Even if TSMC decided tomorrow to sit back and extract margin, GloFo could still probably get back in the game. That ability decays with time, of course, and that decay is what TSMC buys with their low margins. They are suffocating their competition to build a moat. It's a much better strategy than "extract margin and make big bets on fancy new tech to stay one node ahead," with Intel as Exhibit A.
A) spend tens of billions of dollars on a new fab facility B) shut down production in one of the existing facilities while spending billions of dollars to upgrade it to 7nm
While not having the volume to be able to court the largest customers who are the most interested in 7nm.
They decided to stick with their existing 14nm infrastructure and specialize in the long tail of smaller customers who want custom chips and don't need a leading-edge node.
Large Nvidia chips might not be able to move quickly, but all of the MCU, WiFi, etc chips would migrate to cheaper fabs like SMIC. Which reduces TSMC's revenue, which reduces their ability to invest in new processes. Cents (and even fractions of cents) matter to those chips.
They have the money and can attract the talent
They only build their own stuff when they see a competitive advantage to doing so — when IBM stopped delivering on chip efficiency or when Intel’s chips started stagnating.
... and invest in making their own chips from end to end in a couple of years. Apple has proven that they do not hesitate to completely take over stuff with the recent Intel-ARM switch when the quality or price does not match what they perceive to be fair.
TSMC is in an extremely powerful but also very precarious position - they are worth ~400B $ in market cap, and Apple has ~200B $ in cash reserves and a market cap of 2T $. If Apple wants they can either buy up TSMC whole or they can set up a competitor.
=)