Market/Economic Impact of FSD and Tesla Robotaxi Network

[neroden]

Doesn't Tesla have a substantial amount of FSD revenue that can't be booked until they roll it out? So, as features become useable and are released, they will add revenue to each quarter in which the features are delivered? In other words, a lot of people have already paid for FSD but the money hasn't been recognized as income yet?

I've always avoided accounting so I'm no expert!

Yes, for GAAP profit/loss under accrual accounting. But it's already been collected as cash. So for cash flow, any refunds would be cash out the door.

 

[PeterJA]

...Autonomy Day demonstrated that Musk is actually delusional.

This trolling might be taken more seriously if you explained how you know the inner workings of Tesla's autonomy project better than Elon does.

Also, your claim that Tesla Network will fail like Tesla's current customer relations because they are both "services" is unconvincing logic. Tesla Network will be software. Tesla does know how to write software.

SpaceX competitors sneered at Elon's plan to reuse rockets. They're not sneering now. I watched Elon tell a lecture audience: "Reusability is difficult, but we're going to try to do it." Now he says: Self-driving is difficult, but we ARE going to do it.

Show us evidence that he's lost his mind, instead of just repeating it over and over.

 

[StealthP3D]

FSD, though it will be another “moat”, is entirely too dependent on government agencies and lawmakers to count on in the near future. I liked what I saw on the autonomy day video so I’m by no means discounting its importance. However, I consider the Tesla Network a extremely profitable bonus.

I don't see FSD (autonomy) as being too dependent upon government. When it's ready (meaning it's been proven to be significantly safer than human driving), I don't see how bureaucracy could hold it back for long. It will be such a stupendous achievement with such concrete benefits that it would be criminal to hold it up for long. Auto insurance companies and health insurance companies will lobby hard to allow it. If the responsible agencies try to delay approval, the pressure will quickly mount and approval will not be far behind.

 

[jbcarioca]

OK. I'll bite. Limited robotaxis in a few tightly-geographically-bound areas with particularly well-marked street grids... will have zero positive financial impact on Tesla.

Insignificant at best... financially dangerous if Tesla assumes the insurance risk. Since Tesla routinely overestimates how good their systems are (they're still the best on the market, but they have excess optimism), there *will* be accidents, even in the geofenced case (probably due to really weird stuff happening), and the liability *must* not be Tesla's if Tesla is to remain financially stable.

The availability of "best in class driver assistance features" worldwide for car *buyers* will be substantially more meaningful than some geographically limited (and still unreliable) robotaxi scheme. But all it really does is allow Tesla to raise prices; since Tesla's production limited, it doesn't increase car sales.

Frankly your perspective and mine differ somewhat while largely agreeing on the overall trends and their potential direct impact on Tesla.

Several specifics are relevant:
Large cities have a vastly disproportionate share of every type of non-exclusive transportation. Your perspective might well underestimate the influence of such cases due to your living circumstances just as mine might overestimate the same. After all I have lived most of my adult life in major cities, while you have lived in less overpopulated areas.

As for the commercial impact of geo-fenced well documented major urban areas vs the larger world we differ substantially. Just as taxi usage is almost all centered in major cities (measured by number of trips, financial impact, number of vehicles) so the Tesla Network will have similar concentrations. I have not yet searched for documentation but could do so if anybody seriously doubts those assertions. Major urban areas are exceedingly well documented, and present the extreme activity concentrations that will allow for major supplemental efforts to cope with edge cases.

The most difficult edge cases tend to deal with ephemeral issues such as extreme weather, sudden traffic diversions, road accidents and spontaneous entry of people, animals or other foreign objects in the roadway. All of those are less difficult to treat when they are in more dense areas.

By contrast the current level of Tesla autopilot IME is brilliant in sparsely populated well defined limited access highways. That is wonderful, but has limited application to TN unless long distance trucking is included in the equation. Managing the edge cases in sparsely populated areas might well be more difficult precisely because of the relative sparsity of observations available to train NN.

If these questions are to be material for Tesla valuations within the next year or two we might actually do some research on the subject rather than speculating.

I participated in a substantial analysis of the future of transportation while at SRi in the 1980’s. Even in those ancient times we were all convinced that some form of autonomous taxi solutions would become ubiquitous in large cities by the 2030’s. It seems likely to me that our 50 year horizon probably will have been quite accurate. As usually with those old studies with which I was involved I choose to remember the prognostications that have proven accurate. Of course I choose to ignore the ones that did not...

but I cannot fail to recall that several of the people involved in the Transportation project also were providing HP with advice about the future for handheld calculators and concluded the market would be only engineers, who all could do better with slide rules. Oops! In 1971 I had a Bowmar Brain which did not help my analysis that told my then bank employer that there would never be a market for a ‘mechanical teller’. Oops again. Oddly both of my mistakes later provided major career opportunities for me.

Sorry for being slightly OT. However, we all shroud be very modest in our assurance that we see what will happen with FSD. Advances in machine intelligence of every type are happening far faster than we an absorb them. Similarly sensor advances are also happening with astonishing speed.

It is that history and current events that convince me that anybody with too much confidence and/or self assurance is likely to be missing something very important. One clear advantage Elon Musk has over almost all potential competitors is that he changes his mind when he finds clear evidence he’s been wrong. I prefer to admit my mistakes and learn from them rather than pretending I know things I do not know. Unless I find I am in error I remain quite confident in my views.

Finally, I will be astonished if the top 10 SMSA in the US do not produce >80% of TN US revenues and >100% of profits.

 

[DurandalAI]

That would be interesting to know how much unaccounted revenue there is for that, they've been selling it for ? years now. It's called the revenue recognition principle, cash can come in whenever, but the revenue doesn't get booked until the product/service is delivered.What is the Revenue Recognition Principle? - Definition | Meaning | Example

I believe that they will consider FSD as full feature released when it can do the following as merely a DRIVER ASSIST feature:

Stop at stop lights/stop signs and go at the appropriate time.
The car gets into the proper lane at intersections and can make turns into the correct lane ala NoA for in the city.
The car looks for oncoming traffic to determine whether or not it is safe to take an unprotected turn, such as a 2 way stop sign in a 3 or 4 way intersection.
The car can come pick you up from its parking spot in a parking lot to the front of the building.

I think part of the reason highway NoA and self-parking was moved to FSD was so that a portion of that $ can be realized as released and counted as income.

I think they will have FSD as fully released by Q1 2020, early access in Q4 2019 at the earliest. I don't foresee Level 4 for at least 2-3 years, and I don't think we will see level 5 until 5 years from now.

 

[StealthP3D]

Yes, for GAAP profit/loss under accrual accounting. But it's already been collected as cash. So for cash flow, any refunds would be cash out the door.

Are you saying this revenue has already been recognized by Tesla under their non-GAAP profit/loss statements?

As far as having to refund any payments for FSD I will point out two things:

1) Unlikely to have to return any because there was no deadline promised. Kinda like if you buy IPO shares from a company and they go down in value. The company doesn't have to refund the difference because there was no guarantee the company would perform well enough to make the stock go up.

2)The only way I see any possibility of any refunds happening is if Tesla isn't successful at developing FSD after the life of the car is expired (Average car life is 12 years I think). I don't see that as happening but, if it did, Tesla's production volume would be so large by then that revenues would dwarf any tiny amount they might have to refund. Remember, Tesla is no longer selling FSD as an add-on, access is now included at no extra charge with every car.

As to when Tesla will recognize revenue from FSD, I think that is likely beginning Q3. Because as they roll-out FSD features, they can recognize portions of FSD revenue. This could be a nice bonus to their earnings over the next two-three quarters. And I think Tesla is focussed like a laser on delivering positive earnings reports so they can be included in the SP500 sooner, rather than later. This move to the S&P was foiled by the poor Q1 performance caused primarily by the transition to international sales so the best is yet to come. Inclusion in the SP500 will be huge.

 

[DurandalAI]

2)The only way I see any possibility of any refunds happening is if Tesla isn't successful at developing FSD after the life of the car is expired (Average car life is 12 years I think). I don't see that as happening but, if it did, Tesla's production volume would be so large by then that revenues would dwarf any tiny amount they might have to refund. Remember, Tesla is no longer selling FSD as an add-on, access is now included at no extra charge with every car.

Um, no, Tesla is still selling FSD as an add-on. Autopilot is now included at no extra charge, but not FSD.

 

[StealthP3D]

Um, no, Tesla is still selling FSD as an add-on. Autopilot is now included at no extra charge, but not FSD.

Thanks, I stand corrected on that point.

I still maintain Tesla is unlikely to ever be forced to return payments for FSD, even in the unlikely event that FSD is not approved in the next 10 years. This is what they say before you click the FSD add-on:

The currently enabled features require active driver supervision and do not make the vehicle autonomous. The activation and use of these features are dependent on achieving reliability far in excess of human drivers as demonstrated by billions of miles of experience, as well as regulatory approval, which may take longer in some jurisdictions. As these self-driving features evolve, your car will be continuously upgraded through over-the-air software updates.

In other words, as long as Tesla regularly updates those who bought this feature with the latest and greatest results of their on-going development efforts, they have met their part of the bargain and will not have to offer refunds. This seems pretty obvious to me. The problem would arise if Tesla shut down their development program or stopped coming out with updates/improvements.

 

[abasile]

Common belief. Also totally false.

I really do not want to rehash this for the umpteenth time, it's actually *textbook* transportation modeling. It just flat out doesn't. Things like carshare sometimes displace at *less* than 1 for 1, and never much more than 1 for 1.

The two-word explanation for why is "rush hour". I do not want to repeat the long explanation why because I said it within the last week. If you get rid of rush hour (the Soviets tried, but failed -- but there are some communities which naturally lack rush hour for weird reasons), or serve it with high-capacity trains (which is often done), you can change this. However, in places with high-capacity trains, there are often a lot of people with no car at all, so when you introduce carsharing and they use it, it ends up increasing the number of cars on the road. Robotaxis would do the same.

I apologize for the rehashing, but now we have a new thread for this! I'll admit that, due to time limitations, I skipped and glossed over many of last week's posts in the main investor thread.

Yes, economical robotaxis will capture many riders who currently use mass transit. It's hard to beat the convenience of door to door transportation. Whether we see this as a good thing or not, it's coming sooner or later. One benefit for mass transit operators, though, is that they'll be able to stop running many of their under-utilized, non-economical routes/times since robotaxis will be able to pick up the slack. Many commuters may in fact be more motivated to use rail-based mass transit during rush hour if they know for sure that it'll be easy to take a robotaxi home when they need to stay extra late at the office.

In car-dependent areas, in the long run, families will reduce the number of cars that they own as robotaxis become ubiquitous and cheap. It's a simple matter of economics. I think it'll take some years, though, for people to become comfortable letting go of some of their vehicles. In the meantime, the total number of registered vehicles may indeed increase.

Maybe it'll never come. I mean, demand for Tesla purchases will keep being very high indefinitely. Who needs Tesla Network? Certainly not Tesla.

I agree with @jhm that to maintain high demand for its vehicles, Tesla must remain competitive in vehicle autonomy. Assuming FSD becomes good enough to run Tesla Network, I think Tesla would be foolish not to reap the benefits of what may initially be a virtual monopoly in robotaxis. But I suppose you're right - Tesla could just stick to selling vehicles and let others handle the "network".

OK. That sounds wrong to me, though. Do you know the law in this area?

What I know is that you can get refunds from a company even if the company was acting in good faith, if they simply never delivered the promised product. (Generally without interest or penalties.) The question would be, what was the promised product, which would involve reading the marketing text carefully.

While I have in the past carefully studied the law in other areas, and I have close family members who've practiced law, I'm no expert on consumer law.

However, when Tesla buyers purchase FSD (as we did for the Model 3), the associated text makes it clear that the timeframe for FSD is uncertain for various reasons. As long as Tesla continues making good faith efforts to achieve FSD with demonstrable progress, and provided FSD can be considered an attainable goal, it seems to me that it would be tough to convince a court to force Tesla to issue refunds. The one thing that could perhaps undermine Tesla's case would be Elon's optimistic public statements about the timing of FSD.

If anyone with real legal background in this area would like to chime in, that would be great.

 

[tander]

I believe that they will consider FSD as full feature released when it can do the following as merely a DRIVER ASSIST feature:

Stop at stop lights/stop signs and go at the appropriate time.
The car gets into the proper lane at intersections and can make turns into the correct lane ala NoA for in the city.
The car looks for oncoming traffic to determine whether or not it is safe to take an unprotected turn, such as a 2 way stop sign in a 3 or 4 way intersection.
The car can come pick you up from its parking spot in a parking lot to the front of the building.

I think part of the reason highway NoA and self-parking was moved to FSD was so that a portion of that $ can be realized as released and counted as income.

I think they will have FSD as fully released by Q1 2020, early access in Q4 2019 at the earliest. I don't foresee Level 4 for at least 2-3 years, and I don't think we will see level 5 until 5 years from now.

Is current autopilot consider part of the full self driving features? I wonder if they plan to make features available as they are ready or lump them all together at once. I don't know how many people have actually paid for it but say 400,000 have, that would be an extra 2B revenue they could put in whatever quarter people download it if isn't being accounted for on a feature by feature basis. They've been selling FSD for a year or two now haven't they?

 

[StealthP3D]

Is current autopilot consider part of the full self driving features? I wonder if they plan to make features available as they are ready or lump them all together at once. I don't know how many people have actually paid for it but say 400,000 have, that would be an extra 2B revenue they could put in whatever quarter people download it if isn't being accounted for on a feature by feature basis. They've been selling FSD for a year or two now haven't they?

On the Model 3, FSD was an option from the start. At some point (last year if I recall) Tesla changed the dividing line between AP and FSD. My wife's Long Range has some limited FSD features that are grandfathered into the AP feature set but the vast majority of AP sales do not include these more advanced features. I believe "Navigate on AP" is one such feature.

As to the timing of recognizing income, I imagine GAAP accounting rules require the recognition of that revenue in proportion to the value of the features as they are delivered and Tesla will make the judgment call as to what proportion of the value is delivered in each quarter's software updates.

 

[Maitri982]

I've a lot of personal examples to say this is not always correct.

I've also lots of examples where we thought the target was realistic but we still missed them.

Problem is when the team doesn't even try - because everyone thinks the target is unrealistic.

In terms of FSD, the problem is that its never been done and telling a developer team that it needs to be done by X is just making up artificial deadlines based on nothing. Its like telling a team they need to solve viable fusion in 6 months. You can set that goal but since no one has ever done it you are just making up an arbitrary deadline based on your your wishes and not based on anything else.

FSD is not a solved problem and may not be for a long time. They are clearly still experimenting with things and probably will be for some time.

BTW...This is very different than landing rockets on barges...also very hard, but actually much easier than FSD. The parameters of the environment for landing a rocket are fairly well known and will be within a relatively tight margin for each booster return. For a car driving the situational variation approaches a very large number and is therefore hard to even know what you don't know. And if you even know what all the situations are to be handled its not even clear *how* to handle them. Each may require its own NN or its own rules based approach.

The point is that Tesla and everyone else is still in the lab trying different things and under that environment pressing deadlines that are basically arbitrary can be defeating to the team. It is a delicate balance with such things though...you don't want them lollygagging around either, so they need some motivation. Pick your managerial poison.

 

[EVNow]

In terms of FSD, the problem is that its never been done and telling a developer team that it needs to be done by X is just making up artificial deadlines based on nothing.

Tesla is not right now solving FSD. They are trying to get to feature complete on features which Waymo (and Cruise) has already comepleted. We are not talking about getting to six 9s - just basic feature and some validation.

 

[neroden]

I believe that they will consider FSD as full feature released when it can do the following as merely a DRIVER ASSIST feature:

Stop at stop lights/stop signs and go at the appropriate time.
The car gets into the proper lane at intersections and can make turns into the correct lane ala NoA for in the city.
The car looks for oncoming traffic to determine whether or not it is safe to take an unprotected turn, such as a 2 way stop sign in a 3 or 4 way intersection.
The car can come pick you up from its parking spot in a parking lot to the front of the building.

I think part of the reason highway NoA and self-parking was moved to FSD was so that a portion of that $ can be realized as released and counted as income.

I think they will have FSD as fully released by Q1 2020, early access in Q4 2019 at the earliest. I don't foresee Level 4 for at least 2-3 years, and I don't think we will see level 5 until 5 years from now.

Seems like slightly aggressive marketing, but from looking at the materials... you're probably right. That would probably be sufficient to avoid more than token settlements. Though you know some guy is going to sue claiming that it wasn't "full self driving" because the car didn't drive itself halfway across rural Alaska while he was sleeping in the back seat. The question is how much money he'll get.

 

[tander]

...As to the timing of recognizing income, I imagine GAAP accounting rules require the recognition of that revenue in proportion to the value of the features as they are delivered and Tesla will make the judgment call as to what proportion of the value is delivered in each quarter's software updates.

If it's considered a product, they can recognize the full amount as soon as delivered, if it's a service it's usually spread out over the life of the contract. Makes me wonder if they'll start treating it as software as a service at some point since since it will probably need continual updates, also makes me wonder how much of FSD will be applicable to other things...like air, sea, and space...

 

[StealthP3D]

If it's considered a product, they can recognize the full amount as soon as delivered, if it's a service it's usually spread out over the life of the contract. Makes me wonder if they'll start treating it as software as a service at some point since since it will probably need continual updates, also makes me wonder how much of FSD will be applicable to other things...like air, sea, and space...

FSD is obviously a product that hasn't been delivered in its entirety (I think it only just began with the Nav on AP). So it will be recognized as the software updates are gradually delivered to the cars. It will be in Tesla's best interest to accelerate the recognition of this revenue as much as possible (within reason).

 

[neroden]

I think StealthP3D is correct. The actual marketing materials are pretty cautious about what they promise. As long as Tesla keeps adding new features, even if they *never* get full self-driving, it sounds like people wouldn't have much of a case to get refunds.

Good.

 

[TNEVol]

So if the AI taxi is registered with the Tesla Network I'd expect it to be an 'all in one' service, where the revenue sharing finances the commercial insurance. Even Uber, with an actual owner driving the car, offers insurance for the duration of the trip. (Uber drivers do have to insure their cars when there's no passenger in it though.)

There's no reason for owners to insure individually: they have no impact whatsoever what happens with an AI taxi, so why should they have to acquire insurance individually? The owners are no factor at all to risk levels.

It's not just nonsensical but also inefficient: if Tesla insures externally with one of the large re-insurance companies, they can insure all AI taxi owners as a group and reduce the per unit insurance underwriting costs significantly.

So Elon's answer that Tesla is going to insure cars in the Tesla Network while they are working for the Tesla Network is the only correct answer, the only open question is whether they'll re-insure or self-insure entirely. (They probably should as @neroden suggested.)

My guess is that Tesla will self-insure for damages up to a certain limit, and re-insure for tail risk liabilities above that. This would allow Tesla to leverage their considerable underwriting cost advantages via Tesla Insurance. As Tesla's capital reserves become bigger and bigger they can increase the re-insurance liability limit more and more, and above certain levels of capital reserves (above $100b and after the 10th Gigafactory) they'll be able to 100% self-insure.

FC as usual you are logical and as usual I admire your analysisu but your legal conclusion may be off somewhat. I am not a tort lawyer but liability for the owner of a robotaxi is not a tough stretch, especially if it is a business venture
Tort liability begins with a duty owed to the injured party. IMO ifor example it wouldn't be a hard sell to a jury that you as the owner of the robot taxi have a duty to inspect, maintain and warn about any known defects. If the car has a little too much dirt on the front bumper and causes the radar to react to ghosts by slamming on the breaks causing a rear end accident probable by data, the owner could be liable due to not washing the car. This frequently occurs now on my car. What will happen is owners and Tesla will be named in the suit and will have to defend themselves. Any decent tort lawyer will include enough allegations against the vehicle owner to make it through discovery. To meet the risks the.owner will at least need business liability insurance.
I agree that most liability will be on Tesla. Tort lawyers will be looking for the case where a toddler is seriously injured by a robotaxi when all the lawyer has to do is convince a sympathetic jury that TN knew or should have known about a defect or edge case. These might then be lumped into a class action that FSD is fundamentally dangerous. Emotion from the public is not easily overcome by statistics
Because of this emotional aspect I expect the first reaction of government will be to ban any use of FSD without a driver in control.
We had an example here recently. A drunken young tourist turned his re.tal electric scooter in front of a car and was killed. The immediate reaction of his family and the government was to ban the electric scooters. I'm not sure it will be very soon that insurance actuaries will feel able to calculate the liability which may be largely based on emotion not science.

 

[StealthP3D]

I agree that most liability will be on Tesla. Tort lawyers will be looking for the case where a toddler is seriously injured by a robotaxi when all the lawyer has to do is convince a sympathetic jury that TN knew or should have known about a defect or edge case. These might then be lumped into a class action that FSD is fundamentally dangerous. Emotion from the public is not easily overcome by statistics
Because of this emotional aspect I expect the first reaction of government will be to ban any use of FSD without a driver in control.

FSD (with no driver in control) will not be approved until it is statistically safer than human drivers. The banning of Autonomy, ironically, could be considered a serious crime due to the needless deaths the ban would cause. I disagree that the government would ban a service they had already approved unless new statistical data shows it is less safe than humans (that their initial analysis was flawed).

One dead baby, no matter how emotional of a reaction that might raise, is not going to cause a ban.

 

[ggr]

Mod: I have temporarily made this thread sticky so it's easier to find. But I expect to unstick it in a month or so. --ggr