Waymo

[Bladerskb]

LOL. NOA is better than FSD Beta.

Next.

Yet neither is an autonomous driving system? Why can't NOA or FSD Beta drive you from your house to the airport while you sleep? which you demand of Waymo? Why can't Tesla do anything you demand of Waymo?

Simple. Just deploy entire SF. Is there a reason for not deploying in entire SF - they have been testing since before my kids were born !

Simple. Just deploy entire SF. Is there a reason for Tesla not deploying in entire SF - they have been testing since 2015 (probably) before your kids were born!

Better yet, just deploy on one route, one entire street????

Why can't you start your own forum to crib about Tesla ?

You mad that Tesla can't do a single thing you demand daily of Waymo and may never will on current hardware?
While Waymo keeps advancing and you keep moving the goal post as they advance?

 

[diplomat33]

Some interesting tweets from Dolgov on the recent news of going driverless in SF:

https://twitter.com/x/status/1509169859657023491

https://twitter.com/x/status/1509169963575103493

@EVNow I think you wanted to know about unprotected turns. In the video clip in the tweet, you can see two unprotected turns that the Waymo did really well and confidently. And the 5th Gen I-Pace seemed to do the unprotected left turns more confidently than what we've seen with the 4th Gen Pacificas in Chandler IMO.

 

[diplomat33]

I know CEOs say a lot of marketing stuff but Dolgov tweeting that Waymo has entered the "next phase of acceleration and scaling" makes me hopeful that we will see Waymo accelerate their scaling a lot in the next few years. The fact that just 2 years after deploying driverless in Chandler, they are now doing driverless in SF and expanding their Phoenix area is promising.

 

[aronth5]

I know CEOs say a lot of marketing stuff but Dolgov tweeting that Waymo has entered the "next phase of acceleration and scaling" makes me hopeful that we will see Waymo accelerate their scaling a lot in the next few years. The fact that just 2 years after deploying driverless in Chandler, they are now doing driverless in SF and expanding their Phoenix area is promising.

People have to realize the autonomous driving space is huge and there will be lots of different successful solutions to meet different requirements. Haters of both Waymo and Tesla just need to chill. Let's encourage all vendors that take on this very difficult challenge. Solutions for driverless cars or solutions with drivers can and will both be successful.

 

[diplomat33]

People have to realize the autonomous driving space is huge and there will be lots of different successful solutions to meet different requirements. Haters of both Waymo and Tesla just need to chill. Let's encourage all vendors that take on this very difficult challenge. Solutions for driverless cars or solutions with drivers can and will both be successful.

Oh I agree. That is why I hope Waymo is successful in the ride-hailing space while I hope Tesla can be successful in the consumer self-driving car space. And even though I personally don't think the current hardware is good enough for L5, I think Tesla can still be very successful if they can achieve safe and reliable "door to door" L2 that works everywhere. And both products can fill an important role. Waymo's product can work great for people in select cities who need a ride, especially people who can't drive due to a disability. Tesla's product can work great for Tesla owners who want a system that takes most of the burden of driving away, especially on long trips.

 

[diplomat33]

This tweet is big. It confirms that Waymo has not just completed the 5th Gen hardware but also completed a "heavily rewriting software stack":

https://twitter.com/x/status/1509236262888169479

@willow_hiller This might explain some of those "software error" disengagements we were discussing since Waymo was in the process of a major software rewrite.

@Bladerskb I do think this is another promising sign for Waymo scaling. Now that they have completed new hardware and new software stack that works better, I think we will see Waymo expand faster now.

 

[Doggydogworld]

The fact that just 2 years after deploying driverless in Chandler, they are now doing driverless in SF and expanding their Phoenix area is promising.

I'm also encouraged by Waymo's recent steps forward. But let's be honest - they started driverless trips in Chandler in 2017. So it's been 4.5 years from first trip to first trip. Here are AZ milestones with matching "4.5 year" SF dates:

Late 2017 - First driverless "employee" ride -- March 2022 in SF (CHECK)
Early 2018 - First Trusted Tester driverless rides -- late 2022 in SF (LIKELY)
December 2018 - Paid public rides (Waymo One) with safety drivers -- mid 2023 in SF (TBD, need permit)
Early-mid 2019 - First paid Waymo One driverless rides (see September 2019 promo video) -- late 2023 in SF
October 2020 - Driverless Waymo One for anyone who downloads the app -- March 2025 in SF

The 3rd one is kind of meaningless, so let's focus on the last two. Late 2023 for paid driverless rides in SF sounds pessimistic, but experience has taught me to keep expectations low. Also, the gap between first paid public rides and "anyone can ride" was lengthened by COVID in AZ. On the other hand, it was easy to open to everyone in AZ because "everyone" was eight people (Just kidding. Kind of). Opening to everyone in SF creates huge logistics and cost issues. Dealing with those is not Waymo's forte, to put it nicely.

 

[diplomat33]

I'm also encouraged by Waymo's recent steps forward. But let's be honest - they started driverless trips in Chandler in 2017. So it's been 4.5 years from first trip to first trip. Here are AZ milestones with matching "4.5 year" SF dates:

Late 2017 - First driverless "employee" ride -- March 2022 in SF (CHECK)
Early 2018 - First Trusted Tester driverless rides -- late 2022 in SF (LIKELY)
December 2018 - Paid public rides (Waymo One) with safety drivers -- mid 2023 in SF (TBD, need permit)
Early-mid 2019 - First paid Waymo One driverless rides (see September 2019 promo video) -- late 2023 in SF
October 2020 - Driverless Waymo One for anyone who downloads the app -- March 2025 in SF

The 3rd one is kind of meaningless, so let's focus on the last two. Late 2023 for paid driverless rides in SF sounds pessimistic, but experience has taught me to keep expectations low. Also, the gap between first paid public rides and "anyone can ride" was lengthened by COVID in AZ. On the other hand, it was easy to open to everyone in AZ because "everyone" was eight people (Just kidding. Kind of). Opening to everyone in SF creates huge logistics and cost issues. Dealing with those is not Waymo's forte, to put it nicely.

Waymo has already started driverless rides for employees in SF. So I think Late 2023 for first driverless ride and March 2025 for driverless for everyone in SF is very pessimistic. I think it will happen before then. I think first paid driverless ride in SF will happen late 2022/early 2023 at the latest. And I think driverless for everyone in SF will happen in 2023.

You can't assume "4.5 years" timeline for SF since the pace will likely be faster. That is because Waymo has learned from Chandler so their deployment process is likely more efficient. Also, Waymo has better hardware/software in SF than they had when they started in Chandler. Waymo is starting driverless with employees in SF with a fully validated new hardware and software stack (5th Gen). That was not the case when they were started driverless rides in Chandler.

Also, see my last post. Waymo confirmed that they completed the transition to the new 5th gen hardware and a major software rewrite. So I think we will see Waymo scale faster now that they have completed the transition to their latest and best hardware/software stack.

I am optimistic. But we shall see soon enough in the months ahead.

 

[Daniel in SD]

Waymo is looking more and more like one of those dot com companies with no real plan on how to get to profitability.

Without all that Monopoly money, they would have gone belly up by now.

Or May developed a real business plan.

Yeah, they should just make robotaxis that work everywhere with no maps or remote assistance using <$1k in hardware. haha.
The plan is to make a robotaxi and then bring down the cost low enough to be profitable. As far as I can tell that's everyone's plan except for Tesla's (Is Tesla's FSD business profitable either?).
I wouldn't invest in Waymo though because I don't think there will be a big first mover advantage even if they are successful.

 

[Doggydogworld]

Waymo has already started driverless rides for employees in SF.

Yes, this week. That's why I said "March 2022 (CHECK)".

I think first paid driverless ride in SF will happen late 2022/early 2023 at the latest. And I think driverless for everyone in SF will happen in 2023.

You can't assume "4.5 years" timeline for SF since the pace will likely be faster.

Maybe, maybe not. They had shiny new "validated" h/w and s/w when they started in AZ, too. Gen 5 is better, but the technical problems in SF are harder. And the logistical problems are MUCH harder. So it's not at all clear they will proceed faster. Heck, if not for Cruise pushing them I'd half expect them to proceed even slower.

The downtown Phoenix announcement interests me more. It shows their first hint of entrepreneurial thinking. Also, it's absolutely necessary to ramp in multiple cities simultaneously (why I found "N+1" arguments here kind of silly). Waymo developing that ability now is the clearest signal I've seen of them thinking about this like a business since they "ordered" those 82k cars.

(Is Tesla's FSD business profitable either?).

Are you kidding? Selling vapor for 12k a pop deserves a prime spot in the business model hall of fame.

I wouldn't invest in Waymo though because I don't think there will be a big first mover advantage even if they are successful.

Robotaxis are a scale business with massive entry costs. First mover has a huge advantage in each locale. I just don't know if the business model works beyond a few dense cities.

 

[Daniel in SD]

Are you kidding? Selling vapor for 12k a pop deserves a prime spot in the business model hall of fame.

Tesla is investing huge sums of money into it though and the $12k includes other features that cost them far less to develop.

Robotaxis are a scale business with massive entry costs. First mover has a huge advantage in each locale. I just don't know if the business model works beyond a few dense cities.

But what are the fixed costs per locale? Mapping obviously, a location to service and clean the cars, what else?

 

[Doggydogworld]

Tesla is investing huge sums of money into it though and the $12k includes other features that cost them far less to develop.

Tesla has invested very little. They avoid anything that costs money, like good maps. They use old, cheap cameras. Karpathy has a small team. The upfront cost of chip development was paid back many times over in unit savings. It's a spectacular business model.

But what are the fixed costs per locale? Mapping obviously, a location to service and clean the cars, what else?

The biggest cost by far is the fleet. Fixed cost of stuff like a customer support/remote assist center and mapping are much smaller, but drive a certain minimum fleet size. But the real advantage of sScale is critical for quick response time. It also helps with utilization, which is key to amortizing fleet cost.

 

[Daniel in SD]

Tesla has invested very little. They avoid anything that costs money, like good maps. They use old, cheap cameras. Karpathy has a small team. The upfront cost of chip development was paid back many times over in unit savings. It's a spectacular business model.

Tesla spent $2.6 billion on R&D in 2021. What percentage of that was FSD? Their Nvidia supercomputer is definitely not cheap. Developing custom ICs is very expensive but I agree it's probably paid off versus buying Nvidia chips. The claim I've heard is that Tesla hasn't booked any revenue for the FSD L5 (pre April 2019), FSD "Automatic driving on city streets", or FSD "Autosteer on city streets" and I would argue that the vast majority of Autopilot R&D has been spent on those products.

The biggest cost by far is the fleet. Fixed cost of stuff like a customer support/remote assist center and mapping are much smaller, but drive a certain minimum fleet size. But the real advantage of sScale is critical for quick response time. It also helps with utilization, which is key to amortizing fleet cost.

I think we're in agreement. You do need a minimum fleet size for quick response times which means you need a big enough market share to get enough utilization. It does seem like robotaxis would have the same issue as Uber and Lyft where customers drive down prices by switching between them...

 

[diplomat33]

It does seem like robotaxis would have the same issue as Uber and Lyft where customers drive down prices by switching between them...

So far, we have not really seen any direct competition since robotaxis are still early and customers don't have a choice between two robotaxi services yet. The closest is perhaps SF where in some areas, customers might be able to choose between Cruise or Waymo. But I don't know if there is overlap between the Cruise and Waymo ride-hailing areas. Once there are enough robotaxis that customers do get to pick between different services, things will get very interesting. Competition does tend to bring down prices. But also things like wait time, convenience and amenities will factor in to customer's decisions of which service to use. I think that is one reason why we see AVs like Cruise and Waymo focus on things like the comfort and convenience of the ride. It is not good enough for the robotaxi to get to the destination without crashing. That's critical of course. But once there is widespread deployment, every AV will be able to reach their destination without crashing, so what will make the difference is comfort, wait time, convenience, amenities etc... The robotaxi that can provide a superior ride experience will win.

I also wonder if robotaxi companies might try pick different service areas, even in the same city, to avoid overlap and direct competition. After all, it might be more profitable to deploy in an area where you are the only robotaxi service available so you get all the potential customers for yourself, rather than deploy in a service area with other robotaxis where customers will be split.

 

[EVNow]

But what are the fixed costs per locale? Mapping obviously, a location to service and clean the cars, what else?

R&D.

 

[EVNow]

This tweet is big. It confirms that Waymo has not just completed the 5th Gen hardware but also completed a "heavily rewriting software stack":

https://twitter.com/x/status/1509236262888169479

@willow_hiller This might explain some of those "software error" disengagements we were discussing since Waymo was in the process of a major software rewrite.

@Bladerskb I do think this is another promising sign for Waymo scaling. Now that they have completed new hardware and new software stack that works better, I think we will see Waymo expand faster now.

That’s funny. I thought only Tesla rewrites software

 

[diplomat33]

That’s funny. I thought only Tesla rewrites software

Of course not. Everybody rewrites software at some point. It is a normal part of software development. Nothing wrong with that.

 

[EVNow]

Of course not. Everybody rewrites software at some point. It is a normal part of software development. Nothing wrong with that.

It’s only wrong when Tesla does it

 

[diplomat33]

It’s only wrong when Tesla does it

No. I am not saying it is wrong when Tesla does a rewrite.

 

[t3sl4drvr]

Robotaxis are a scale business with massive entry costs. First mover has a huge advantage in each locale. I just don't know if the business model works beyond a few dense cities.

I would expect "Waymo Driver" to become the largest product - not "Waymo One" robotaxi.

As soon as "Driver" really works well enough and with more subtle sensors, there will be a long like of car manufacturers who want to license it. This is a feature they just cannot live without. Just look how wide-spread Android Auto and Apple CarPlay are today.

"One" should be able to operate and win where-ever Uber has decent availability. 24/7 availability and 0 driver cost should pay back added HW quick.