LIDAR prices are falling fairly rapidly over time (although I’m not sure about the impact of tariffs). The car and computing resources are otherwise boring costs, maybe they are at around $200k/vehicle? That would be pretty easy to pay for a with a few months of rides. If most of the rides are going through the Waymo app, they aren’t paying a lot to uber and Lyft, and I doubt they are paying the full 30% to the ride sharing platforms anyways.

▲

leoc 3 days ago | parent | next [-]

They're also spending an unclear amount of money on human driving-assistance workers https://www.nytimes.com/interactive/2024/09/03/technology/zo... and the supporting infrastructure. Presumably it works out to a lot less per vehicle-hour than an in-vehicle, US-resident human driver, but a lot more than nothing.

I suppose another wrinkle is that driverlessness isn't only a cost saving (aspirational or real), it's also a positive attraction eg. for anyone who worries about their safety with a rando taxi driver or Uber guy. There are also cost savings achievable by timeshifting antisocial-hours work to elsewhere in the world, though presumably a significant part of the savings will be simply be the result of outsourcing to lower-wage countries.

▲

Zigurd 3 days ago | parent | next [-]

Waymo has published videos about how their remote operator system works. Waymo's are never remotely driven. They're just told what the operator thinks is the best place for them to go next in a situation where the vehicle can't make that decision. This is critical to controlling opex. Waymos operate 24/7, that means three shifts of operators plus coverage for weekends, vacation, and sick time. This is where the difference between a demo and a product is going to be decided. Alphabet could be reckless about spending on remote operators. But I think they're still processing the trauma of buying Motorola and won't tolerate a big jump in headcount.

▲

Animats 3 days ago | parent | next [-]

> Waymo has published videos about how their remote operator system works.

Although not, as far as I've been able to find, videos that actually show the control center. Just semi-animated videos of the advising process. The number of customer support people per vehicle is not disclosed.

Employee stats are available.[1] Unclear how accurate they are. According to Unify, Waymo has 2,406 employees. Headcount is up only 6.6% since last year, even though the number of vehicles deployed has reportedly tripled. So they seem to have a big fixed labor cost in engineering, and but a low variable cost per car. Which means this scales well and becomes profitable.

[1] https://www.unifygtm.com/insights-headcount/waymo

▲

Zigurd 3 days ago | parent [-]

The relatively low headcount growth enables estimating values for the number of remote operators per vehicle. It takes four FTEs to fill one seat in the remote operation center 24/7.

If the ratio of remote operators to vehicles is 1:10, and the fleet size is 2500, that's 1000 FTEs. That's more than 5X the headcount growth.

If all of the headcount growth goes to remote operations, it would be one seat for 50 vehicles. I'd guess it's more like one to 100.

That fits with the relative rarity of seeing Waymos stopped waiting for instructions.

	▲	leoc 3 days ago \| parent [-]
		Do we know that the remote operators (all? mostly?) show up in that data as employees, though? Waymo's development has presumably reached the point where it would want to start filling remote-assistance seats with contractors. (There's also offline but relatively labour-intensive work like updating maps, adjusting routes and the like.) One remote assistant per 50 or 100 vehicles would also seem to put Waymo now far ahead of where Cruise was in 2023, when according to the NYT's sources Cruise's remote assistance staff "intervened to assist the vehicles every two and a half to five miles", though I suppose it's not especially hard to believe that 2025 Waymo is beating 2023 Cruise handily.

▲

leoc 3 days ago | parent | prev | next [-]

Sure: I didn't say that the remote workers were driving the autonomous vehicles. The NYT article I linked gives a pretty clear account of what they're doing. (The big exception here would of course be Tesla, assuming it ever gets its current "robotaxi" plans to fruition.) In any case, while the "what the humans are doing is not driving" thing always seems to come up early in these discussions, it's a relatively secondary issue when it comes to the cost and profitability question, while (as you say yourself) things like human-assistant-hours per vehicle-hour are central.

▲

voxic11 3 days ago | parent | prev [-]

Also (at least in the SF market anyways) their operating license does not allow cars to be remotely controlled.

▲

seanmcdirmid 3 days ago | parent | prev | next [-]

I guess, but my interaction with a real person in a waymo was limited to 15 seconds on my many rides, so I don't know if I'm just an anomaly or not.

> it's also a positive attraction eg. for anyone who worries about their safety with a rando taxi driver or Uber guy

Yes, this is the main reason to like it. I've had too many experiences with an Uber driver nearly falling asleep while transporting me.

> though presumably a significant part of the savings will be simply be the result of outsourcing to lower-wage countries.

Given how many Uber drivers are immigrants already, this is already happening if Waymo employs call center employees from the Philippines or not.

▲

Scoundreller 3 days ago | parent | prev [-]

Also cost savings because you can do things like work the 8AM rush on the east coast and then 3 hours later start picking up the 8AM rush on the west coast with its 3h time zone difference.

▲

Closi 3 days ago | parent | prev [-]

> maybe they are at around $200k/vehicle? That would be pretty easy to pay for a with a few months of rides.

The 'few months' bit doesn't seem quite right - the cost to get a human to drive a car would maybe be $50k per year, so i'm not sure how a $200k vehicle can pay itself off in a few months vs a $50k car + $50k per year driver.

I'm aware that the cost / ride is higher for Waymo, but it doesn't sound like that would be enough to cover the extra $200k and not certain that scales to other geographies outside of SF.

I mean to pay off a $200k vehicle in a few months you would need each car to be clearing $3k a day in revenue or something like that.

It's probably a few years per car if they are at $200k. If they are 'in the black' or not will probably depend more about their accounting rules (i.e. depreciation) more than anything else.

▲

maxlapdev 3 days ago | parent | next [-]

Note that you probably need 2 or 3 drivers per car to get a similar level of usage of the car that Waymo can do.

	▲	leoc 3 days ago \| parent \| next [-]
		Though that will also result in faster depreciation of the (mostly-)autonomous cars.
	▲	Closi 3 days ago \| parent \| prev [-]
		Very true, although even at 2 or 3 drivers it’s still more than months.

▲

ddeck 3 days ago | parent | prev | next [-]

Just to add another data point, the co-CEO indicated on a podcast 18-months ago that the sensor package cost was <=$100k for the then current generation:

"But saying, you know, picking an upper bound, $100,000 worth of equipment on it, you amortize it over, you know, the lifetime, call it, say, 400,000 (miles), 25 cents per mile. Right. And, you know, it gives you some margin compared to the cost of paying a human driver."

He also mentioned that the next generation would see a "drastic reduction in the cost".

https://www.shack15.com/podcast - ep41

	▲	SoftTalker 3 days ago \| parent \| next [-]
		Those Jaguars they drive are not going to last 400,000 miles without several significant overhauls. And interiors will wear out and have to be refurbished. Exteriors will get dented and scratched. Is all that part of the $100k they are amortizing?
	▲	Zigurd 3 days ago \| parent \| prev [-]
		Waymo Driver 5 was introduced in 2020. Driver 6 was introduced August of last year. So if the interview was 18 months ago he would be referring to Driver 5. Driver 6 hardware is probably built in batches of a few hundred each. Still a low volume CM job, but incrementally cheaper. Probably. If they start building in batches of thousands, that's going to drive costs down significantly.

▲

jedilord 3 days ago | parent | prev [-]

[dead]