Latest metrics in US robotaxi competition, free market at work
Number of Taxis
Tesla 10ish
Waymo 1500+
Number of L4 Autonomous Taxis
Tesla 0
Waymo 1500+
Number of Metro Areas
Tesla 1
Waymo 4 (L4 only)
Service Area
Tesla 40 sq miles
Waymo 700+ sq miles
Number of L4 Miles per Week
Tesla 0
Waymo 1+ million
Availability to Public
Tesla “invite only”
Waymo “general public”
Hours of Service
Tesla 6am to 12am
Waymo 24 hours daily
Way more crashes:
Just saying…
The Captain
Rather a silly comparison days after Tesla has started … not to mention careful wording to make the comparison more dramatic.
I respectfully disagree.
Those are all perfectly fine metrics of a robotaxi service and L4 is the autonomy goal.
I don’t see how this list of factual metrics is in any way silly.
It is a true representation of current state.
The future may differ, but as of today, it’s how things are.
Do you have an additional metric we should consider for the robotaxi business for which we have data for Waymo and Tesla?
I’m open to other relevant data.
I agree… as long as they are ‘real world’ metrics and not some bureaucratic regulatory designation.
The Official NHTSA ‘L’ levels are 'bureaucratic regulatory’ and have no actual, real world FUNCTIONALITY.
We’ve all heard about high school athletes who ‘skipped’ being validated in the ‘college farm team system’ and went straight to ‘pro’, in Basketball, Football, and Baseball.
We’ve all heard of the ‘teen/s’ who is a world class software coder and went to Microsoft, Google, Apple, etc, sometimes even without a high school diploma.
They ‘jump the line’ without ‘paying the dues’ cause they have the actual, real world functional ability.
And… sometimes are wildly successful.
I keep seeing the ‘but Tesla is not L4’… as in does not have the NHTSA seal of approval for L4.
Tesla/Elon explained 3 or 4 years ago, that they were not going to get NHTSA L4 regulatory approval, until they/Tesla saw the benefit of jumping through that hoop.
Tesla/Elon explained that being NHTSA certified L4 has the downside of being ‘regulated as to how many cars could be ‘deployed’ and ‘where’ the cars could be deployed, etc. By delaying getting NHTSA L4 approval, Tesla had much more freedom.
Tesla was developing L4 functionality, while still only being ‘regulatorily approved’ for L2.
I agree with @Tamhas (not @pauleckler). Saying Tesla is not L4, merely cause the NHTSA approval has not been granted… is silly.
Ok, ok, fine. The ‘Levels’ are the ‘industry approved descriptors’ for the different ‘milestones’ from basic driver assist to full AV, in the real world functionality sense. And this is useful.
I argue that Tesla is being criticized as “not L4” from the NHTSA regulatory approval POV. Not the real world ‘level of functionality’.
Is there ‘real world’ information/data that suggest Tesla has ‘L4 (real world’ functionality? IMO, yes - all the miles and miles of Teslas driving itself, while having the L2 required ‘human driver’ paying attention in the driver’s seat.
There are (apparently. If the YTs are ‘true’) many many many examples of ‘end to end’ self drive with no human interventions.
These YTs are individually derided as ‘anecdote’… but, at some point a whole bunch of datums becomes DATA…
ralph
macroeconomically, having those ‘outliers’ jump the line without getting ‘official certification of ability’ has greatly benefitted human development.
Today I counted 30 Teslas and not a single Waymo!
The Captain
Depending on how you define the start date. Robotaxis were in Tesla’s Master Plan Part Deux back in 2016 and announced as being right around the corner a number of times since then. So it isn’t like Tesla is just getting started on this. Robotaxis have been a core part of their business strategy for a decade.
Regardless of the start date, Tesla is measurably behind Waymo. There’s no other way to slice it. Tesla might catch up, but right now they are behind.
There is no formal NHTSA “Level 4 certification” process for autonomous vehicles. Tesla is operating its ride hailing vehicles with a human safety operator in the vehicle. That’s Level 2 by definition.
Most jurisdictions require proof that AVs are safe for ride hailing. A notable exception is Texas. Texas does not require special proof of AV safety and Texas does not require a safety driver. Logically, Tesla’s robotaxi service in Austin is using safety operators because Tesla thinks they need them. That is certainly prudent on launch, but it is easy to see that Tesla is measurably behind Waymo in this area.
Here’s what ChatGPT says:
{ NHTSA does not have formal federal regulations that specifically “govern” SAE Levels 1–4 as distinct categories. Instead, it approaches vehicle automation through a mix of standards, voluntary programs, and regulations that apply broadly but can impact different SAE levels in practice }
So… A difference without a distinction?
As they say?
Which also allows Tesla to continue to avoid whatever bureaucratic regulations comes with “no safety driver”.
Tesla, from my POV likely already HAS FUNCTIONAL L4.
YMMV.
ralph
I thought the human in this case was in the passenger seat.
Consider two alternative scenarios.
Scenario 1
Assume:
Tesla has the data that clearly shows that their AI driver, when unsupervised by a human safety watcher, has a safety and performance record about as good or better than the Waymo AI driver.
What should Tesla do?
Shouldn’t they follow their business plan and commercialize their product into robotaxis and L4 personal vehicles?
Yes, right?
They should follow their plan.
But they are not doing this.
Instead, they are testing a tiny sample of vehicles in a geofenced area for 18 hours of the day to an invite only ridership.
Scenario 2
Assume:
Tesla does not have the data that clearly shows that their AI driver, when unsupervised by a human safety watcher, has a safety and performance record about as good or better than the Waymo AI driver.
What should Tesla do?
They probably should not yet commercialize their product into robotaxis and L4 personal vehicles.
Instead, they are testing a tiny sample of vehicles in a geofenced area for 18 hours of the day to an invite only ridership.
Tesla’s actual behavior is way mo consistent with scenario 2 (maybe there is another scenario that explains lack of commercialization of L4 product, a top business goal, I’m open-minded to ideas).
I’m pretty sure, given what we know about the company, that they are not shy about promoting any progress, so if they could really do it, I think they would be doing it: removing safety monitor.
If they are careful with measuring safety and performance, then when Tesla removes the safety monitor that is a good sign that they have the data to support that decision.
There is big downside risk for them to getting that wrong. They need customers to trust the product.
I agree 
%.
Consider scenario 3. Tesla has a plan for the roll out, and they are following that plan.
For me, that’s the KISS explanation.
ralph
I also believe that FSD and the robotaxi market WAS Tesla’s to lose.
The longer time it takes, (and rising anger toward all things AV, EV, AI, etc) the more likely Tesla will lose a significant portion to other AV technologies.
The race is getting tighter.
NVDA is backing companies working on alternative AV technologies.
Here’s one ( I own a piece):
Aurora Innovations.
Huang, NVDA CEO, has said he thinks AV technology will use simulations to train it’s models. Ie, no real world driving data. No time “wasted” no $ wasted on real world data collection.
Elon has said they’ve reached the limits of real world data and will focus on simulations (of increasingly rare “edge cases”).
Some Chinese company reportedly has a $200 lidar tech.
Grok 3 says: { Hesai Technology, a Chinese LiDAR manufacturer, offers the ATX LiDAR for advanced driver assistance systems (ADAS) at a price under $200, starting next year. This is a significant reduction from their current AT128 model, which costs around $400. }
At least those are the 
I’ve heard around the YTverse.
ralph
The part I thought was important was your statement:
But there is no NHTSA L4. Tesla doesn’t have L4 in the way L4 is commonly defined and understood in the industry. Waymo does. That really is a difference.
In Texas though, they don’t have those bureaucratic regulations regarding a safety driver. Not required for AV ride hailing in Texas.
But those regulations do apply in other jurisdictions–notably California. If Tesla wants to expand its robotaxi model outside Texas it will have to jump through those hoops eventually.
It is. That is why I said “safety operator” not “safety driver.” But it is important because it is difficult to scale if each robotaxi requires a safety operator.
Clearly, Tesla wants to get rid of the safety operator, but Waymo was able to do that four years ago.
I argue that bureaucratic regulations have no bearing on whether the vehicle is functionally fully autonomous or not in a real world, practical application.
My investing premise is that Tesla either already has fully autonomous capability or is very close.
Tesla is currently fine tuning it’s software, crossing the t’s n dotting the i’s.
The “Tesla delivers itself to a customer” story, IMO, supports my investing premise that Tesla either already has or is very close to fully AV capability. Ie “true” FSD.
I’m confident that Tesla will remove the “safety driver/operator” when it suits Tesla to do so.
I’m confident that Tesla will expand it’s range, using geofencing tech and then just going “bareback”, when each step suits Tesla.
I’m confident that Tesla will use “First Principles” strategies that release incremental upgrades, each one a step toward true FSD.
Why am I confident? Cause I see Tesla already doing these things.
ralph