Electronic – Effective transconductance of a CMOS inverter

M2 is essentially acting like a pullup resistor in this case. Real resistors are difficult to make on silicon chips, so a PFET in on-state is good enough for this purpose.

The chip designer can vary parameters like the channel length, width, and possibly doping level. Depending on the characteristics of the transistor, it could act more like a current source than a resistor at the operating point. Sometimes a "long tail FET" is used to make a rough current source. Without knowing the parameters of M2, we don't know if it is more like a resistor or more like a current source, although in this application that wouldn't make much of a difference. Ideally you'd want a current source for a pullup, but lots and lots of places you see resistors doing that job well enough.

The voltage shifts because additional over drive voltage is required to generate additional \$ I_{ds} \$ to match the \$ I_{ds} \$ of the PMOS. This arises because of the reduction of W reduces the \$g_m\$ of the NMOS transistor.

The curve will have it's roughly sigmoid shape, which is dictated by how the transistors transition from triode to full active, but they will not be perfectly symmetrical. \$V_{th} \$ , \$g_m\$ are very unlikely to be perfectly matched.