One train is like 500 autonomous vechicles on one road, and more comfortable. It will still win energy-wise. That you can hop on robotaxi after arriving makes it even more desirable.
What are we basing that on? Let's look at some numbers:
https://en.wikipedia.org/wiki/TGV
Let's take the Eurostar trains as an example:
> The Three Capitals sets are 394 m (1,293 ft) long and have 766 seats, weighing 752 tonnes.
Most of the other trains seem similar. The mass is about one tonne per passenger. So in terms of mass of metal you need to drag around per passenger, a full train is very similar to a small car (say, the latest model Mazda Miata which weighs 1056 kg) occupied by a single passenger.
But of course trains are very rarely entirely full (and cars average more than one passenger). But the whole mass of the train needs to make the journey regardless of whether there's people onboard or not! The train I caught the other day was probably dragging around ten tons of metal for every passenger. This is what makes trains a horrendously energy-inefficient means of transport.
Aerodynamics? Self-driving cars could draft behind each other at high speed. Probably not a drag coefficient quite as low as the aerodynamically designed train, but it also has a colossally lower cross sectional area, because a train is ridiculously high and ridiculously wide (in the interests of not being ridiculously long, and also for historical reasons).
What else? Oh yes, the car doesn't have to stop 'til it reaches the destination, trains tend to have stops along the route.
For speed, I don't see any reason why, once self-driving is solved, you couldn't have cars cruising at 200mph+ in self-organised "trains" on the existing I-5. No present car is really optimised for that kind of thing, so it's difficult to say how hard it would be, you'd need different tyre compounds for a start.
As for comfort I'll take the privacy of my own little bubble over travelling in the close company of four hundred random strangers.
And not stuff like "trains are very rarely entire full". Of course they're not. But cars are almost always entirely empty.
Rolling resistance is the big issue that usually favours trains. On the other hand, could we design cars with an extra set of wheels or something?
For long distance/high speed aerodynamics is more important than weight.
And even that is very far away. Meanwhile trains just work. No high-speed connection between LA and SF is an anomaly by world standards given size, proximity and economical importance of two cities.
Also, modern high-speed trains don't have 4-person compartments (or have those at premium), they are rows of seats not unlike plane business class seats. You usually have power, sometimes wi-fi. You can get coffee and a newspaper.
I don't want to own it. I don't have to own it. The transit authority can own and operate it.
This will be so much better than fixed-track transit it's not even funny.
You fly at 500 MPH, I'll drive at 65-70 or whatever the speed limit on I-5 is these days.
I will win this bet, most likely. You will lose. A train won't help you much, if at all, because it's going to embark and disembark on its own fixed schedule, probably in some crappy/scary part of town miles away from where you need to go. (Here's another reason: eventually some terrorists are going to attack the train. The TSA will muscle its way in, and you'll have to arrive at the station an hour or two early, just like we do now when we fly.)
There's just too much hassle and overhead involved in multi-modal transit... which is why people who can afford to drive still do, even when they have alternatives.
500 autonomous vehicles on the road are a train. Only without expensive, single-purpose hardwired tracks.
People don't understand just how insanely different things are going to look in another couple of decades. Trains make sense if, and only if, everything stays exactly the same.
But between the cities, you still have tons of cars going together, in a line, all the time. There is no flexibility here, almost everyone is going the same direction for most of the way. I don't see how even autonomous electric cars could beat trains on that. I think that the best way for the future is bullet trains/hyperloop connecting the large cities, and PRT within the city. I'm open to math that shows otherwise though.