I agree with a lot of what you say, but I'm not as confident as you are that Intel could be successful in the near term ( ~5-10 years) going to a foundry model.
As you point out, the business has a design side and a manufacturing side. Intel has always been so far ahead on the design side, and their product was so valuable that they never tolerated any shortage of capacity so they built a lot of fabs. This was justifiable because the loss of revenue from being unable to meet demand due to supply constraints was much more significant than the extra expense of having a lot of fabs not at full capacity, so they overbuilt. That worked out great for them and was almost certainly the right decision at the time. But margins are too thin to run like that in a foundry. And its not as simple as just saying "we'll fill up our factories and get more money out of them." Everything about running the factory has to change. Business processes involved QA, repair and maintenance, operations, have not over the decades been subjected to the same sense of time based urgency as is the case at memory fabs or a pure play foundry. Changing that mindset and culture is quite an undertaking.
A second issue is all of the internal organization and infrastructure that has to be created to deal with all the different products and customers. At as single Intel fab they may make 4 families of processors, with several different very similar versions in each family. And the customer for every single one of those chips is Intel, with very clear Intel requirements. At a TSMC fab they will be simultaneously making everything from the SOC for a iPhone, image sensors for an Indian digital camera customer, a display driver for an LG TV, infotainment parts for Ford, ASIC chips for bitcoining mining, etc. There would be hundreds of types of chips and over a 1000 different products and the insane number of reticle sets that go along with that. You have to have an organzied way to maintain SPC charts and monitoring for that many devices. Device engineers who understand the layout of those chips to diagnose issues during manufacturing have to be hired. There need to be people that maintain relationships with hundreds of customers, each of which has different pPk requirements, different requirements for manufacturing turn around time.
Intel is fucking great at designing CPUs based on x86. They are great a manufacturing very very expensive chips in a slow and steady way. That has very little to do with running a foundry.