Hacker News new | ask | show | jobs
by spikels 337 days ago
Actually it's mass public transit that can't scale. Despite $1.3T in cumulative subsides US public transit usage is tiny: ~1.5% of trips and ~1% of passenger-miles.

Getting 100 people in a bus is not "easy" because very rarely do 100 people all want to from/to the same place at the same time. We end up bunching people in both time and space and making multiple stops and/or connections - which all make the journey longer and less competitive with alternatives.

And even then buses are only near capacity in our very biggest cities (top 10-15) for a hundful of hours during weekday rush hours. For example, in SF while buses/railcars may occasionally get crowded they are usually empty: Muni averages only 6 passengers per vehicle, BART only 8 per traincar. On a passenger-mile basis this makes the service very expensive (Muni ~$4 and BART ~$2 including capital) and very energy inefficient (Muni(LRV) ~400Wh and BART ~600Wh). Explaining why both systems are struggling financially and require massive ongoing subsides keep operating.

Because busses/subways/trains (i.e. mass transit) require lots of passengers to make financial - or even environmental - sense they will always be a niche, if very important, part of transportation. Robotaxis are competing in the much, much larger car/light truck market (80-85% of trips and passenger-miles). Uber/Lyft (plus traditional taxis) is already much bigger than public transit (~2% vs ~1-1.5%) and still growing and with what should eventually be a much lower cost structure robotaxis should greatly expand this. In other words robotaxis like Tesla will likely be much more scalable than public transit.

1 comments

This is so wrong outside US.

SE Asia and Europe is proof of public transport being viable.

Not as familiar with EU statistics but for shorter trips (under 300km) public transit is well under 10% of trips while cars are over 50%. For all intra-EU travel in passenger-miles cars are around 75% while busses/trains of all types is well under 15%. Viable sure, more scalable probably not.

You have numbers for SE Asia?