Hacker News Reader

Comment by erikpukinskis | Hacker News Reader

I don't think people full comprehend the price drop that is imminent. The total cost of ownership per mile for a car in the U.S. right now is about $0.60. I was watching a panel the other day where someone calculated the cost per mile of a fleet of self-driving on-demand electric cars at $0.08.

That an 86% reduction. People might have their misgivings about any of this stuff... electric, self-driving, carshare, and of course many people will go their own way. But for an 86% reduction in your monthly auto bill, it's going to be an avalanche.

View on HN

On demand self driving cars will be amazing. A lot of people I know work part time, or have two jobs. Car sharing kinda works, but it is complex to organize over facebook chat at 11PM.

Being able to rent the appropriate car to match the passengers would make it much easier. 7 people? 2 people? 3 people + 6 surfboards? 2 people with kyaks? I can not wait for this to happen.

If a Tesla Model S costs $70000, even if electricity and maintenance are free (and there's no premium for self-driving) you would have to drive it 875000 miles for it to cost $0.08/mile. This seems like the type of thing that is literally unbelievable without more evidence than "I heard somebody on a panel say this".

What's crazy about 875,000 miles?

Most personal gasoline cars last approximately 200,000 miles, but people rarely keep them that long because for the average driver the car is 10-20 years old. The interior will wear out long before the engine. If the car manufacturers built the drivetrain to last 400,000 miles, they would just be wasting money.

If the car is running 24/7 at 30 mph, you would top 250k miles per year. I imagine they will design the cars to last more than a year, potentially a few, so 750,000 miles is not off the table.

Also in the world where a car is an autonomous income producing asset, initial cost matters a lot less and total lifetime cost matters a lot more. Businesses will do the math and buy vehicles that minimize their cost. Automotive engineers will put more effort into reliability because fleet owners will have data to measure.

It's a different world. A lot of our current assumptions will get rethought.

I find it hard to imagine a conventional petrol powered car only lasting 200,000 miles. Which manufacturers only build cars to last 200,000 miles?

I can't think of a car I, or my family, have owned which hasn't lasted at least twice that long.

My current car is a 2002 BMW E39 530D. It has almost 900,000 kms on the clock and is still going strong on the original engine. The cloth interior is still in quite reasonable condition. There has of course been the need for minor maintenance (rear shocks were required at the last inspection). Right now I need new door seals and there is some minor rust attended to. However, at a consumption of approximately 7 liters per 100 km on the autobahn, I don't see any need to buy a newer car for quite some time yet.

In some sense, if a vehicle is well built, properly maintained, and driven gently, it will last forever. You just replace the car part by part over time.

The issue is if it's poorly maintained or it's abused, or if it's not well built. And a sizable percentage of cars in the fleet fall into at least one of those categories. The number that we need to worry about from a policy perspective is not the life of the car in the best case scenario, it's the life of the car in the expected scenario.

I had 2001 Suzuki Swift MKII 1.3. I think it had 65,000km on odometer in 2007 when I bought it, but can't be sure because odometer was only going to 99,999km. Now at (1)24,000km it's still going but I had to weld chassis or it would snap in half. Engine is still in very good condition though, only minor expected fixes were required. But yes, Suzuki DOES make cars which are not expected to last more than 100,000km. I've also recently got 2004 Opel Astra H which had already 209,000km and it's in perfect condition both engine and body, so YMMV.

Of course you won't see 24/7 usage on most cars, just like you don't see 24/7 usage on all electrical power stations.

Unless someone finds an economic activity that you can throw basically unlimited car-hours at in the night time. The yield can be low, but it has to be higher than the variable cost of one more car hour.

Surplus of night time car-hours can support a market for cheap delivery.

I thought about that, too. But I'm not sure whether that's not done better and cheaper by more specialised vehicles (eg trucks)?

The $0.08 is the _marginal_ cost of operation. You're not comparing apples to apples. The $0.60 cost of operating a gasoline vehicle does not include the original cost of the vehicle. (Using your same math: $70,000 car driven 100,000 miles would be $0.70 per mile.)

Ya, those are good numbers to point out. I was too lazy to link the discussion, but here it is:

https://youtu.be/HnQs1k_0Yys?t=29m36s

The speaker is Stefan Heck from the Precourt Instute for Energy at Stanford. No citation, so I guess it's just math he did.

I assume they're factoring in some reasonable price drop. I would expect the vast majority of self-driving cab rides will be in two-seaters like the Smart Car, which starts at $12,000. You'd buy them wholesale, so it should be even less than that. Not to mention you can get rid of the entire steering column, etc. At $10,000 per vehicle, you only need to drive 125,000 miles to break even.

If the vehicle is self-driving, and managed by Google infrastructure for ridesharing so its highly utilized, how quickly do you cover 875000 miles? A quick Google search indicates that traditional cabs cover 70K miles a year; so that's a <10 year payback period. Not terrible for an asset that's going to be ancient tech in 5-7 years.

You won't buy the car outright, and you won't drive all those 875,000 miles. You will rent it as required and others will use it the other 99% of the time.

Look at taxis/Uber and think about what portion of the cost is licensing (Uber's cut in their case), wage and fuel.

This would not be referring to the Model S, but something like the $20-30000 Model 3.

That's not an at all implausible distance for cars that are used all the time - compare it to a taxi or a delivery vehicle, not a personal car.

Reply on news.ycombinator.com