To me it simply looks like Apple buys out the first year of every new TSMC node and that is the main reason why the M series is more efficient. Strix Halo (N4P)

To me it simply looks like Apple buys out the first year of every new TSMC node and that is the main reason why the M series is more efficient. Strix Halo (N4P) has, according to Wikipedia, a transistor density about 140 MTr/mm2, while the M4 (N3E) has about 210 MTr/mm2. Isn't the process node alone enough to explain the difference? (+ software optimizations in MacOS of course)