Fabs are extremely expensive, and an uncompetitive fab is literally just a pit in which you toss money. That's Intel's current problem.
> uncompetitive fab
To be clear, do you any inside information / career experience that you know that existing Intel fabs are losing money? I have not read any such reports in media.
As a counterpoint, Europe and Japan are no longer cutting edge in their fabs, but continue to produce many ICs for industry.
Nevertheless, it is hard to know how real are these numbers, because the prices charged by the Intel fabs for making Intel CPUs are set internally, so they could be manipulated to move all losses on the fabs from the consumer and server product divisions.
Moreover, for the latest Intel fabrication process for which anything is known, i.e. Intel 4, which is used for making the CPU tiles in the Meteor Lake CPUs, it is known that the fabrication yields have been much less than desirable, which had two outcomes, both less profit margins when selling Meteor Lake and the inability to produce as many Meteor Lake CPUs as they could have sold. Both effects have increased the financial losses suffered by Intel during 2024.
In general, "uncompetitive fab" can mean two things. Either there are things that the fab cannot do at all or the fab has too low fabrication yields at the things that it can do.
The Intel fabs are in the second situation. They can do about the same things that TSMC can do, but at much lower fabrication yields, which means much higher production costs. High production costs means low profits and Intel is a company that has been habituated to have huge profits, so it cannot manage well the case when its profit margins are very thin.
(Speaking of which, what does Intel do with old fabs for the past 20 years after moving to a new node?)
Old Intel fabs are refurbished, by installing modern equipment, so they are converted from old CMOS fabrication processes to recent processes.
Before being converted, the fabs spend a few years making peripheral chips that complement the Intel CPUs, e.g. the so-called south-bridge, which may be packaged together with the CPU for laptops or it may be a separate IC package for desktops, and which does not need a fabrication process as dense as the CPUs.
The problem is that the latest transistor sizes are the ones that have profit margins. 90% of TSMCs chips aren't on EUV processes, but 50% of their profits come from EUV. If you don't have leading (or close to leading) processes, you're completely shut out of the phone, laptop, server, and GPU market. Intel could survive like that, but it would involve laying off ~90% of their staff and exiting all of their major markets.