Yes I am aware. Whats your point? Waymo is actually doing fully autonomous miles, what are Tesla's doing? An intervention every 13 miles on avg. But Elon says they will be deploying fully autonomous Tesla's in 2025. Yeah ok. Ill be waiting.
> Waymo decided to solve the problem using whatever they deemed necessary.
Solving the technical side is only half the problem, otherwise we'd all be flying in Concorde jets.
I'm curious to see if Waymo can scale up to mass production of their sensor suite and maintain them economically, where it makes sense as a business. There may be a big business in LIDAR factories and repairmen soon.
That was mostly a social/policy problem which destroys the economics of it because you're limited to a very small amount of airports.
Waymo is currently limited to a small amount of towns for a mix of economic, policy, and social reasons. Probably in that order. Just being sufficiently better than human drivers was never the entire barrier to entry for this to be a real business for Google, at least one that becomes nation-wide or global.
Predicting it on a macro level is a boring question IMO, the influence of policy interference and social outrage every time there's an accident - which is never fully connected to the realities on the ground of true capabilities/potential - is basically a black box. Like betting on elections.
My interest atm is mostly the economics question of whether LIDAR will scale widely in near term combined with the social risks (ie, millions of cars with $30-50k in sensors, which atm cost as much as the car itself, so basically only e-taxis, no personal ownership + high cost of maintenance) vs cameras being good-enough... that's a far more interesting question. It's pretty obvious LIDAR/ultrasonic or a combination with cameras is superior. So maybe some ambitious person will find some mass-producible middle ground within <5yrs shrug
Calling San Francisco a "town" diminishes what they've achieved. It is the second densest urban center and ride hailing market in the US, and a difficult place for humans to drive. I think being able to effectively replace Uber drivers there means that once they crack freeway driving and snow, the regulatory, economic, policy, and social hurdles will start to fall like dominos.
I don't think this is how most people expected this to go. My personal belief was that self-driving cars would be limited to medium-density suburban areas initially (i.e. places where the driving is relatively simple, but there's still enough demand for it to be economically viable). It's astonishing that they were able to move from suburban trials to high density environments so quickly.
Everyone obsesses on the technical side which was necessary but scaling it up country wide as a functioning profitable business entity, even just urban/suburban, requires far more than just heavily subsidized R&D work and experiments as a minor taxi player in a few places.
I have doubts the very slow pace of expansion was just a matter of regulations and nailing down the tech. It has to make sense as a business otherwise they aren’t going to dump billions into sensor factories (where each LIDAR suite currently costs around $75k to make), cars, and service infrastructure.
Just like how Tesla wasn’t just a of matter of making the first prototypes of electric sports cars. The real magic and barrier to entry was the factories and infrastructure to make the economics work.
This was when I really started to doubt that guy's genius. That was a phenomenally stupid assertion. I can drive with 2 eyes because I have a bunch of other sensors as well as an entire world model in my head with 30 years of driving experience.
Is it Halloween tonight? Better drive extra slowly. Did that white semi with the bright background sky just disappear? Probably not. Is that deer going to bolt in front of the car coming the other way? Maybe. Is the road slick because it recently rained here? etc.
I've ridden a few times and their prices seem close to Lyft. I would happily pay a premium for the service they are providing right now. New, clean cars that drive very smoothly and lower rates of accidents than an 'average' driver. No weird smells, distracted drivers, inexperienced drivers, etc.
I think in a few more years with the amount of training their AI will have Waymo will be a truly incredible taxi service. It can only get better!
> New, clean cars that drive very smoothly.. No weird smells.
One concern I had is that once self-driving cars are widespread people might take advantage and treat them poorly with no human driver watching. Where a human would likely notice and deal with weird smells or things added by the past customer. Waymo-type service probably rely on reporting by passengers once the car already arrives, no? (I've never driven in one)
People might be treating them nice today because Waymo is a fancy new service in Jaguars that people treat as a novelty. Once those e-taxi services are under aggressive economic demands I'm curious to see how it plays out.
Waymo's have interior cameras on their vehicles. Any rider that leaves mess will get a strike and probably be banned for repeated infractions. You're right, some problems will only be noticed by the next rider.
Waymo is closer to a proper liveried taxi service than a rideshare. The only time I use rideshares is for NEMT. I use a human-driven taxi, albeit much less than Waymo, and I book through the taxi company's app.
The taxi company uses traditional medallions allocated by the authorities here. Their dispatch is traditionally over-the-phone, in fact they require me to answer the phone if I expect to get a ride. Their drivers--I don't know their employment status--seem to own their vehicles, which are fleet models, liveried, and usually identical. So I can always count on uniformity when I get in a car like that.
The human taxi service usually charges/estimates prices at least 20% higher than an identical itinerary with Waymo.
Waymo service is not Uber or Lyft. First of all, it's a liveried fleet vehicle. When I summon a Waymo, I know that they'll send a Jaguar I-PACE, and I'll know how to board it and fasten my seatbelt. That is absolutely golden knowledge to have. (Rideshares routinely give me panic attacks as I try and figure out how to seat belt when they've already started moving.)
The Waymo has an absolutely uniform experience, because it's always "the same Driver". Human drivers I've met have been drunk, belligerent, rude, scammers, incompetent, unsafe, late, lost, wrong car, overly friendly/nosy, you name it. Waymo is none of these.
Waymo's power trunk door means I can always stow my cargo, because the driver isn't living out of his car. Unfortunately, there is no human to assist me in lifting it, but there's room and there's a button to close it again.
Of course the in-car experience is luxurious and relaxing. I don't have to be concerned with a jerk driver or entertaining the guy or pretending to ignore him. I can basically curl up and zone out. I feel safety and security in Waymo alone.
The in-car display has plenty of telemetry to watch the 3D visualization, play music, adjust temperature, do what I want. There are always charging ports for my phone, cup holders, and sun visors.
Waymo is a Godsend for me. And it's a Godsend for anyone in a vulnerable population who may be at-risk while alone with another human like that. It's about time.
I comment so frequently on Waymo posts that I'm sure folks think I'm a shill... That said, I agree with this post fully and I'm not a vulnerable population at all. I can't say enough positives about the experience. I hope to God the safety record (no fatalities, minimal or zero serious injuries caused by a Waymo) continues because it will be a huge bummer if vasts swaths of the US (world?) population aren't able to take advantage of this in their daily lives. I really think it will radically change personal transportation. I also think competition would be great because that will force prices to remain somewhat reasonable so bring on Tesla, Cruise, whomever else can safely deliver a similar product.
It is completely honest. The question was "What fraction are driven by a remote safety driver???". Fleet response agents do not have the ability to remotely drive the vehicles. From your own source: "The Waymo Driver [...] is in control of the vehicle at all times".
Phoenix was the first place that Waymo offered service, which the selected for the weather, IIRC. The video highlights the spinning lidar sensors. How does that do in freezing rain? How long until a solid state lidar is a practical replacement (and would it be any better)?
I would presume at this point they don't drive when the weather is near freezing. icy conditions are going to be a huge issue for the cars. LIDAR and Cameras don't help with black ice.
My guess is that getting local permits is easier to get for Waymos (never faster than 45mph, only within bounds of a single city) than long haul cargo (on highways faster than 65mph, crossing multiple city boundaries).
My guess is we don’t have any current infrastructure for outside the city depots. Right now trucks pick up the cargo after long distance shipping and take them straight to their final destination. Setting up what you’re talking about would require a paradigm shift in how the trucking industry works.
Highways are simpler on paper, but it turns out that any crazy thing (pedestrians, obstacles, etc.) that happens on a city street can happen on a highway, so you have to be ready for that. Energies are many multiples higher at highway speed, so "contact" events are much more serious.
The solution for this is to make it so the cars physically can't go everywhere they want to go. Meaning, instead of a highway designated for freight, use a rail.
Same. Ten years ago I was predicting that the long haul truckers would be the first major segment of the workforce to be replaced by AIs. Turns out it is going to be junior devs and entry level artists. Oops.
A ride takes 15 minutes. Costs $20. 15 minutes, on average, of downtime between rides (late-night offsets shorter during day/evening)? So $20 per 30 minutes. 24 hours in a day. 48 rides per day. $960 per car per day. Holidays/weekends probably fewer riders so say 300 days per year to keep it conservative. $288,000 revenue per year. Car costs $150,000 per year (costs are going to drop dramatically as they scale up, I would think. And I'm assuming that after driving 300,000 or so miles per year that the car will be cooked, but that's probably overly conservative). Still, though, good gross margin (or at least the margin on the vehicle cost alone; maintenance is going to cost something but tires and a few other things are not going to be material relative to the cost of the car itself). Note that a quick Google (eh) search shows Google overall gross margin of 58% so this wouldn't be too far off of that. That is to say, pedal to the metal on the roll out if these numbers are even marginally accurate.
Please pick apart at your leisure. I'm most interested in knowing the truth, but just thinking through this and how it might look internally at Waymo/Google. If anyone has actual numbers (maybe they talked about on earnings call?) would love to hear them.
Like I said in another reply, there seems to be a good bit of debate about whether that cost, which was cited several years ago, is still accurate (or ever was).
That said if it is then 2x would drop the margin in half. Still a healthy margin and I do wonder what the Geely car cost will be. Note when I say "car" I mean everything for the car to drive itself, which is the car, the sensor suite, etc. I did not figure on the operational staffing as another comment pointed out.
The best source for the price is what the ex-CEO said in January 2021:
> Let me paraphrase it like this: If we equip a Chrysler Pacifica Van or a Jaguar I-Pace with our sensors and computers, it costs no more than a moderately equipped Mercedes S-Class. So for the entire package, including the car - today,
What do you mean by the car costs $150,000 a year? The price of the Waymo is approximately 250,000 for each car, so if you’re trying to amortize that over 10 years, it would probably be closer to $25,000 a year.
Amortized over 1 year. Was trying to be super conservative. Zero chance a Waymo lasts 10 years. I mean maybe you could say the body will last 10 years, but would imagine quite a bit of the car, over the course of 300k miles driven, would need to be replaced? Not too many cars on the road with over 300k miles.
But still your point stands that maybe that $150k car cost would be spread out over a longer time frame, which makes the economics much better. Like I said was trying to be conservative.
As for the $250k price point, when I search for that most pundits think that figure is several years old and may have been overly high to begin with, but if it's true at $250k then obviously a lot of margin pressure there. Still think it's likely the car cost will drop precipitously and the $150k figure won't be far off.
How do you price engineering samples for an iPhone, Tesla Model 3, or a Chromebook,
or an Xbox, or anything else? The $250k number includes a lot of custom engineering work amortized over a relatively small number,
since, the build hasn't been mass productionized and they're not pumping out thousands of cars per week.
Hey, not sure I fully understand your question, but are you talking about the initial capital costs to get to a fully "productized" self-driving car? I would leave those out of the unit economics because (I think) those get treated differently financially/accounting-wise and also are not really representative of whether the future business is sustainable (it's more about whether the initial investors get the proper return on their investment).
Might not be addressing your question with that, but I think that's what you're asking. And I think you're assuming the opposite of me which is that they haven't yet productized their offering, but I think they actually have at this point. I definitely could be wrong about that. Just an opinion.
my point was that using $250k as the cost of the fully loaded waymo car, and then doing math based on that, is basing the math on what is essentiallu a made up number.
The last info from early 2021 is that it costs no more than a moderately-equipped Mercedes S-Class (so around $150k?). Good chance that the next car will be under $100k and $50k is possible in the long term.
The cars were estimated to cost 150k, based on information that's now a few years out of date. Given that they've had time to push down the costs I wouldn't be surprised if it's significantly less now. (Probably still over 100k since the Jaguars are not cheap vehicles.)
watch that live stream of the depot to hey a feel for how few people it takes on the ground. And because customer support isn't remote driving the vehicles, but merely aiding the classifier, they don't need to be somewhere close enough that latency is totally minimized (aka cheaper than the Bay Area or an expensive part of socal).
Of course, however, your can't forget about the literally billions of dollars and decades of time that were plowed into R&D for them to get this far. All that wants to be recouped at some point.
But also, each intervention only makes up a fraction of the trip's duration. So one operator might be enough for hundreds of cars. That doesn't include operators for cleaning, charging, maintenance, and other downtime activities though.
The issue is fleet utilisation, the fleet is not going to be fully utilised at night and will be capped out during rush hour. Compared to rush hour, daytime usage is about 50% of peak and it falls to below 10% or lower at nights[1]. There was an article or blog looking at what happened if Uber replaced all their drivers with robotaxis a few years ago, couldn't find it though, but conclusion seemed to think going asset heavy an owning cars/parking would be the same or worse than paying drivers more but drivers owning their cars
I'm not convinced that's true, esp since Covid and WFH. Since they only travel within SF/LA/Phoenix, it doesn't follow typical suburban commuter patterns, nor totally Uber/Lyft ride patterns either since they can (currently) travel further.
If Alphabet/Waymo leadership were more theatrical... I wonder if the Tesla event delay could have given them time to expand 3 miles up the 101, across the Cahuenga Pass, so they could offer fully-autonomous Waymo rides through LA traffic to and from Musk's WB lot Tesla demo.
though it's not at all their style. And realistically I'm glad they're being methodical and prioritizing safety.
I heard it speculated that maybe Tesla chose that location for their event specifically because it was outside of the Waymo area and they didn’t want a bunch of journalists showing up in Waymos to the Tesla robotaxi event.
The other thread on HN about the funding waymo received just went to show how little Tesla and their fans know about Autonomy and that it cannot be solved with their hardware stack. They talk a lot but wouldnt dare put their money where there mouth is and sit in the backseat and take a ride across town with supervised FSD (which is an oxymoron). It is such a clear stock pumping scheme by Elon to maintain their Market cap and people fall for it. Im long Uber and Goog on this and when Tesla fanboys finally get struck with reality its going to be hilarious.
I took my first Waymo trip this past weekend. I will be using it every time from now on.
I try to take an actual taxi (Flywheel) whenever possible, because Uber and Lyft feel... I dunno, exploitative to me. But my last 3-4 cab rides have been annoying and/or terrifying. One cabby stopped in the middle of a busy intersection to check directions. Another cabby got lost the minute I got into the cab, and I had to guide him. Then he kept peppering me with personal questions. Another cabby tried to convert me to Islam...
Waymo was quiet. Obeyed traffic signals. Took a rather tricky backroad to get to my apartment, and navigated it flawlessly. Tried to convert me to the Cylon god.
I'm gonna root for the robots on this one, I'm afraid.
A few weeks ago I was driving down Santa Monica Blvd in West LA during rush hour and a Waymo was stopped with its hazard lights on in the middle of the road blocking traffic right by the 405 entrance. I finished my errand like 40 minutes later and it was still in the same spot blocking traffic, so idk I’m not super confident in these things yet.
76 comments
[ 2.4 ms ] story [ 135 ms ] threadWaymo decided to solve the problem using whatever they deemed necessary.
Solving the technical side is only half the problem, otherwise we'd all be flying in Concorde jets.
I'm curious to see if Waymo can scale up to mass production of their sensor suite and maintain them economically, where it makes sense as a business. There may be a big business in LIDAR factories and repairmen soon.
That means very limited routes. That is a technical problem.
Hopefully, NASA solves the technical problem and commercial supersonic flight returns
https://www.nasa.gov/centers-and-facilities/armstrong/taming...
Waymo is currently limited to a small amount of towns for a mix of economic, policy, and social reasons. Probably in that order. Just being sufficiently better than human drivers was never the entire barrier to entry for this to be a real business for Google, at least one that becomes nation-wide or global.
To me it sounds like you’re just being sort of vague. Yes, absolutely everyone knew there were social issues too.
Where will self driving be in 5 or 10 years?
My interest atm is mostly the economics question of whether LIDAR will scale widely in near term combined with the social risks (ie, millions of cars with $30-50k in sensors, which atm cost as much as the car itself, so basically only e-taxis, no personal ownership + high cost of maintenance) vs cameras being good-enough... that's a far more interesting question. It's pretty obvious LIDAR/ultrasonic or a combination with cameras is superior. So maybe some ambitious person will find some mass-producible middle ground within <5yrs shrug
I don't think this is how most people expected this to go. My personal belief was that self-driving cars would be limited to medium-density suburban areas initially (i.e. places where the driving is relatively simple, but there's still enough demand for it to be economically viable). It's astonishing that they were able to move from suburban trials to high density environments so quickly.
I have doubts the very slow pace of expansion was just a matter of regulations and nailing down the tech. It has to make sense as a business otherwise they aren’t going to dump billions into sensor factories (where each LIDAR suite currently costs around $75k to make), cars, and service infrastructure.
Just like how Tesla wasn’t just a of matter of making the first prototypes of electric sports cars. The real magic and barrier to entry was the factories and infrastructure to make the economics work.
This was when I really started to doubt that guy's genius. That was a phenomenally stupid assertion. I can drive with 2 eyes because I have a bunch of other sensors as well as an entire world model in my head with 30 years of driving experience.
Is it Halloween tonight? Better drive extra slowly. Did that white semi with the bright background sky just disappear? Probably not. Is that deer going to bolt in front of the car coming the other way? Maybe. Is the road slick because it recently rained here? etc.
Still pretty cool.
I think in a few more years with the amount of training their AI will have Waymo will be a truly incredible taxi service. It can only get better!
One concern I had is that once self-driving cars are widespread people might take advantage and treat them poorly with no human driver watching. Where a human would likely notice and deal with weird smells or things added by the past customer. Waymo-type service probably rely on reporting by passengers once the car already arrives, no? (I've never driven in one)
People might be treating them nice today because Waymo is a fancy new service in Jaguars that people treat as a novelty. Once those e-taxi services are under aggressive economic demands I'm curious to see how it plays out.
If you purchase a GoPass, you'll receive a 10% discount on all rides, and 5 free cancellations.
Or will you ever be able to reach support once car takes you to wrong location?
https://sfstandard.com/2023/08/11/san-francisco-robotaxi-cru...
Waymo is closer to a proper liveried taxi service than a rideshare. The only time I use rideshares is for NEMT. I use a human-driven taxi, albeit much less than Waymo, and I book through the taxi company's app.
The taxi company uses traditional medallions allocated by the authorities here. Their dispatch is traditionally over-the-phone, in fact they require me to answer the phone if I expect to get a ride. Their drivers--I don't know their employment status--seem to own their vehicles, which are fleet models, liveried, and usually identical. So I can always count on uniformity when I get in a car like that.
The human taxi service usually charges/estimates prices at least 20% higher than an identical itinerary with Waymo.
Waymo service is not Uber or Lyft. First of all, it's a liveried fleet vehicle. When I summon a Waymo, I know that they'll send a Jaguar I-PACE, and I'll know how to board it and fasten my seatbelt. That is absolutely golden knowledge to have. (Rideshares routinely give me panic attacks as I try and figure out how to seat belt when they've already started moving.)
The Waymo has an absolutely uniform experience, because it's always "the same Driver". Human drivers I've met have been drunk, belligerent, rude, scammers, incompetent, unsafe, late, lost, wrong car, overly friendly/nosy, you name it. Waymo is none of these.
Waymo's power trunk door means I can always stow my cargo, because the driver isn't living out of his car. Unfortunately, there is no human to assist me in lifting it, but there's room and there's a button to close it again.
Of course the in-car experience is luxurious and relaxing. I don't have to be concerned with a jerk driver or entertaining the guy or pretending to ignore him. I can basically curl up and zone out. I feel safety and security in Waymo alone.
The in-car display has plenty of telemetry to watch the 3D visualization, play music, adjust temperature, do what I want. There are always charging ports for my phone, cup holders, and sun visors.
Waymo is a Godsend for me. And it's a Godsend for anyone in a vulnerable population who may be at-risk while alone with another human like that. It's about time.
Meanwhile, I have bought and own a Tesla that drives me autonomously anywhere in the country, from a company that just posted a record quarter...
Yeah, Google is surely winning this race.
Waymo was doing winter testing in Buffalo, NY in February of this year: https://www.wkbw.com/news/local-news/inside-the-self-driving...
A ride takes 15 minutes. Costs $20. 15 minutes, on average, of downtime between rides (late-night offsets shorter during day/evening)? So $20 per 30 minutes. 24 hours in a day. 48 rides per day. $960 per car per day. Holidays/weekends probably fewer riders so say 300 days per year to keep it conservative. $288,000 revenue per year. Car costs $150,000 per year (costs are going to drop dramatically as they scale up, I would think. And I'm assuming that after driving 300,000 or so miles per year that the car will be cooked, but that's probably overly conservative). Still, though, good gross margin (or at least the margin on the vehicle cost alone; maintenance is going to cost something but tires and a few other things are not going to be material relative to the cost of the car itself). Note that a quick Google (eh) search shows Google overall gross margin of 58% so this wouldn't be too far off of that. That is to say, pedal to the metal on the roll out if these numbers are even marginally accurate.
Please pick apart at your leisure. I'm most interested in knowing the truth, but just thinking through this and how it might look internally at Waymo/Google. If anyone has actual numbers (maybe they talked about on earnings call?) would love to hear them.
That said if it is then 2x would drop the margin in half. Still a healthy margin and I do wonder what the Geely car cost will be. Note when I say "car" I mean everything for the car to drive itself, which is the car, the sensor suite, etc. I did not figure on the operational staffing as another comment pointed out.
edit: changed thread to reply in first sentence
> Let me paraphrase it like this: If we equip a Chrysler Pacifica Van or a Jaguar I-Pace with our sensors and computers, it costs no more than a moderately equipped Mercedes S-Class. So for the entire package, including the car - today,
So I think the $150k is a good estimate.
They’d just make their own like zoox.
But still your point stands that maybe that $150k car cost would be spread out over a longer time frame, which makes the economics much better. Like I said was trying to be conservative.
As for the $250k price point, when I search for that most pundits think that figure is several years old and may have been overly high to begin with, but if it's true at $250k then obviously a lot of margin pressure there. Still think it's likely the car cost will drop precipitously and the $150k figure won't be far off.
Might not be addressing your question with that, but I think that's what you're asking. And I think you're assuming the opposite of me which is that they haven't yet productized their offering, but I think they actually have at this point. I definitely could be wrong about that. Just an opinion.
Of course, however, your can't forget about the literally billions of dollars and decades of time that were plowed into R&D for them to get this far. All that wants to be recouped at some point.
But also, each intervention only makes up a fraction of the trip's duration. So one operator might be enough for hundreds of cars. That doesn't include operators for cleaning, charging, maintenance, and other downtime activities though.
[1] https://www.researchgate.net/figure/ehicle-use-by-time-of-da...
I'm not convinced that's true, esp since Covid and WFH. Since they only travel within SF/LA/Phoenix, it doesn't follow typical suburban commuter patterns, nor totally Uber/Lyft ride patterns either since they can (currently) travel further.
though it's not at all their style. And realistically I'm glad they're being methodical and prioritizing safety.
I try to take an actual taxi (Flywheel) whenever possible, because Uber and Lyft feel... I dunno, exploitative to me. But my last 3-4 cab rides have been annoying and/or terrifying. One cabby stopped in the middle of a busy intersection to check directions. Another cabby got lost the minute I got into the cab, and I had to guide him. Then he kept peppering me with personal questions. Another cabby tried to convert me to Islam...
Waymo was quiet. Obeyed traffic signals. Took a rather tricky backroad to get to my apartment, and navigated it flawlessly. Tried to convert me to the Cylon god.
I'm gonna root for the robots on this one, I'm afraid.