Yes, being a TSMC is very different than being an Intel. One is a services company and the other is a product company. Being a service provider means you are necessarily market following and at least one level below the product company in the food chain, which it is reflected in margins, brand, product roadmaps, business dealings, etc. Intel providing foundry services automatically means they have accepted defeat.
Intel's mistake is not due to integrating design with manufacturing, but sticking with x86, or more correctly, not developing new alternatives with enough enthusiasm. Their only other serious product different from x86 that I recall is Itanium (XScale doesn't count). It also shows how monumentally difficult changing philosophies is when even Intel with near monopoly power over their customers could not ram through a new product (x86-64 won).
nVidia may be a one trick pony too, but they were visionary enough to see beyond polygon pushing and have been beating the "GPU is a better multi-thread processor than CPU" drum for many years. They worked with scientific community and developed parallel computing solutions from workstations to supercomputers. So, they were in the right place at the right time to take advantage of the current machine learning and crypto mining craze.
AMD is the small and sometimes abused sibling of Intel, where Intel will take the whole dining table and leave scrapes. Very hard to compete and develop when you are starving. Credit to them (despite couple of dumb leadership phases) for hanging on.
Coming back to Intel, they are seriously in trouble. They are the apex predator and hence need the whole jungle to do well to thrive, otherwise they become just another component provider. But that is what is happening with new ecosystems that are vertically integrated. Apple/Samsung don't need them, they need Apple/Samsung. Google/Amazon/Facebook will/are making their own chips. Microsoft has rediscovered their identity as software first company with hardware as enabler of software development. If they feel like they need a Surface-like product line for chips to do selective AI/ML stuff, they will do it. They have already done many chips for Xbox, Hololens, etc. That leaves only system aggregators (HP/Dell/Lenovo) and tier II (HTC, Huawei, etc.). So this means only Xeon servers are real money-makers, and how this market will be with big guys bringing chip design in will be interesting to watch.
So, Intel desperately needs a hit. They are investing heavily in ML/AI and have acquired many companies. Can they come up with a solution when the big vertical guys have far more ammunition (algorithms, customer data from photos to shopping history, etc.) to develop better systems? The same question largely applies to Qualcomm too. I think this is what is driving Broadcom to swallow giants whole to somehow become big enough to survive.