It was not obvious to Intel leadership that the future of manufacturing was to become a contracted fabricator of the design of other companies. Vertical integration and total control over both design and manufacture of chips was the obvious (and, until recently) successful path. What board would keep an Intel CEO that presented the idea of shifting the primary focus of the company to manufacturing the designs of other companies whose products are in direct competition to Intel?
Intel is a perfect example of why companies need to regularly kill their old selves in order to survive. But it's also a perfect example of why so many companies fail to do so, since it often involves making what appears to be a terrible decision at the time, and convincing a bunch of people who know better to back you up. Very few leaders have the clout to pull this off.
The success of TSCM was the result of an interesting confluence of events: including the existence of ARM; Apple deciding that they were going to design their own hardware and outsource its manufacture; Nvidia developing CUDA; and the machine learning/AI revolution that drove demand for Nvidia cards.
If you take away any of those pieces, I think TSCM doesn't become the powerhouse it is today.
It's TSMC. Their rise is waaay before Apple or ARM. Plug and play fab model made getting into semiconductor design as a start-up possible. Broadcom, Qualcomm, Nvidia, Conexant etc all used TSMC in 90s, way before smart phones and ARM-mania. Intel saw this coming but they were ahead and they didn't worry about it. At the end they got stuck at getting bad yield from their FinFETs, which is invented by Intel, while TSMC and Samsung moved forward. Iphones would have been using Intel processors if Intel had a competing product to be honest. They screwed that up too. TSMC and Samsung just had the right mindset and focus, and a lot of smart and hardworking people.
+1 this.
Intel let MBA-think drive too many decisions (ie EUV). Maybe there was some misperception there, even cultural bias.
Pat's biggest real change has been to push the company towards process leadership. They always had good designs, though perhaps also damaged by MBA-think there too.
Still not sure about the decision to get out of non-volatile memory, though perhaps that was just a scope decision. Obviously, you could make a pretty amazing AI device with a whole bunch of NVM in-package...
The US does of course need on-shore chip foundries for strategic reasons.