That's because M2 was on the same TSMC process generation as M1. TSMC is the real hero here. M4 is the same generation as M3, which is why Apple's marketing here is comparing M4 vs M2 instead of M3.
Actually, M4 is reportedly on a more cost-efficient TSMC N3E node, where Apple was apparently the only customer on the more expensive TSMC N3B node; I'd expect Apple to move away from M3 to M4 very quickly for all their products.
Saying tsmc is a hero ignores the thousands of suppliers that improved everything required for tsmc to operate. Tsmc is the biggest, so they get the most experience on all the new toys the world’s engineers and scientists are building.
It's almost as if every part of the stack -- from the uArch that Apple designs down to the insane machinery from ASML, to the fully finished SoC delivered by TSMC -- is vitally important to creating a successful product.
But people like to assign credit solely to certain spaces if it suits their narrative (lately, Apple isn't actually all that special at designing their chips, it's all solely the process advantage)
Saying TSMC's success is due to their suppliers ignores the fact that all of their competitors failed to keep up despite having access to the same suppliers. TSMC couldn't do it without ASML, but Intel and Samsung failed to do it even with ASML.
In contrast, when Apple's CPU and GPU competitors get access to TSMC's new processes after Apple's exclusivity period expires, they achieve similar levels of performance (except for Qualcomm because they don't target the high end of CPU performance, but AMD does).
Tsmc being the biggest let them experiment at 10x the rate. It turns out they had the right business model that Intel didn’t notice was there, it just requires dramatically lower margins and higher volumes and far lower paid engineers.
