Track utilisation is limited by a number of factors, including headways, much of which is governed by stopping distance.
I believe a general minimum practical headway allows about 10--12 trainsets per hour on a given track. At 400 passengers/trainset, that's roughly 4,000 to 4,800 passengers/hour, per track.
That compares with about 1,500 vehicles/hour per highway lane, or about 2,550 passengers/hour at the U.S. average of 1.7 passengers/vehicle.
I've seen some suggestion that headways could be reduced to 3--4 minutes, or 15--20 trainsets/hour. I know that some urban rapid transit systems (e.g., BART in San Francisco) achieve this, but only on a small section of the system (Colma to Oakland, transiting downtown SF, IIRC). Train speeds there peak at 70 MPH, and I believe are much slower for most of this portion of track.
And, I suspect, low speeds, as braking distance increases with the square of speed.
The New York subway has a top speed of 55 mph (88 kmh), but an average speed of 17.8 mph (28 kmh). I'm not sure what downtown intra-station peaks are, but suspect they're low on shorter sections. The London Underground is similar, with an average of 33 kmh (20 mph), max ~100 kmh (60 mph). This varies considerably by lines, the Bakerloo stock peaks at 45 mph; the Circle line averages just 24 kmh (15 mph). Chicago's 'L' is similar to NYC: max 55 mph, average ~20--25 mph.[3]
Note that averages include acceleration, deceleration, and dwell times (standing still at stations). But central portions of subways probably achieve low headways with relatively low peak speeds for congested track segments.
The single best thing you can do to drive down train cost is to put more people on each train.
Based on these sources: https://www.statista.com/chart/20169/gb-rail-fare-spending-b...
https://transport.ec.europa.eu/document/download/f5c82d86-23...