The inverter ties the gates of a pMOS and an nMOS together as the input. The pMOS pull-up, when on, connects the output to V_DD; the nMOS pull-down, when on, to ground. Each body is tied to its source. The output drives only the next gate, which takes no steady current, so by KCL the current the pMOS supplies equals the current the nMOS sinks. In the netlist, the nMOS has V_GS = V_in and V_DS = V_out; the pMOS has V_SG = V_DD - V_in and V_SD = V_DD - V_out.
This chapter assumes V_DD > V_Tn + |V_Tp|, so both are on for V_Tn < V_in < V_DD - |V_Tp|. By the boundary, an on nMOS is saturated while V_out ≥ V_in - V_Tn (V_DS ≥ V_GS - V_Tn), an on pMOS while V_out ≤ V_in + |V_Tp| (V_SD ≥ V_SG - |V_Tp|).
The currents decide the regions. As V_in rises, the nMOS saturation current grows and the pMOS one shrinks; V_M is the input where they are equal. Below V_M the nMOS one is smaller, so both cannot be saturated. Nor can the nMOS be linear: that needs V_out < V_in - V_Tn, which saturates the pMOS, and its current would exceed the most the nMOS can carry. So below V_M the nMOS is saturated and the pMOS linear; above V_M the roles swap.
Nearing V_M from below, the pMOS current nears its saturation value, so V_out nears the pMOS edge V_M + |V_Tp|; from above, V_M - V_Tn. At V_M any V_out between these meets both conditions, V_M included. For V_in ≤ V_Tn the nMOS is off, so the on pMOS carries zero current, which in the linear region means V_SD = 0: V_out = V_DD. The last row mirrors this.
An arrow shows the output falling across a row. The middle row is a vertical drop: with channel-length modulation neglected, both devices are saturated only at V_M. In the first and last rows one transistor is off, so no path joins V_DD to ground: a steady input below V_Tn or above V_DD - |V_Tp| draws only leakage. A static CMOS gate works the same way: steady inputs at 0 or V_DD turn off its pull-up or its pull-down network.