[ajross]

> why are there shortages for these basic chips? Can’t they be made anywhere, and more easily?

Not really. Semiconductor fabs are built around "tools" from manufacturers like AMAT and Nikon. Those tool vendors make most of their money from selling new tools for fancy new processes, not supporting 20-year-old stuff. Eventually stuff breaks, and fabs have to offline these older processes.

The way this works in the tech industry is that "chips" are actually software, so if your old manufacturer isn't keeping up you resynthesize your VHDL or Verilog for a new fab, rev your board design or whatever, and keep going.

But other industries aren't so agile. They have older designs without design teams to support them, or even chip designs that they retain only as masks and not HDL. Those parts don't port cleanly to newer high-volume logic.

[gostsamo]

Actually, ASML provides lifetime support for their machines. I don't know for the rest.

[parsimo2010]

But lifetime support doesn’t help if the parts for your machine aren’t available anymore. If your 20 year old machine breaks and there aren’t parts available to fix it, you might get offered an equivalent replacement. If your old chip masks are incompatible with the replacement machine, you’re not immediately able to make what you need. So for some companies, having lifetime support might not help with the manufacturing slowdown when an old machine breaks.

[gostsamo]

I can't vouch for the parts, but they supported machines from the eighties. This is not some cheap consumer product that has half-lifetime of 13 months.

[rhodozelia]

The parts were made by humans once, they can be made by humans again. The only question is is it worth it

[salawat]

Underrated comment. Though the "worth it" bit is the trick.

In my estimation these older parts that "just werk" should be getting inherited and iterated on as a public good.

The idea that means of production should phase into public trust tends to get everyone in a tizzy though. I'd like to see a public "foundry of last resort" that focuses on being able to make anything.

[azeirah]

Not completely sure on this one, for example how the Romans had better concrete than we have now, even despite all of our scientific advancements

[gostsamo]

We know how they did it, it just does not scale, because it is not worth it causing a volcano every time we need some more ash.

[baybal2]

Most 20-30 years old machines on the market are Japanese, ASML wasn't that dominant in the long tail market up until DUV era.

And Japanese almost as a rule have whack a good leasing, and service business, including replacement parts for close to 30 years old equipment.

[xadhominemx]

> Eventually stuff breaks, and fabs have to offline these older processes.

Absolutely not. You just made this up.

[ajross]

The first is a statement of the third law of thermodynamics. The second clause is just obviously true. Go call up Fujitsu and try to order more of a chip they made for you in 1.5um in 1988.

[xadhominemx]

Yes of course the equipment breaks down but older equipment is easy to repair. It is very rare for a fab to be decommissioned and the equipment scrapped - in fact I have never heard of this happening to any production facility with 6” or larger wafers. That equipment will go to de-bottlenecking at some other fab and net production capacity for the node will increase.

Obviously many very old chips are out of production but not because the equipment broke down and was never repaired.

[ajross]

The corrollary to your point then is that all these fabs have immense idle capacity of exiting installed tools which they aren't using but retain simply because nothing ever "broke down"? Obviously that's ridiculous.

You're interpreting me pedantically while actually agreeing with my point, I think. Old processes don't have the capacity they used to[1]. If you don't like "stuff breaks" then how about "eventually the ROI on the equipment goes negative relative to the business so the line is idled and the fab real estate repurposed to make more profitable modern stuff." OK?

[1] Which, again, is just a "duh" kind of point and I can't believe we're arguing about it.

[xadhominemx]

No, old processes have very nearly the same capacity the used to, some even more. Several foundries are adding 8” capacity right now.

[blueblisters]

It’s not just about supporting new processes. Many tool/machine vendors are backlogged by years because they simply don’t have the capacity to make more than a few of those machines every year. Even if someone wanted to invest in new manufacturing, they would likely have to wait a few years to start production.

Secondly, some legacy manufacturers of semiconductor parts lost money on their capacity-building investments during the dot-com burst. The semiconductor industry is brutal and there is a genuine fear that overcapacity will make it hard to deal with any bust that happens after this boom.

[joe_the_user]

Correct me if I'm wrong, it seems like in the modern economy you have a situation where for many companies, most spending decisions are made long term, on the principle of what promises profits long term, regardless of immediate factor and with the perspective not overcompensating.

This has all sorts of bizarre consequences. In the middle of the PPE shortage - hospitals prevented their employees from buying PPEs themselves but would still only buy PPEs at the lowest price with a long term contract. And you had the Texas company that loudly proclaimed they couldn't sell their PPEs but they also only sold by long term contract. And this was all with people dying.

It's easy to see how manufacturer isn't going to be adding capacity for a puny short-term shortage.