Fair enough, but “assume Tesla solves L4” is the crux of the problem, in my view, and it’s certainly the prerequisite.
It’s the hard AI robotic thing that people haven’t done before.
Manufacturing and fleet operations are things people have done for a long time.
Sure, manufacturing and fleet can be further optimized, but in this thread and many other threads, it’s been argued that the economics of robotaxi are not great, even with cheap vehicles.
But those operations are all irrelevant without first doing the new, big, difficult technical AI thing: L4 autonomy safely over many million miles.
So assuming Tesla achieves L4 autonomy in a scalable and timely way, is a huge premise.
And so far, the evidence shows they are struggling, and so that assumption is struggling.
It is but let’s move past that since it’s not relevant to my specific premise.
Again, I think this plays into my argument that it then puts heavy pressure on the businesses ability to manufacture at increasing lower and lower margins. Telsa has demonstrated that ability as the first new US car company in nearly 75 years and with margins better than most other auto manufactures. Musk has further demonstrated this with gutting nearly 85% of X and still growing the business and reducing launch costs by nearly 80% at SpaceX while others still struggle to compete.
Yes it is. Regardless, the broader point may be that autonomy is just one battle and the war is won by the company that has the advantage to scale. As great as Google is, it’s had the luxury of operating in a high margin business where it can blow capital in lots of different areas. Capital intensive businesses require far more discipline and this is where Musk’s almost maniacal focus on vertical integration, efficiency, and starting with the premise of eliminating something rather justifying its existence, may pay dividends. I think WAYMO thinks about solving autonomy but not how to make it profitable. Tesla thinks about how does it deliver profitability if it can solve autonomy.
I don’t think that’s true. At all. Mostly because the obstacles to scaling are largely going to be unrelated to the technology, and also I think you’re overestimating the limitations of Waymo’s (and others’) tech.
Even if you had “solved” full autonomy (and it’s not binary, see below), you’re still not going to be able to identify a new city on a Monday and stand up 500 cars to service it by the following Wednesday. You’ll need a pilot, because you don’t know the traffic patterns or demand cycles or consumer behavior of that city. You’ll need to identify and acquire (mostly via short-term leases) a number of properties within the city for storage, charging, and cleaning the vehicles - as well as to be a physical dispatch for IRL tending to their needs on the road. You’ll need to hire people to do all that. Then when you’ve done that for six months or so, you’ll need to do it over and over again as you expand into new areas of the city - new data on usage patterns, new properties for storage etc., new employees. And, of course, the permitting and licensing and all that takes time as well.
That stuff is absent in the Uber/Lyft model. It would have been absent in the Tesla Network. It’s 100% real for anyone running a fleet operation. So none of them are going to scale like we’re used to tech companies scaling.
So Waymo doesn’t really have much in the way of a disadvantage here. Sure, they do city-specific mapping more than Tesla does - but that’s all something that can take place simultaneously with all the other stuff. And as noted above, “full autonomy” isn’t going to be a binary - Telsa’s still going to need to ground-test and adjust their autonomy package to identify weird idiosyncratic issues in a new city as well.
As for “manufacturing prowess,” that’s a complete non-issue. Waymo’s partnered up with Hyundai and Geely. By the time any of these companies are ready to scale up to many thousands of cars, rather than the many hundreds (Waymo) or many dozens (Tesla) of today, those companies will have stood up a factory for Waymo that can produce as many as they need on whatever timeframe they need. We’re well more than a decade away from when these companies will need more than the 50K or so annual cars that a company would need to support a fleet of 200K. Maybe - just maybe - Tesla has some special sauce that gives them an edge on pricing of those cars (though I highly doubt it). But no advantage on scaling.
And of course, you can’t scale without building the brand. Ride-hailing isn’t selling cars. People will actually invest their own time into researching cars before they buy one, so just having a good product can get you into consumers’ headspace even without advertising (though Tesla’s recent performance shows the disaster that not managing your brand can become). But no one’s doing that for their ride-hailing apps. Once you get past the pilot-intro phase, it’s going to take time to build consumer demand for your product, which slows your scaling as well. And because Uber and Lyft exist and do more or less exactly the same thing that a Waymo/Tesla TaaS do, you’re not going to be able to just have word of mouth.
Which is a long way of saying that Tesla doesn’t really have any scaling advantages over any of the other competitors, because the technology in the cars isn’t going to be the main scaling bottleneck at any point when it matters. The only way that’s different is if Tesla actually were able to deliver on the Tesla Network model, and it’s now pretty clear that’s going to be impossible for them to do for close to a decade or more.
I never claimed that Tesla would solve autonomy in add on cities in a matter of days. Assuming all other things are equal (admittedly they may not be), roll out that does not require continuous mapping and the ability to manufacture the vehicles and operate them at lower margins, has a scaling advantage.
When you have outsourced the manufacturing, you have automatically introduced additional margin costs and supply chain dependency that is beyond your control.
I don’t think individuals are going to buy ride hailing vehicles. The manufacturers will operate them or sell leasing agreements of the technology to third parties. Consumers will care little about which service they use, assuming they meet the safety and reliability expectations. In that sense, the service becomes a commodity and the low margin producer likely wins. Do you care which yellow taxi you get into in New York?
I suspect Uber or Lyft may sit as a platform on top of these services. Consumers can then choose driver or driverless, size of the vehicle and other aspects.
Again, I don’t think that’s true. Continuous mapping doesn’t seem to have materially slowed Waymo’s rollouts, and it’s very likely that all the other things that have to be done slowly when entering or expanding a market are going to be much slower than the mapping. The mapping won’t be the bottleneck. Nor is there any reason to think that Tesla has any advantage in manufacturing or operating the vehicles at lower margins.
So what? That only matters if the margin costs or supply chain dependency are going to be relevant constraints. It’s highly unlikely that they will be.
Not at all. Which is exactly the point. I use Uber and Lyft because I know they exist and are on my phone. There are probably a dozen or more ride-hailing apps that I could use here in Miami. I don’t use them because I don’t know they exist. For all I know they might be cheaper or lower-margin services. But I don’t know that they are, and am not going to invest any time or effort into finding out.
Low margin producer doesn’t likely win unless they can convey that lower margin into a competitive advantage. You need to build your brand to get that information to consumers, and Tesla (famously) can’t be botherd to do that.
Plus, again, the “lower margin” here is probably negligible. I think Tesla won’t have much of an advantage on cost of production (they don’t have any better operating automotive margins than Hyundai does), and the cost of production will be a very tiny part of the overall operating costs of this service. It’s going to come down to cost of operations, not the cost of the cars.
Maybe, but that’s not the way things are shaking out so far. Waymo’s building their own service, eschewing their former partnerships with Uber. Tesla 100% should partner with Uber - and if they did these marketing concerns would vanish - but they seem to be unwilling to do so as well.
I think you are taking a very narrow view of how these businesses may operate in the future and the significance of the low cost per mile provider in terms of consumer decision making, which is why margins matter. Do you really think if robotaxi went fully operational tomorrow and provided a ride hailing app, it would need a marketing and brand development program to give it exposure?
Further you’re applying contradictory ideas as evidence to support it. WAYMO’s progress in its rollout of autonomous vehicles city to city is evidence its winning the game. Tesla and Musk’s history of manufacturing expertise and where WAYMO has none or has to outsource it, is not an advantage for Tesla.
The cost of operations will include the replacement cost of vehicles. It also will include replacement cost of parts and systems. The more complex those systems the more costly it is to maintain them. The more whole vehicles or parts are externally manufactured, the more likely there is an increase in the marginal cost.
I don’t know what the operational costs will be for Tesla. Starlink is suppose to provide some real time data services as part of robotaxi, eventually. I could see where that adds to Tesla’s advantage in terms of operational efficiency, unless WAYMO is working on something similar. But, it’s really all quite theoretical at this point. Neither of us really knows what the operational cost per mile will be for the businesses.
Nearly every car on the road today, including our ICE car, has cellular connectivity. And it does not have Starlink. Or, as I like to call it, N@ziCon.
Well, regardless of marketing, and speaking objectively, there could easily be 1/3 of US population that would not use Tesla robotaxi for standard use, regardless of product or price, because of the damaged brand.
So there’s 1/3 of TAM gone right to start.
Pretty sure that’s a larger impact than any of the operational excellence mentioned in this thread.
I sort of suspected that might work its way into the discussion. I don’t think people care that much about it, not in the long run. VW built vehicles for the N&zi Regime and there was once a substantial anti-Japanese auto bias in the US. The more important aspects that drive most consumer behavior decisions ultimately won. History may not repeat itself but it tends to rhyme.
Yes. Which is why Waymo is advertising. Because everyone already has a ride-hailing app on their phone. And again, no robotaxi is going “fully operational tomorrow.” Every city is going to have pilot programs, followed by small rollouts in limited services areas, followed by gradual expansions - all over the course of a year or more. You need more than media coverage.
Tesla is not only bad at brand management - they don’t do it at all. Which, again, you can maybe get away with for a while in selling automobiles, when people will research before a purchase on their own and are sensitive to small percentage changes in cost. That’s not going to be true for ride-hailing. A consumer will care if a car is 4% cheaper than a competitor (OTBE), because they will care about saving $1,500 on their purchase - so they’ll find you if you have a margin advantage. That consumer doesn’t give a fig if you’re 4% cheaper on a ride hail, because they don’t care about saving $0.23 on their taxi ride - and they’re not going to research ridehail services to see which one is 4% cheaper.
Yes, because the former is what matters and the latter isn’t. Tesla isn’t partnering up with anyone to manage the tedious process of getting support sites or assessing market patterns or all that stuff. And they’re behind in doing it, which means Waymo’s going into every city first and getting the better spots and building their brand. But on the other hand, Waymo is partnering with established auto manufacturers who are perfectly capable of supplying everything that Tesla can do. And once more, it’s the operational side - not the cars and software side - that’s likely to determine the pace of scaling.
Again, it’s not likely to matter much at all. The cost difference between a $35K vehicle and a $40K vehicle spread over 300,000 miles is two cents per mile. Give or take. It’s going to be swamped by cost differences on the operational side of things. And even if it isn’t, no one’s going to make their ride-hailing choice based on saving twenty cents compared to the other one.
Starlink is nearly useless for this application. It requires a clear view of open sky to function. Any time a car goes under a bridge/overpass, or into a downtown area with high buildings, or even a highly-landscaped roadway, they’ll lose contact. For urban environments, Starlink offers no advantages over conventional ground-based telecomms.
My intent when making an investment is a holding period of forever, so long as the thesis remains intact. Further, the world is a very big place and opportunity.
consumers will drive half way across town to save 5 cents a gallon on gas. I flip all the time between lyft and Uber to find the cheapest ride.
Not sure what to tell you, every successful business tycoon going back to Carnegie and Ford understood the advantages of pulling every aspect of the operation in house, where it financially made sense to do so.
I don’t think anyone has suggested, even the most optimistic predictions, the WAYMO vehicle will come in at that price point.
Co-Pilot Response
Tesla’s Starlink Advantage for Autonomous Robotaxi Operations vs. Waymo
Tesla’s Starlink integration gives its robotaxi service a unique operational edge over Waymo by enabling real-time, low-latency communication and remote monitoring that can significantly improve safety, efficiency, and scalability.
1. Real-time remote monitoring and teleoperation
Starlink provides a private, high-bandwidth satellite internet connection for Tesla’s robotaxis. This allows Tesla to remotely monitor vehicles in real time, send corrective commands, and perform teleoperation when needed Reuters**+1**. In contrast, Waymo’s vehicles, while also teleoperated, rely on terrestrial networks that can be congested or unavailable in remote areas. Starlink ensures continuous connectivity even in regions with poor cellular coverage, which is critical for safety-critical interventions.
2. Faster fault resolution and safety interventions
With Starlink, Tesla can detect anomalies, reroute vehicles, or take control instantly without waiting for ground-based network handoffs. This reduces downtime and improves safety margins, especially in complex urban environments or adverse weather WebProNews. Waymo’s multi-sensor, high-precision mapping approach is robust, but it still depends on terrestrial infrastructure for updates and control.
3. Support for OTA (Over-the-Air) updates and AI optimization
Tesla’s Starlink-enabled fleet can receive OTA updates for its FSD software while driving, allowing continuous AI model refinement without physical depot visits Gate.com. This enables rapid optimization of autonomous behavior based on live data, which is a key part of Tesla’s “hyper-exponential” scaling strategy. Waymo’s updates are typically more controlled and slower due to its multi-layered safety systems and regulatory scrutiny.
4. Scalability and cost efficiency
By using Starlink, Tesla can avoid costly terrestrial infrastructure in new markets, reducing deployment costs and speeding up expansion. This aligns with Tesla’s goal of scaling to “half the U.S. population” quickly Reuters. Waymo’s approach requires extensive city mapping and infrastructure investment before autonomous rides can launch, which is slower and more expensive.
5. Data and safety synergy
Tesla’s billions of miles of real-world driving data combined with Starlink’s real-time monitoring creates a feedback loop for AI training and safety validation digitalhabitats.global. This allows Tesla to fine-tune autonomous decisions in live conditions and address edge cases faster than Waymo’s more conservative, rules-based approach.
In summary:
Tesla’s Starlink advantage lies in real-time remote control, instant fault resolution, continuous OTA updates, and reduced infrastructure dependency. These capabilities complement Tesla’s camera-only, AI-driven FSD system and give it a faster, lower-cost, and more scalable path to autonomous operations compared to Waymo’s sensor-heavy, map-dependent model. While Waymo has a proven safety record, Tesla’s Starlink integration could accelerate its ability to deploy and optimize robotaxi services across a wider range of environments.
Some consumers do. Most don’t. And while you’ll flip between the apps on your phone, I suspect you’re not continually searching for new ridehail apps that you don’t already know about to find a cheaper alternative.
Yeah - where it financially made sense to do so. Apple doesn’t manufacture their own phones. Nike doesn’t manufacture their own shoes. Even Tesla has suppliers - they don’t make everything in house.
Waymo isn’t making their cars in house because it’s going to be years and years before they or any other robotaxi company needs enough new cars for it to make financial sense to be their own manufacturer. Probably never. No reason at all not to have a company that knows how to make cars just manufacture your cars for you, just like Apple has Foxconn make their phones.
And I don’t think Tesla can make their vehicles at $35K, either. Certainly not a five-passenger car, and they’ll need a bunch of those to supplement the two-seater coupe.
The numbers are there to illustrate that the amount of difference that Tesla might capture because of alleged manufacturing margin advantages due to vertical integration over other suppliers like Hyundai or Geely would be relatively meaningless. Tesla might have some margin advantage over Hyundai…but thinking they’d have 10 full points of margin would be ridiculous. Hyundai’s got the same gross margins as Tesla automotive right now. So if it turns out that Tesla is correct that a small two-seat coupe is the way to go, rather than a standard sedan, and we reach the point where its the correct move to make any car without steering wheels or pedals or any driver controls, Waymo’s going to be able to pivot and get basically the same car at somewhat the same cost - and if they lose a few points of margin, it won’t matter much.
Hey - it’s fun to just post AI stuff without distilling it into an argument! Here’s what I get (this is from ChatGPT)
Cellular is actually better in cities
Ironically, the places where robotaxis will earn the most money are exactly where Starlink performs the worst.
Dense urban areas have:
Excellent 4G/5G coverage
Numerous nearby cell towers
Low latency
Continuous handoffs
Starlink excels where cellular coverage is poor:
Rural highways
Deserts
Mountains
Remote communities
Those are not the first markets most companies target for robotaxis.
Tesla has another possible advantage
Tesla is reportedly working on smaller, automotive-oriented Starlink terminals.
The consumer dish is far too large and power-intensive for widespread vehicle installation, but SpaceX has already developed flat, electronically steered antennas for aviation and maritime use. An automotive version could eventually fit beneath a glass roof or body panel.
Even then, physics doesn’t change:
Trees still block signals.
Parking garages still block signals.
Tunnels still block signals.
Dense downtown skylines still reduce satellite visibility.
My expectation
I would be surprised if Tesla’s long-term communications architecture were “Starlink instead of cellular.”
A more likely approach would be:
95%+ of communications: cellular (5G/6G)
Supplemental connectivity: Starlink where it provides an advantage
Driving decisions: entirely on the vehicle, independent of either network
That architecture combines the strengths of both systems while avoiding dependence on continuous satellite visibility.
***** End AI cut and paste ------------
So, no - no reason to think that Starlink would materially affect robotaxi service.
According the the investigation, that’s not what happened.
The accident resulted from a panicked pedal pushing driver who reacted to some squirrelly FSD action. If you can override FSD by slamming your foot on the accelerator, that’s a problem.
Not usually a huge fan of it. However, it seemed to concisely lay out the arguments for it. The more credible thing to do would be to address why the summary is wrong rather than the author of it.
I disagree. The human needs to be able to override the machine. Though I am surprised that emergency braking didn’t kick-in. Our EV has it, and -so far- it’s engaged three times. Each time it wasn’t necessary (i.e. I wasn’t going to hit anyone/anything), but it slams the breaks and is very startling. Our EV would see a house as a big stopped truck, and if I didn’t hit the brakes I’m pretty sure it would have.
Humans are far from infallible, but we are motivated by survival (in general). We can generally be relied upon to react to ensure that survival. As we’ve seen multiple times, automated driving -usually due to faulty or incomplete input- can smash into things. Whether it’s a house, or a semi-trailer turning in front of the vehicle (several years ago).
If I couldn’t override some auto-drive feature, I’d never turn it on in the first place.
I did. I posted the summary prepared by ChatGPT on why the argument is wrong. Starlink offers no advantage over terrestrial telecomms services in the areas where robotaxi services are likely to operate. Those are dense, urban areas that are heavily served by terrestrial telecomms infrastructure and have numerous and frequent obstructions that prevent a clear-sky view during the course of a trip. Under such circumstances, even if SpaceX developed a Starlink dish suitable for a passenger car (itself debatable!), it wouldn’t be advantageous in any of the likely robotaxi markets.
It’s just not an advantage. It doesn’t do anything that ordinary cell service can’t do in an American city of any size. One can imagine that if you wanted to offer robotaxis in remote exurban environments, or in developing nations that don’t have robust terrestrial cell service, that might be useful. But those markets are so unsuitable for robotaxis that the possible availability of Starlink doesn’t affect the business case for robotaxi in the slightest.
The failure of Tesla to make the hardware in existing Model 3/Model Y cars actually work for FSD has fundamentally changed the business case of autonomy for Tesla. It still can be labelled as “FSD” or “robotaxi” or “autonomy,” but what it is and the potential value to Tesla has been completely altered. IMHO dramatically reduced - the best reasonable case for Tesla now is vastly less valuable to the company that what they were aiming for even a few years ago. And apropos to this thread, that change in the business model has all-but-eliminated any of the advantages that Tesla might have had to distinguish it from competitors in the field.
I’m not suggesting that humans shouldn’t be able to override the machine. I’m trying to imagine a scenario where slamming on the accelerator would be the most effective means to override the system, and should therefore be part of the design.
But then again, when Tesla allows drivers to select a “Mad Max” speed profile, maybe shat is supposed to get outta hand.
