... This is in line with an AC simulation of the same circuit. Bias point is set at vds=1.5V, vgs=1V, ac/dc blockers are used to separate dc bias point from ac signals. From this simulation the unity (current) gain ft frequency is lower, around 70GHz, may be because the current gain (calculated as the ratio of currents passing thru vd and vg ammeters) includes other parasitic effects, not only cgs and cgd. From the AC plot the current gain at 10GHz is in line with the value found in transient sim.

Yes! DC is not the better way to calculate ft for extreme high frequencies, but it can bring a feeling to where a transistor can be pushed. A better way is to use s-param to calculate it, but it will be another sleepless night to put the schematic in shape.

Sure, the big unanswered question is these models are valid at such frequencies. Most of the work on skywater is in the digital domain, at lower frequencies. Moreover the real layout of these devices (multi-finger, multiple gate contacts) affect RF behavior immensely.