Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Germany paying $5.5B for Intel fab

electronicsweekly.com

Post reply on HN
Print view View original

261–270 of 593 posts

Re: Germany paying $5.5B for Intel fab

#261

Post by kken »

riedel wrote: post #189
Sakos wrote: post #145

Earlier quoted context omitted.

Ideally, the EU would be trying to build its own semiconductor industry. I don't see that kind of foresight in any of our political leaders, much less mine in Germany. So the next best thing is to ensure that the current company we're reliant on builds local production.

Germany used to have a semiconductor industry like Intermetall (acquired by micronas acquired by TDK). The problem is that IMHO German automotive industry is not willing to pay a cent more than necessary. They do only care about short term profit. They triggered the chip crisis in Germany by stopping there orders and then restocking. I am not sure if any government can fix this.

>Germany used to have a semiconductor industry like Intermetall (acquired by micronas acquired by TDK). The problem is that IMHO German automotive industry is not willing to pay a cent more than necessary.

How about the three fabs that Bosch has in Germany? Or Infineon, X-Fab, Globalfoundries, TI, Prema, Elmos (now Siltech), Vishay, Nexperia?

Re: Germany paying $5.5B for Intel fab

#262

Post by audunw »

ChuckNorris89 wrote: post #21
Tade0 wrote: post #12

Earlier quoted context omitted.

We are - indirectly. Intel uses EUV machines provided by ASML.

The ASML machines Intel uses use EUV tech wich ASML acquired through their acquisition of Cymer, California and use licensed tech from Sandia Labs, also a US lab, meaning the US has full veto rights over to whom ASML gets to sell their EUV machines (spoiler alert, not to China).

> meaning the US has full veto rights over to whom ASML gets to sell their EUV machines (spoiler alert, not to China).

That's false. For USA to be able to block exports of the technology, more than 25% (if I remember correctly) of parts has to be of US origin. When the EUV program started, the share of US technology was much higher, but as the program progressed, it was diluted.

USA could still block sale of EUV machines, by refusing export/licensing of critical components, but they'd have to change their laws first.

Fact is that ASML voluntarily chose to block exports to China. In part because they want to keep good relations with USA. It's the US that funded the EUV program in the beginning after all. But I'm sure ASML was tired of China copying their stuff too, so it probably wasn't hard to convince them.

Re: Germany paying $5.5B for Intel fab

#263

Post by dqpb »

qiskit wrote: post #162
jakobov wrote: post #85

Earlier quoted context omitted.

It's about national security. A fab is necessary for national security

> It's about national security. A fab is necessary for national security How does germany funding a foreign/non-german company's fab part of national security? What about expelling a foreign occupying force? If the germans cared about national security, shouldn't they be looking in that first? If germany was funding a german company's fab, it would be national security. What germany is doing is paying tribute to a do…

The US and Germany are allies. In fact, I believe “German” is the largest ancestry group in the US.

Besides which, it’s extremely common for entities to invest in mutually beneficial agreements with other entities.

Re: Germany paying $5.5B for Intel fab

#264

Post by f6v »

alaricus wrote: post #15
f6v wrote: post #8

Earlier quoted context omitted.

Why? Germany doesn’t pursue political objectivity anyway.

I don't want my tax money filling the coffers of an American monopoly.

Understandable, but that’s why western liberal democracies have been established. G7 summits leave no room for this kind of reasoning.

Re: Germany paying $5.5B for Intel fab

#265

Post by Matthias247 »

DocTomoe wrote: post #99
april_22 wrote: post #77

Earlier quoted context omitted.

there was a strong chip industry in saxony decades ago, but they became way too expensive when china came. they are slowly starting to built back there, but progress is slow. Bosch and Infinion have the fabs there

You are refering to Robotron, which mostly is known for their Z80s knockoffs (which often were less buggy than the originals.) Given that they did not create chips of their new design, I would not describe them as having been 'strong', though.

Maybe more about AMD having had fabs there, which then transitioned into Globalfoundries?

Re: Germany paying $5.5B for Intel fab

#266

Post by colinsane »

sneak wrote: post #220
bottled_poe wrote: post #168

Earlier quoted context omitted.

Seems to me that most software is horribly inefficient anyway. If push came to shove we might just consider making better software?

The cutting edge of software that mostly uses these advanced process node sizes is not horribly inefficient. The new applications enabled by, for example, GPU-accelerated ML are not the same as running 400 chrome tabs on an M1. The hot paths in most important high-end computing are mostly quite efficient.

> The hot paths in most important high-end computing are mostly quite efficient.

it’s hard to determine the lower bound on a lot of these things. in scientific computing (e.g. physics sims) i still run across a lot of runtime branching that could easily be compiled out if we were using languages that had stronger type systems/templates. we use linear approximations for a lot of things, but it’s been shown that smarter interpolation can let you decrease your resolution without losing accuracy/stability. there’s an easy 2x or more perf gain every time you do that successfully. there’s also the boundary conditions: we always simulate more volume of space than we’re actually interested in so as to avoid certain types of distortion (reflections from the boundary of the simulation). over time we’ve learned tricks for reducing those reflections: we’ve applied a lot of general-purpose optimizations, but there’s evidence that we can go further if we encode more simulation-specific information at these boundaries. this whole area has just been slow, steady, compounded optimizations.

i don’t work in ML, but i find it very unlikely that (a) optimizing neural networks is a completely solved problem or (b) that you all know how to train them using the fewest iterations.

you make sweeping statements that just aren't obvious to me.

Re: Germany paying $5.5B for Intel fab

#267

Post by mikkergp »

verisimi wrote: post #224

Wow. These deals amaze me. Imagine you got given subsidies to live somewhere. I'm not talking about your wages - I'm talking about something extra. Well that is a direct equivalent here. How any governance structure can think this is ok boggles the mind. Is it not clear how we already live in a fascist state - where government + corporations are working together at the people's expense? $5.5 billion.

Serious question - How do you propose governments encourage companies to do what they want? Yes it’s a subsidy but it also seems like the government in this case is essentially buying a service. I guess Germany would hire people to build this factory directly rather than paying a foreign company to do it? Which I imagine would be more expensive if it’s even possible to do.

Do you think there are better ways to do this within capitalism or are you proposing we dismantle capitalism?

Re: Germany paying $5.5B for Intel fab

#268

Post by ajmurmann »

pasabagi wrote: post #230
ISL wrote: post #171

Earlier quoted context omitted.

It is very important for national stability. A highly-interconnected global economy is arguably better at securing international stability.

> highly-interconnected I think the problem is, decision makers need to understand what interconnectedness means , and not just on an abstract level. If you make things, you're intimately familiar with just how many components come from China. You're deeply aware of the fact the machines you use are made in basically every country in the globe. That the materials you consume come from literally every corner of the pl…

This is also the problem with something like the Buy American Act. I recently heard about a case where pipes(?) had to be pulled back out of the ground because it was discovered that the steel for the bolts had come from China. All this creates enormous overhead and worsens inflation, not make it better.

Re: Germany paying $5.5B for Intel fab

#270

Post by wazoox »

ISL wrote: post #171
mmastrac wrote: post #2

Investing in local chip fabs is long overdue. These are incredibly important for geopolitical stability and should be distributed world-wide. We've centralized too many fabs in too few places in the search for low cost.

It is very important for national stability. A highly-interconnected global economy is arguably better at securing international stability.

That was the main argument for global trade already in 1914. France was Germany's most important commercial partner, and vice-versa. All German rifles had stocks made from French walnut wood! All French locomotives had German-made tubing! This stupid mantra was repeated ad nauseam after 1991, too, with silly op-eds from Friedman and friends about how two countries with McDonald joints in both can't be at war and similar inanities. Surprise surprise, such wars happened many times since then (the latest being, of course, Russia vs Ukraine).
Post reply on HN