There are ~14000 licensed taxis in New York. So all of the taxis in New York could carry the equivalent of 5.6 trains. Any given subway line has about 15 trains per hour and there are 25 or so subway lines. So you could assume there's maybe 350 trains running at any given time vs. 5.6 train-equivalents in taxi space.
So unless self-driving pods can somehow squeeze together 60 times more densely than taxis as well as not need to deal with crosswalks or traffic lights, the subway has a rather large advantage here.
https://www.quora.com/How-many-people-are-on-New-York-subway...
https://www.quora.com/How-many-trains-are-on-a-New-York-subw...
https://en.wikipedia.org/wiki/List_of_New_York_City_Subway_s...
Victoria line in London can do 34 trains per hour currently and there is a plan to get it to 36 (anything further is really difficult because you hit a physical limitation of getting people on and off trains).
AFIAK only the L train reaches anywhere near "world class" capacity (as it has CBTC), but is still only 26tph.
Other lines on the NY subway must be far less than that. From my experience 3-4 min headways per track are pretty common, which equates to 15-20tph. Obviously if you have both express and local then you're up to 30-40, but it could be ~double that with modern CBTC.
https://en.wikipedia.org/wiki/List_of_New_York_City_Subway_s...
The lines with the same colors run on top of each other in Manhattan so you have to sum them. Supposedly the 456 does get 44/hour, counting express and local tracks together.
I agree that most lines do not have modern signaling and most do not run at 45/hour most of the time. But also you know, 24x7 and all that.
If you combined that with walk through trains you could easily get a 200-300% capacity increase. There is a lot of theoretically low hanging fruit on the MTA, technically speaking.
Keep in mind how screwed cities like london are which do have the theoretical max per track and no other option but to build new subways. NY doesn't have this problem, it just needs to use its assets way better and there is room for at least a decade or two of further growth.
It's analogous to small packets vs jumbo packets. You're going to always have better throughput for the overhead, the larger the unit... so for just bulk moving of data (people) to/from a limited number of destinations, you'd be better served with the "jumbo" option. If you have a small number of people with a large number of destinations, a smaller option (cars), is better.
Unless we're talking completely reimagining the transportation network (which isn't a 5-10 year time frame project), you won't see use of autonomous pods displace subways -- especially in a place like NYC.
Maybe paved roads are cheaper to maintain than subway track, but that's hard to really compare apples to apples.
Honestly, I find it easier to imagine a NYC where there are no private vehicles allowed at all during rush hour than I find it possible to imagine one w/o subways.