Why is CAPM theoretically flawed, and why does every banker use it anyway?

Theoretical weaknesses:

• Assumes a single risk factor — empirically there are several (size, value, momentum, quality in Fama-French).
• Assumes investors hold the 'market portfolio' — unrealistic.
• Beta is unstable.
• Roll's critique: the true market portfolio is not observable. In practice: CAPM is used anyway because (a) it's simple, (b) it's the market standard for pitches and defensibility, (c) more complex models don't have much better out-of-sample performance empirically, (d) in M&A valuations the CoE is an anchor number — comparability matters more than theoretical perfection.