formation flying is more difficult and risky than solo.
intercst
But why should any of us think he has “caught” them all now. It even begs the question “what is all”? I don’t think anyone including Elon has a meaningful basis for thinking the magical mileage to actually uncover all the edge cases out there is 10 billion miles. It could just as easily be 110 billion miles. I think Elon has been off base on way too much of his speculation around autonomy.
1 Like
But who cares? How is it an achievement that contributes to any real objective?
1 Like
Yep. And Elon is slowly adding vehicles to the unsupervised Robotaxi fleet as he reduces the risk of legal liability and reputational harm from an accident.
intercst
The goal is to catch enough edge cases to make FSD safer than a human driver. Elon thinks 10 billion miles is sufficient. You may have a different view. That’s what makes a market. You bet your money, take the risk, and suffer the consequences or reap the rewards.
Elon attracts a large army of skeptics and has a history of making them look foolish.
intercst
(not a Tesla shareholder, but I do drive a Model Y)
It’s been the reverse of that with AI, including vehicle autonomy.
Skeptics have been correct.
For a decade.
I invite you to make a prediction on vehicle autonomy.
So far none of the Tesla bulls here have made a prediction other than maybe “stock will go up.”
I have a few here:
What specific prediction for Tesla vehicle autonomy would you like to contribute?
1 Like
I believe that Tesla will achieve unsupervised FSD with the upcoming release of Version 15 to consumers. The software is in testing and validation now in the Robotaxi fleet.
intercst
1 Like
You are saying
- available for general public consumer purchase? a material number (not 100 vehicles)? or just robotaxi?
- Tesla accepts liability (they are the driver, after all)?
- in what geography? certain metros? nationwide (like the dream states)?
- Would you like to put a time frame on this?
It should be a material amount of unsupervised miles, not the marketing demos going on currently.
I mean they could get it eventually.
And they can continue their marketing demos endlessly. Remember driverless delivery? Chase cars?
PS: I admire and like your spirit to play Tesla prediction. I’ve been lonely.
2 Likes
But weren’t the accolades for V12, V13 and V14 pretty much the same? And haven’t they all disappointed thus far – including v 14 which is currently struggling? Why would we think V15 is going to change that? Adding more miles of test data doesn’t seem to accomplish much. Will elon ever stop and re-evaluate how the current design approach just simply doesn’t seem capable of getting there? Are we seeing an obstinate level of stubborness?
1 Like
Having enough data and having software that responds correctly to all that data is a different point in time.
Maybe that was a part of Elon’s rationale 10 years ago. But why has he not re-evaluated that stance in recent years?
I have always thought dropping LiDAR was a big mistake. I think it is less so now with more advanced cameras and the evolution of 4d imaging radar. However Elon may still be one of a small handful still hanging on to a belief that he will succeed via trying to brute-force his camera only myth now despite limited success in the best of driving conditions.
Despite Elon’s obstinance, LiDAR has grown and broadened its customer base inclusive of most other developers of autonomy. Having invested a lot of money in LiDAR companies, myself I can assure you that my stock in these LiDAR companies has greatly outperformed Tesla stock in recent years, despite bankruptcy of one holding, Luminar.
Why? Two reasons — the performance of the devices has dramatically risen and the cost has even dramatically fallen more. All leaders in actual Robotaxi deployments use the device. But a bigger reason in that costs have allowed a very serious rampup of LiDAR usage in vehicles only intended for L2 operation.
Despite very active monitoring of the LiDAR market, I can’t say definitively what the unit pricing is these days. The crazy high $$$ levels from sources that one might see need to be rejected. The biggest issue now is that the lowest costs are typically a function of large volume purchases which generally are not readily available. But it seems like a lower end L2 type sensor is around $200 in China whereas higher end LiDAR is more like $500. Prices from the non-Chinese resources both in US and Europe seem to be somewhat higher but still not bad – perhaps approaching $1000 for the high end automotive devices - typically long range forward facing devices – but still under $500 for rear facing and other device placement. All of these are still dramatically falling and LiDAR stockholders are reaping the benefits. Another development is that imaging radar has become a viable option in some cases to substitute for LiDAR though seeming not for the long range front facing LiDAR.
While Waymo costs are less available, the Waymo sensor array is way lower than what is often seen. The newer Ioniq5 vehicles are expected to be around $50,000 fully equipped with base vehicle cost around $30,000. All modifications for robotaxi use inclusive of sensors is only about $20,000. Forget the $75,000 nonsense for sensors.
Another vehicle, the newer Waymo offering via ZEEKR multi passenger van is expected to have full cost somewhere $50k- $75K
Edit: Base cost of Waymo Hyundai was supposed to be $30,000 - now changed
A lot depends on the definition of all … like the march of 9s.
Yes - and this is where the Cybercab basically broke Tesla’s robotaxi business model.
Tesla’s largest competitive advantage was going to be that it had a small enough equipment cost (using just camera) that all of that stuff could be equipped standard in their fleet of ordinary consumer vehicles. The main cost of making a Waymo car isn’t the cost of the sensors - it’s that it’s incredibly expensive and labor-intensive to do all the after-market installation of that sensor suite.
However, the fleet sizes are still too small to make it worthwhile to do a full production line version of their car. I don’t know what the minimum annual run would be for that kind of production line to make sense, but I assume it’s close to 20K per year. Waymo doesn’t need that many taxis yet - they have nowhere to put them. So they still have to run with “artisanal hand-crafted” vehicles that are modified from OEM vehicles (even if the OEM is willing to put some of the internal wiring and whatnot inside).
The “Tesla Network” would have allowed Tesla to avoid that entire “artisanal hand-crafted robotaxi” phase of the process while you build up the network to a large enough scale that you can mass-produce cars for it. You never need to hand-craft your taxis. Even better, you have all these slightly used cars lying around that will cost much less than a new car, driving your costs down even further.
By switching to the Cybercab, Tesla’s given all that up. Just like everyone else, they’re going to be using new cars. Just like everyone else, it’s going to be ridiculously costly to do a production run of new cars at very low volumes - until they get up to that 20K (or so) level of need, their per-unit vehicle costs are going to be astronomical.
Now that the cost of LiDAR components have fallen so far, when Waymo or other companies do reach the scale where a mass-production run makes sense, their per-vehicle costs are going to be only a modest incremental cost over that type of car without the sensor suite - maybe only a few thousand dollars more. Not enough for Tesla to have much of an advantage on pricing. So the real competition will be on how well the software stacks work, and how much live personnel have to be used to staff the support network that will be required for each geographic area for the self-driving to operate.
3 Likes
Speaking of LiDAR, take a look at what is soon available in Canada, from China:
Level 3 with integrated LiDAR.
I invested in the Xiaomi ADR when they first announced they would be offering their EV in early 2024. When it dropped well off its 52 week high of $40+ I doubled that original purchase despite substantially raising my combined cost basis. Overall, the doubled position is doing well and I would recommend it to anyone, though each has to make their own call about investing in China.
Xiaomi offers Hesai LiDAR across all trims of SU7 sedan and YU7 SUV. I have never seen any plan to offer L4 or L5 autonomy. They have received a testing permit for Level 3 right now but to my knowledge are still actively in the testing process.
Regarding Hesai LiDAR, they are probably the worldwide leader in LiDAR now for automotive markets. I highly advocate their shares too with the same caveat about China.
Xiaomi is yet to sell outside of China but have already said they will start offering their cars in Europe next year.
2 Likes
I think the whole “Tesla versus Waymo” is the wrong way to look at it. For a few reasons:
- It is not reasonable to assume that there won’t be other entrants in the field. There are already Chinese companies doing autonomy, and Mercedes is working on it, and no doubt others as well.
- Tesla, just the vehicle business, it entirely different than Waymo to begin with. Tesla manufactures and sells vehicles, AND might someday also manufacture and deploy autonomous taxis. Waymo only retrofits and deploys autonomous taxis.
- Seems like Tesla is slowly backing away from the vehicle business at least to some extent. Perhaps they don’t consider it “future” as much as they used to, or perhaps they figure the margins won’t be good enough, or perhaps it’s just a whim of the CEO.
As far as predictions, it’s very difficult to make any reasonable predictions of things that are paradigm changes (oh how I hate that phrase, but it is the only one that is apt here). It was difficult to impossible to predict the transition from horses to cars, it was difficult to predict the transition into the Internet age, and it is currently hard to predict the transition to autonomous vehicles. About the only prediction I could make it that it is very likely that there will be a transition to autonomy someday. But I can’t predict when, and I certainly can’t predict who the winner and losers will be. Heck, the “winner” could easily be a company that doesn’t even exist yet.
1 Like
- Tesla vs. Waymo only looks at the surface minutia, not at the whole picture. Most businesses are one pony shows. GM and Ford are car maker and have been for close to a century. Amazon, by contrast, morphed from book seller to retailer, to AWS, and to transportation giant.
- Tesla is on a similar path, from roadster to mass market cars, to charging infrastructure, to storage, AI, lithium refining, to…, well, you get the picture, I hope.
The Captain
Over 70% of Tesla’s revenue comes from vehicle sales and leasing. It is first and foremost a car company. Everything else is essentially a side hustle, for now.
1 Like
There are plenty of companies who have branched out using their original raison d’être to leverage into other fields and done so successfully. Disney: from animation to motion pictures to theme parks to cruise lines; Amazon from books to everything to data centers to logistics, to mention just two. Usually it makes sense if one leads to another to another, but not always. GE and Westinghouse pivoted from electricity into all sorts of things: aircraft, shipbuilding, nuclear power, and so on.
Musk’s things are not really related, except by force fit after-the-fact (*not that there’s anything wrong with that). But electric cars did not lead inevitably to rocket ships, and underground tunnels have zero to do with social media but so far he’s juggling them all pretty well. The AI thing might link the cars and the rocketships, maybe, somehow, and the robotics and other stuff. Not truly related, but that could happen; Westinghouse started making radio sets but ended up owning broadcast stations. GE, too. And then cable outfits and satellite backhaul and all sorts of other stuff, so never say never.
That said, the car business is in a mildly precarious place. China is coming. FSD is a horserace which Tesla is not, so far, leading. Robotaxi is an unknown both in scope and operability.
Yup. But it’s wayyy too soon to declare the winner, but arguing about it is fun.
1 Like
Did I say otherwise?
Didn’t I mention Amazon?
What’s your point?
The Captain