" brought to mind something: if automotive didn't fuck-up, would that have reduced the shortages ? In the automotive sector, yes, of course, but not the general shortages, surely ? Would we'd just have shortages elsewhere, notably in home electronics, which is the industry that picked up all the spare production capacity that automotive freed.

Is one better than the other ?

The problem is the ripple effect that automotive has across societies. It's one thing if shitcoin miners or gamers have to wait for GPUs, but when the automotive industry with its many millions of employees and in sequence its supply chain with even more millions of employees is grinding to a halt, the impact is immense.

If anything, corona has shown us how extremely unhealthy the dependency of modern society on the automotive industry is. We need to wean ourselves off of that, and the sooner we begin the better - climate change is the next catastrophe waiting in the starting block.

There is more that goes inside a Fab than a litho tool. Granted it’s a major one, but a fabs has 400 process steps.

See KLA Tencor, Applied Materials, TEL, Advantest, ASM, Hirata, Fujitsu, Mitsubishi, etc probably dozens more.

Yes, and "dozens more" needs to be changed to hundreds more, with most not being known even to industry insiders. Lots of them are single vendor globally, and are Taiwanese.

For example, FOUP cleaning machines are made only by Gudeng (as well as ultra low particle FOUPs themselves,) and don't think that "cleaning machine" is any much less sophisticated than other equipment. It's the size of a truck, made of space age materials to survive fluorines, and costs many megabucks.

I want to add other reason: It's astronomically expensive and really slow to grow.

TSMC started in 1987, and AFAIK, the government of Taiwan invest in the company to help their grow.

EU started some years ago, investing in R&D in this sector, some years ago. We need to wait at least a decade to see a real impact starting to grow with this initiatives.

ASML and ARM. Even if ARM is now owned by SoftBank, it’s still HQ’d in Europe, right? Combine the best lithography machines and the leading architecture and Europe probably should be dominant. Interesting that they choose instead to export so much of the value chain around the world.

I'm not an expert so this is entirely armchair reasoning, but I can think of a few factors.

We don't have a culture of crazy investors, nor one of hypergrowth - I mean how many US startups went from nothing to a $1B valuation or takeover in a year, maybe two after they first started making waves?

Wages are not as crazy. Neither is cost of living, but it does mean the most talented people will pursue a career where the pay matches. Anyway when you consider things like income taxes going to a government that doesn't spend more on the military than any other country, socialist policies, and sane health care, it does even out a lot.

It's not as much of a union as they want to make believe. Anything that grows to a certain size will deal with a lot of international headaches. Freedom of movement and trade has made things a lot easier (something the UK wasn't aware of apparently, probably still stuck in English exceptionalism, and they're getting fucked over left and right because of it at the moment).

I mean one project going on right now is "they" are trying to build a EU focused AWS / GCE competitor, but instead of empowering or investing in a single company, or letting the market work for a company to rise in the ranks, they're making it some big international semi-government affair, where each country is fighting hard to protect and promote their interests. It's turned into not a technical or business challenge, but a political one. I mean I'm a complete fuckwit myself but give me a few billion in spending money and absolute power and I'll build you a secure EU cloud provider.

Random counterpoints: how many US hardware startups managed to have hypergrowth. How many US hardware startups are involved in chip fabrication, period?

Taiwan doesn't have either crazy investors or hypergrowth. I suspect the chip hardware business requires many years of steady-but-compounding growth.

As for the AWS/GCE competitor: I suspect the policy goal is not to have one platform/product in X years that will soon be outdated. Rather, it's to nurture and grow a competitive domestic ecosystem that is self-sufficient. That way, you may have dozens of AWSes in thr future that keep you on the leading edge (not just inplementors)

This factory won't be purchasing last gen stuff, more like 10 year old gen machines (I heard 25ish nm machines).

Other sources[1] say its 65nm, and the chips won't even be used for automotive purposes.

[1]: https://www.electronicsweekly.com/news/business/bosch-opens-...

Here's a source from Bosch themselves saying "up to 65 nm", which I assume means 65 nm as the smallest feature. They're making "Application-specific integrated circuits (ASICs), and power semiconductors" mainly for the automotive industry, so they're probably counting on that being enough. And what a great time to start production.

https://www.bosch-presse.de/pressportal/de/en/300-mm-wafer-f...

the text says otherwise regarding automotive: "with output of automotive chips to follow from September" (so, you're correct for now until september, but sep. is soon already I would say)

They'll probably making chips for Bosch parts. IIUC they don't do much advanced computational stuff.

No ones using last-gen for automotive.

Tesla is using Navi 23 RDNA in their new S and X models. That's 7nm chip. NN accelerators will likely move to 16nm or better soon. However, you're right that most automotive chips and power electronics use 180 or 65nm process.

EDIT: Mobile Eye, EyeQ5 ADAS SoC is already 7nm FinFET.

Maybe for computational stuff but older/bigger process nodes are still very capable (and possibly better) for power related things.

I was just establishing frame of reference for the investment amout.

1B is barely news-worthy.

Most cars still default to incandescent bulbs for many uses. Few cars attempt to understand the situation around them, beyond a rear camera and maybe ultrasonic reverse sensor for parking. Engine control units have gotten more complicated, but unless high end or EV, It’s mostly mid 90s tech. Disagree?

I dare you find an incandescent on any car designed in the last five years. It’s insanely inefficient.

I think he's actually talking about halogen bulbs.

And at least in Europe, outlawed since about 10 years

they can easily do that. that's why I also think Intel will turn things around with it's fabs. You don't need to be manufacturing the latest node chips. Just a couple generations old, will do the trick for automotive, defense and non fashionable electronics.

also somehow people forget chips these days are really powerful.

> also somehow people forget chips these days are really powerful.

Only powerful ;-) ? Just found out some FPGAs have embedded processor cores with them (no need to generate them like in 2000) just because their customers already need that and they have enough gates to use.

Other recent stuff like in the RPI2040 microcontroller is a way to code your own "independent" IO sub-programs that run in parallel by a dedicated procesor that does not interfere in your code.

Hardware never stops to amaze me.

Reply on news.ycombinator.com