Because the subthreshold slope has not intersected the
junction leakage floor by the time you hit Vgs=0.
This not only makes the leakage higher, it makes it highly
variable - a nominal 80mV/decade means that if your VT
of 200mV (low-VT) has a +/-100mV tolerance the real
leakage current could vary upward by over a decade at
Vgs=0 (or downward, but nobody cries about that).
And even at nominal, 2.5 decades (for 200mV/80mV)
means you'd leak about 3nA per transistor (where a
good long channel high VT FET should be femtoamp
range). Now figure your LVT transistor count.