> which part of what they said do you disagree with exactly?
I'm disagreeing with:
> > "fabless semiconductor companies" are a significant part of the problem, too.
and
> > Until you can't ship any chips because all the fab owners are making their own chips and you have no leverage to get them to make yours.
because I don't believe the problem underlying the current chip shortage is due to the fabs voluntarily reducing their output in order to screw their customers, or that the fabs would give preferential treatment to in-house customers (which, as I mentioned, in the case of pure-play foundries like TSMC, certainly isn't the case).
> On the other hand, consolidation on foundry as a service makes the system more brittle.
That's true, but OTOH, the alternative to the current fabless + market foundry model isn't a world with tens or even hundreds of vertically integrated manufactures with their own leading edge fabs (that is, some kind of rollback to the 1980'ies), but rather a world where all leading edge chip development would be concentrated in a few megacorporations, with IMHO drastic effects on the rate of innovation across the entire industry, and susceptible to similar shocks like we see today if one of them would hit serious problems.
While I don't see a way out on the dependence on very few suppliers of leading edge fab capacity and all the problems that entails, I guess for slightly larger process nodes (say, in the 20-40nm range) where the capital costs aren't as big, there could be plenty of space for a healthy competitive landscape. Perhaps partially funded by governments (or things like the EU) that want their own chip manufacturing capability for "digital sovereignty" and national security/arms manufacturing reasons.
> because I don't believe the problem underlying the current chip shortage is due to the fabs voluntarily reducing their output in order to screw their customers
Same here.
The issue, I think, is that fabs are very frequently re-tooled in order to remain in geometries that have high demand and profit margins.
When I think about it, the issue is probably similar to the pharmaceuticals industry. Inexpensive / generic drugs have lower profit margins because they are largely figured out. Companies consequently opt to produce something else, despite there still being a market for the generic.
I'm disagreeing with:
> > "fabless semiconductor companies" are a significant part of the problem, too.
and
> > Until you can't ship any chips because all the fab owners are making their own chips and you have no leverage to get them to make yours.
because I don't believe the problem underlying the current chip shortage is due to the fabs voluntarily reducing their output in order to screw their customers, or that the fabs would give preferential treatment to in-house customers (which, as I mentioned, in the case of pure-play foundries like TSMC, certainly isn't the case).
> On the other hand, consolidation on foundry as a service makes the system more brittle.
That's true, but OTOH, the alternative to the current fabless + market foundry model isn't a world with tens or even hundreds of vertically integrated manufactures with their own leading edge fabs (that is, some kind of rollback to the 1980'ies), but rather a world where all leading edge chip development would be concentrated in a few megacorporations, with IMHO drastic effects on the rate of innovation across the entire industry, and susceptible to similar shocks like we see today if one of them would hit serious problems.
While I don't see a way out on the dependence on very few suppliers of leading edge fab capacity and all the problems that entails, I guess for slightly larger process nodes (say, in the 20-40nm range) where the capital costs aren't as big, there could be plenty of space for a healthy competitive landscape. Perhaps partially funded by governments (or things like the EU) that want their own chip manufacturing capability for "digital sovereignty" and national security/arms manufacturing reasons.