[jtorsella]

I really don’t understand this. Chips are tangible goods, not services where, for example, a bunch of hospitality capacity went to waste during the pandemic. Any “extra” chips will go to use, and will only serve to reduce costs for firms and consumers purchasing them, expand the range of products it makes sense to put chips in and the speed and quality of chips companies can afford to build on. While market signals are the best way to determine these things in general, many of the reasons mentioned in the article - long spin-up times for production infrastructure, dependence on other countries with fraught geopolitical situations - are strong reasons to build reserve capacity, even setting aside the direct economic benefits.

[mschuster91]

> Chips are tangible goods

Many chips are not fungible, at least not for large-volume orders that incorporate stuff like customer specific mask ROMs or optimizations, and certainly for built-to-order stuff like Apple's SoCs. That means that let's say Apple can't just go and take a bunch of NVIDIA's chips and slap them into their products, and I'd even take a bet and say that manufacturers of NVIDIA cards can't easily take up cancelled order capacity for the same NVIDIA chip from other manufacturers.

[jtorsella]

Yes, that’s a good point. But it seems to me like that’s a matter of degree, right? Like, even stipulating that the chips themselves are not fungible across products and companies, they are fungible across time for the same or similar products. The price of your 2023 fridge, for example, could be lower because of excess supply of chips for 2022 models, no? And am I right in thinking that even if the chips themselves are not fungible (and there is a genuine possibility that excess production of specific chips goes completely to waste), excess productive capacity is good in and of itself despite some amount of specialization?

[svnt]

No. Chips don’t really expire, and suppliers often have to wean their customers off them by raising the prices regularly once they’re done acquiring design wins.

If the chips went on sale every year then companies would gamble on additional stock at discount, driving revenue down, and further disrupting the sales transitions into new IC designs.