[slackoverflower]

Uber has done the hard work of normalizing ride sharing everywhere and giving consumer pricing expectations of rides, just for Google to sweep in with it fixed-cost self driving car ride sharing service to start collecting massive profits. It is going to take a while to spread nationwide but Google will do it eventually, really just a matter of time (not even money for a company as rich as Google). I'm really wondering where Uber goes from here.

[losvedir]

I legitimately thought this was parody at first. Uber did the hard work just for Google to swoop in with its self driving cars? I mean, I guess it's hard to get VCs to subsidize rides to the tune of billions of dollars, but it's not technically challenging, compared to something like self driving cars which is straight out of science fiction.

[redbrickroad]

At the current cost of a self-driving car (~$250K), it would be ridiculously expensive to build a fully self-driving ride-sharing fleet [1]. So expensive that I doubt it's possible to make real money that way right now.

Self-driving cars are a long game until companies can bring down the costs. It might be years before that happens. Uber has the advantage of being able to ramp up self-driving cars as part of their existing fleet until then.

[1] https://qz.com/924212/what-it-really-costs-to-turn-a-car-int.... [2] https://www.uber.com/newsroom/company-info/

Edit: math mistake

[melling]

In the article it says WayNo has already reduced the cost of Lidar by 90%:

“Waymo, the Google spin-off, claims it has cut the cost of an experimental version of the high-end LIDAR to around $7,500. It did not release details”

More details here: https://arstechnica.com/cars/2017/01/googles-waymo-invests-i...

[hendzen]

They "reduced" the cost by stealing Velodyne's Lidar technology and filing a false patent on it.

https://arstechnica.com/cars/2018/10/lone-engineer-spanks-wa...

[nopinsight]

If a self-driving car can serve for 5 years without a costly overhaul, the amortized cost is only $250k/5 = $50k per annum. Let's say the annual maintenance is $10k, the total would be $60k a year.

It can possibly serve two effective shifts (in fact longer than that but the demand at some hours could be lower and it may need some time to charge, unless battery swapping is performed), which means the cost per shift is only $30k per year which could be in the same ballpark or a bit lower than a human driver would make per shift (depending on the market it serves, etc.)

This is just the beginning though and the costs will most likely come down while the human driver's costs will likely go up.

[adamnemecek]

The cost will go down fast.

[redbrickroad]

What makes you say that? Computers drive a self-driving car, and they've been at scale for a long time.

It might require moving to the Tesla camera-only approach to bring down costs enough to make self-driving cars viable. But that technology is still nowhere near there.

[Tuna-Fish]

The computers needed to run self-driving cars are not the expensive parts. The sensors are. And the sensors are expensive because there are so few users.

High-precision accelerometers used to be extremely expensive, with the best having prices in the high 5 figures. Then companies started putting (initially really crappy) accelerometers into smartphones, just to know which side is up. Then economies of scale hit and suddenly every smartphone has an accelerometer better than those super-expensive ones, and cost per part is pennies. (And then drones became a thing.)

The same will happen with LIDAR. Nothing about any of the parts of a self-driving car requires it to be super expensive, it's just that things that are made in quantity are cheaper than special snowflake parts.

[erikpukinskis]

We’re just bouncing low energy waves off shit bro, and hitting them with some Fourier transforms. If you’ve got a Russian mathematician WFH in his moms cabin, he’ll get this thing running on a Celeron.

[tachyonbeam]

Costs always come down with mass production. The more self-driving cars you produce, the easier it will be to bring down costs. LIDARs themselves can likely be had for cheaper if purchased in large quantities. There are also multiple companies working on less expensive solid state LIDARs.

Also, if we assume the cost of a self-driving car is 250K, that might seem like a lot, but if you amortize it over 5 years, that's 50K a year. Assuming that a car can bring in a mean profit of $10 per ride, that means it has to do 13.7 rides a day for 5 years to break even. That might seem high, but these cars are not human drivers, they don't need to sleep or take breaks, they could be on the road 20-23 hours a day, leaving some time for refueling and maintenance. I think it would be possible to bring in a profit with $250K self-driving cars and an uber-equivalent price points. It will become even easier when these cars drop to 200K, 150K and 100K, which is certain to happen once the technology spreads.

[slededit]

The car may be ready 24/7, but humans have a nasty habit of all wanting to travel at the same times. In reality the vast majority of trips will happen at rush hour and most cars will sit idle the rest of the day.

It’s brand suicide to not have a car when needed so fleets will be sized for peak load. When you apply queuing theory none of the numbers seem to add up.

[yourapostasy]

The numbers for just driving around and getting paid for the drive might not pencil out. But if you include the real time data about who (for payment purposes), intended destination, origin, and historical trips, that represents data advertisers will pay handsomely for.

If an advertiser tried to entice people to place a tracking device in their cars today for a reward, that would go over like a lead balloon. But Waymo will get people to pay for the tracking data and enjoy the riding convenience, while vacuuming up insights into consumer behavior at granularities that are the story elements of Black Mirror episodes. Advertisers are salivating at sending targeted offers that add pinpoint location and time to the axes they can already select upon.

While a teeny bopper rides, the screen screams out a BOGO offer to buy “the latest Kardashian shoe just 10 meters from your destination, press the Buy It Now button in the next 30 seconds and we’ll have both shoes ready for you in your size for immediate pickup at our entrance including a FREE gift $100 value for true Kardashian collectible fans! BE Kardashian Glamorous Tonight!”

[gbear605]

I'm not an expert, but I'd guess that most of the cost is not the physical hardware, but that each one is put together manually by a team of experts. That's a task that seems able to decrease significantly in price.

[lsh123]

$250B

[erikpukinskis]

twirls mustache

[dilyevsky]

How is undercutting the driver (that already don’t get paid shit) going to reap massive profits? Also think about what it takes for uber to start in new city (literally a guy with a phone in a motel) vs self driving company =)

[asdff]

I've never understood why the model isn't for self driving public buses. This should be the primary focus. More riders in one vehicle lowers costs for running the service itself as well as greatly cheapening wide area coverage abilities.

We all love the on-demand, single rider model that Lyft pioneered, but it also isn't sustainable with traffic demands increasing every year with population. Machine learning and productive partnerships with city traffic officials will ensure buses are routed on sensible routes that could be changed with demand without inconveniencing riders. Strictly protected bus lanes would also simplify the programming needed to maneuver a self driving vehicle through traffic.

[hindsightRegret]

Removing one driver from a bus which can carry 30-60 people is less cost cutting than removing one driver for 1-4 passengers for a private/pooled ride. Companies don't attack problems from a societal benefit-analysis standpoint, they attack them from a profit standpoint.

[function_seven]

The driver gets about half to 3/4 of the fare. So on a hypothetical $10 trip, Uber is getting between $2.50 and $5.00 of that.

Google can price a similar trip at $7, saving the rider $3 and boosting the revenue by $2 to $4.50.

[wnissen]

The driver grosses that much, but has to cover gas and the cost of the car itself out of their share. The actual amount the driver keeps appears to be on the order of $10 an hour, so I doubt we'll see very cheap rides any time soon.

[dilyevsky]

Yeah but they are now on the hook for car maintenance, charging (or gas), parking etc etc

[l4yao]

I imagine those details work out to be cheaper per vehicle for a company with a large fleet of nearly identical vehicles and a dedicated operations team.

[mcbutterbunz]

Doesn't Uber have their own driverless car division? Toyota recently invested $500 million in it as well. I think where Uber goes from here is to compete in the space they're very familiar with.

[tigershark]

Yes, and it is so good that they are the only self driving company to actually kill someone.

[fjsolwmv]

Corporate car services have existed for decades before Uber. Googe and partners popularized smartphones and mobile mapping, and developed mobile payments. Uber added nothing of relevance here except creating labor+capital intensive low paying jobs.

[ghaff]

I'm no particular fan of Uber but I don't think that's fair.

Yes, some of the attraction is a price that undercuts taxis. I often take a Lyft to and from the airport in a city I visit fairly frequently. But literally the only reason I do that rather than take a taxi of some variety is that it's 30-40% cheaper.

However, in many cases, Uber/Lyft are indeed easier to use than the alternatives. Not always--I take a private car back and forth to the airport at home--but under many circumstances.