>>510265934 (OP)This is all akaik: 18A, intel's next node that is better in some ways (performance) worse in some ways (density) than tsmc's coming 2nm is in good shape and will be used in Pather Lake - to be announced in Q4.
Intel needs external customers for its semiconductor processes. Building fabs and figuring out the 1000+ steps a wafer goes though to be turned into chips is a hugely expensive task that needs to be amortized over higher volume.
But the way intel designs chips is different than everybody else. For external users to use their fabs Intel needs to develop PDKs (Process Development Kits) that integrate with industry standard CAD tools. Those PDKs from intel are not nearly as good as TSMC's - so even if the transistors are great - external customers can't build with them easily.
Intel's new CEO comes from Cadence a CAD tool vendor, where he was a CEO for some time. He may also be a better salesman for intel's process than Pat - the last CEO.
Intel is teaming with UMC, a smaller Taiwanese foundry to design PDK's for external customers that use its older equipment (iirc intel 12 and intel 16) which will hopefully help them learn how to be a foundry for other customers.
Intel has some external interest for its 18a, but not enough. It has to demonstrate with its next generation of chip that it has caught up with TSMC in technology (or surpassed it), AND for its next process (14A) it needs to have PDKs ready to be used by others.
The good news is that intel has seemingly caught up with TSMC in process technology. The accelerated path from Intel 7, intel 4/3, intel 20A/18A worked. It just has to work out economically.
Then the other side of intel is Intel Products, and it is a bit of a mess too - sapphire rapids taking ~12 steppings to get "all" the bugs out (for example). Intel has serious management issues too. All of this is being worked on.