Ark Invest Believes a Tesla Robotaxi Future is the Most Likely Scenario

ARK’s 2025 price target for Tesla is $3,000. ARK’s 25% bear case is for Tesla at $1,500 in 2025. The 25% bull case is Tesla could be worth roughly $4,000 per share. They published their model on Github and invite you to test your own assumptions and/or craft visualizations from the simulation outputs for our assumptions.

Ark does not model Tesla’s utility energy storage or solar business in our models. Nextbigfuture notes that Tesla making 500 Gigawatt hours of extra batteries in 2025 not used in cars would be worth another $500 billion in share value. They also have not modeled bitcoin assumptions in their model. For ARK’s work on bitcoin as corporate cash, please download our latest Big Ideas presentation.

Ark Invest has growing confidence in Tesla full self-driving becoming a robotaxi service. They think those scenarios are more than 50% likely by 2025.

SOURCES – Ark Invest, Met Kevin
Written by Brian Wang, (Brian owns shares of Tesla)

45 thoughts on “Ark Invest Believes a Tesla Robotaxi Future is the Most Likely Scenario”

  1. Thanks for your reply. Of course there will be some ownership. But they will affect the economics…you get knock-on effects too: think about houses without garages, how much more space you get, or cheaper houses overall.

    I would never say never, but that's just me.

    I think it may be at least competitive with ownership

  2. I'd like to see the source of those numbers, whoever they are they forgot that if your robo car drops you off at work, you don't need to pay for parking.

    and something tells me they didn't factor in all of a taxi companies expenses, (hint these companies aren't CEO free!) the price of ownership for most people in most circumstances is still cheaper than renting, leasing and taxi service.

  3. a service is never cheaper than ownership long term, I can see robo taxi's killing off or seriously changing leases, long term payment plans and the need for families & couples to own multiple vehicles but ownership will always be a thing.
    fact is, the middle man will always charge more than it is to own, always, even without drivers you must pay the salaries of those who work on or own the taxi, that extra overhead always makes it more costly.

    plus electrics are destined to become significantly cheaper to buy, run and maintain, the simplicity of electrics once the cost of energy storage drops, will increase the pool of potential buyers.

    think of it like this, cars have seats, climate control, electronics etc, much like your home does and people today can rent all of those same things.
    from air conditioning to radios, computers, phones, furniture etc. there is even a whole industry to support renting of these appliances, electronics and furniture to people and the cost of rentals is low, because you don't need to hire operators for any of these devices & appliances.

    but the majority of people today, still prefer to own these things instead of renting them, cars will be no different than these rental goods once the driver is eliminated.

  4. "…Why can't another neural network do it?" I believe it's because engineers haven't begun to reverse engineer the cerebellum and the other brain areas responsible for implementing predictive/anticipatory mechanisms.

  5. I'm not actually sure where the children thing applies. I wasn't the person with the camera, I just heard complaints about people not being allowed to eg. video the kids sports.

    Google "video children at school sports banned"

    give me quite a few hits.

    They seem to be saying it's not against the law per se, but is banned at many venues and is under discussion as to future law changes.

  6. I expect robotaxis, will eventually greatly benefit virtually everyone. Getting optimum legislation for this will be a huge task.
    I very much like the 6 passenger, GM Cruise Origin robotaxi that should be in production before 2023.

  7. Is that something local? Pretty sure parents take pictures/videos of their kids' sporting events and such all the time around here…

    It could be one of those laws passed to pile on charges when someone does something else wrong, or to give an excuse for nabbing someone when there's no other evidence of criminal activity?

    Maybe authorities could recognize that it should be criminal use of images, not the taking of images, that should be outlawed.

  8. Usually when discussing SD taxis people mean one or maybe two people in a sedan, carried nearly door to door, no other stops. Used that way, far more SD taxis would be required to cover the heavy rush hour traffic, then left mostly idle the rest of the day – parked somewhere.

    As I've written previously, if people would widely accept robo-van services – that would make more sense. But ride sharing and van-sharing are available today, and aren't used heavily enough to make much impact on traffic, despite the savings they can already provide.

    SD van sharing MIGHT gain acceptance if a software app scheduling service can eliminate any significant delays and other inconveniences due to sharing with other riders – maybe offering riders a trade-off between delays and cost, and never more than a single transfer.

    If people had to bid for the right to use highways, and there were only so many 'slots' available at a particular time (to minimize traffic jams) then people would have to pay enough more for a low-occupancy ride that they'd switch and everyone would benefit from shorter rides. But good luck getting THAT accepted!

  9. I know that it's already against the rules to photograph any group of children. Children are (currently) allowed to wander around all over the place, including public roads.
    You go past outside a school photographing the playground and you will very quickly be asked to explain yourself.
    You drive a car that automatically does the same thing, then as soon as the average punt realizes this we can expect the same response.

  10. As someone who has dealt with computers on a regular basis, I have to say that computers, and their failure modes, are a lot more stupid than you are.

    As someone on the internet you should already have dealt with them yourself.

  11. Airports. Bridges. Tunnels under the city? Past big government offices? Schools?

    Yes. It turns out that most people DO drive to places every day that a government might want to, or already does, prevent photos being taken.

  12. You assume that you'll be allowed to buy and register a personal car with that kind of vigilante economic action in progress?

    I would like to buy a car…. but banks will not deal with me because I'm on The List. No car dealer will accept $25k in cash (because of bans on large cash payments to stop crime.)

    I would like to pay for my registration, but the government office (online only of course) will only accept Visa, Mastercard, Paypal all of which have banned you.

  13. Personal cars are expensive costing about $0.70/mile, so a 10-mile commute costs about $15/day whereas a robotaxi 10-mile commute should eventually cost about $5/day. (And if you pay for parking, add another $10/day to your commute).
    During peak traffic, a robotaxi should have up to 6 passengers but typically about 3 or 4. During peak traffic, personal cars will have not much more than one passenger per car. When most cars on the roads are robotaxis, traffic congestion should reduce to near zero. This accords with the GM major objective of zero crashes, zero emissions and zero congestion.

  14. LIDAR isn't a "magic ingredient" that solves all SD driving problems, and Tesla's vision-based SD generally seems to do well on the aspects of self-driving that LIDAR contributes to – detecting and avoiding curbs and vehicles and pedestrians and other obstacles.

    Where Tesla SD appears to fall down so far is on difficult stuff like reading signs, correctly understanding poorly marked streets and intersections, reading other drivers' intentions, dealing with blocked lines of sight (which could affect LIDAR as much or more), etc.

    If/when Tesla SD relies on detailed street/lane maps for more than route-mapping, I'd expect a big jump in capability – possibly too big. If implemented now, it'd probably land Tesla SD cars in dangerous 'high level 3' territory – where drivers are most likely to get complacent, going long periods without needing to intervene and then suddenly something comes up and they're not paying enough attention to handle it. Also, it'd be a crutch that would prevent Tesla SD from getting real world training and learning from alert human drivers.

  15. Heh – I recall not so long ago police were stopping people from taking photos near "potential terror targets" like bridges. Maybe still do? Places Teslas probably drive past every day.

  16. Tesla SD appears to suffer from not building up shared street maps. Instead it appears to try to 'solve' every driving situation every time it is encountered. Good for training the system, not so good for everyday safe driving.

    E.g. in that video the car apparently thinks it COULD go straight through an iron bar fence, recognizing it late enough that it scared the driver into taking over (though it did look like it might have made the turn – several such close calls scared the driver even though the SD seemed like it was very good at not running into static objects).

    In another case it was turning up a hill onto a one-way and the left-turn-only sign painted on the far left lane weren't very visible at first, so the car got into that lane despite wanting to go straight – apparently planning ahead to turn left further up the road. That's a vision failure that street mapping might have easily fixed.

    Tesla SD also maybe interpreted tall grass on both sides of its single lane as indicating it was off the road – maps would have easily fixed the latter.

    Maybe Tesla will add shared maps as a 'safe mode' eventually?

  17. Because it will be cheaper and they will be orbiting your block. No tax, depreciation, insurance, driving effort, maintenance, cleaning, need I go on? The economics will most likely make car ownership a no-go at some point.

  18. Cars are extremely expensive now in a big city, and not preferred by a fair amount of people when good in-city mass transit options exist.

  19. Robotaxis are not all. There will be subscription models like CaaS that will replace ownership but allow for longer term possession than a short trip.

    Given those possibilities, the entire need to own a car outside of enjoyment/status reasons goes away.

  20. While Waymo has focused on urban driving, Tesla seems to have started at getting highway driving right and is 'learning' its way toward handling the much more varied case of city driving.

    That should give an edge to Waymo for urban taxi services, and to Tesla for commuting from suburbs. But commuting isn't a great taxi application, as demand is concentrated into a few hours a day – it seems a better match to personal cars.

  21. Robo taxi's will expand the taxi market for sure, but ownership will still be a thing. the number of cars owned will for sure be reduced, but never eliminated. I'm looking at that family of X with 2-3 cars making due with one robo car.

    but eliminating personal ownership for the average person? no. the average American is already capable of owning and paying for their own vehicle, why would they pay more for a service when they can own for less?

  22. Actually, the Tesla SW sort of mimics how the brain works. Most of the code is now ANN, not traditional C++ code. And with beta version 9.0, there will be almost zero C++. Did you not know this?

  23. I don't like the idea of becoming dependent on robotaxis if cancel culture continues to metastasize. "You have expressed wrongthink and hatecrime online. Your transportation account with RoboTaxisRUs is hereby suspended."

  24. If anything, Musk has become more humble over the years. Generally, he doesn't come across as overconfident or arrogant when speaking, quite often hesitant and stumbling over his words.

  25. China has ALREADY banned Teslas from a range of places such as military bases because they have cameras that can broadcast photos. Not robo teslas, which don't exist, just the normal Teslas that already have cameras on board.

    I imagine this would become the default for any robo or even camera+computer-with-web-link car that the local nation doesn't have complete understanding and control over.

    So, no taking your Tesla to a Chinese military base. Or airport? Or Xinjiang? Hong Kong???

    No taking a Chinese robocar to a US base. Or airport? Or the 3 blocks around the white house??

    Both would be banned from such places in Russia, or India, or wherever.

  26. But he is trying to solve it exactly with computer vision and ML.

    It's a problem that remains to be solved. Just because existing ML architectures isn't up to the task, that doesn't mean they will not be in a decade or 2.

    Human civilization is still on the primitive end of the ML development curve, no part of it is anywhere close to being mature.


  27. Yes that is true, it is different when living in major city. I myself live near woods and nature so i have different view on that and prefer/need own car.

    it is also true, that 200k people in Us work as taxi drivers, if you replace that with robo taxis, which could work almost 24 hours how much more do you need? 1 million of robo taxis working almost non stop making 60k miles per year with 0.2 cents per mile cut for tesla would be 12 billion per year for Tesla.

    I think that because of the age and competition. Model 3 is amazing car, looks good, electric, powerful, autopilot features,…

    Competition will be catching up and releasing new models. So when having the choice between 6 year old model 3 and nicely looking new model they will go after new one? Car companies usually put out next generation of particular model after 4-8 years, so after 5-6 years they should start developing new gen.

    If you look at model S and X sales and how much did they drop lately. You wont buy model S in 2021, so much other choices. I understand Tesla's mission is to bring ev's to the masses and they don't care so much about 70-140k price range.

  28. But he is trying to solve it exactly with computer vision and ML…

    I would change "imperil" to "blindly applaud" in your statement and get a better picture of what is going on.

    The guy overworked himself to the point of being mentally unstable. He is now jeopardizing Tesla and even Starship by making esoteric choices. Hopefully he will return back to Earth soon and start fixing things.

  29. Another big clue is that a computer does not work like a brain. I go with what the experts say and they say LIDAR.

    Tesla will star putting LIDARs if they don't want to be left behind in that game.

  30. I note that in what I watched so far the roads are dry & visibility is good. Rain fog snow & ice are further issues to get self driving cars to handle. This doesn't mean that self driving cars are impossible, just that there is a long way to go.

  31. LIDAR is unnecessary, you can create depth maps from camera systems. LIDAR is unnecessary. You can use it along with vision, but vision should be able to do the job without LIDAR. A big clue is that humans don't have LIDAR.

  32. I think in many cities, once robotaxis are cheap and convenient, the political support for policies that make private car ownership appealing will fade. Things like parking minimums in zoning, free parking, lack of congestion pricing, etc. If you look at cities like NYC, it becomes culturally quite normal not to have a car.

    Not sure why you think sales will plateau for Model 3/Y. As Tesla rides the cost curve down, they are expanding their addressable market. Of course, these cars will be refreshed eventually. Tesla is developing a new platform with large castings and cell to pack construction. The exterior/interior styling can be changed to keep it fresh, but it's probably not going to be needed for 5 years or so as they went with a very minimalist/modern style. If anything, they are probably setting design trends that other OEMs are beginning to copy.

  33. I can see countries requiring that robotaxi operators be domestically controlled, because they will become critical infrastructure. No chance that China would allow US firms to control their robotaxis or vice versa.

  34. And I think that robotaxis without a LIDAR are impossible

    I think you're wrong, computer vision is simply too unreliable with today's ML systems.

    Musk is too full of himself

    The consequences of the sycophantic devotion of legions of dilettantes and malcontents will continue to imperil all his works. It's not good for anyone's mental health.

  35. In their projections of the industry and those that may develop them in future, it's odd how many manage to exclude those that have operating Robotaxi services right now.

    The requirements of the kinds of support in use by existing Robotaxi services should be an indication of the limitations of current state-of-the-art constructs referred to as "AI". Current architectures are insufficient, robust unconstrained L5 requires further breakthroughs in machine learning.

  36. Robotaxis have big potential. Tesla could undercut uber and earn percentage of each mile driven. On the other hand there are limitations, and after some point market will become saturated. A lot of people like owning their own cars and don't want to rent them.

    I think Tesla has a problem, because they expect sales to keep climbing, but they still haven't presented any model 3 replacement.

    Its is about 8 years since they started producing model S. That is a lot and some minor upgrades are not enough. Sales dropped.

    It is almost 4 years since they started producing model 3. Sales will plateau and start dropping eventually, the model will be too old for people to buy. Small upgrades, refreshments won't be enough. For that particular reason, they should start to think about a new car or a new generation of model 3 in that particular price/fastback sedan segment.

    When releasing new generation of model 3, they will have to make major or at least visible changes to the outer body and so on. If you want to do that then you need to change the manufacturing as well and test it appropriately,..

  37. Robotaxis should become hugely popular before 2030 and by 2035 there should be more than 100 million Worldwide that have replaced nearly all cars, especially ICE cars.

    I expect it is more likely that GM, rather than Tesla, will lead the robotaxi revolution.

    It may be that Chinese companies will lead the World robotaxi revolution. To describe this as problematic would be an understatement.

  38. And I think that robotaxis without a LIDAR are impossible. Musk is too full of himself to realize that. This autopilot of his will be for a long time in beta and even if progresses, it would be nothing like he promises.

    The LIDAR is right there on the picture, just behind Boss Huang. See it?

Comments are closed.