Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[Antares Nebula]

Please read post #60384 of this thread.

Edit: I'd copy/paste but I'm on mobile

Thanks. I would say that is a significant amount of selling in the last six months. Over 100,000 shares, totaling about $25-30M (remember Musk bought about that much in the recent offering, and was seen as a vote of confidence, even though he is worth much much more than JB). JB still has 300,000 or about $75M worth left. But again there could be many reasons for the sale. I never said that JB no longer has faith in the company (as some posters here are mis-interpreting my words). As I said above,

Let me clarify, JB is still a believer and holds a signif amount of shares.

But I don't want to belabor the point about JB and shares sales. That was only an ancillary point to my larger argument:

Perhaps you've found something. I can't say without knowing more. But can I ask WHEN you came to know this info and WHEN you think Musk became aware of this.

The reason I ask is that money talks (and BS walks). If Musk was aware of this fact by the recent capital raise (around May 5th or so), then Tesla could have raised equity at a much higher share price by revealing this info to a select group of investors with an NDA. Even otherwise, with such a silver bullet or trump card, there is no way the share price would have dived to 180, much less break 300 (after all, you're willing to double down here at 250 based on this knowledge). Tesla, Musk, and upper mgmt would have used this knowledge to secure capital and support the share price. It would have been extremely foolish and reckless not to.

Also, you would not see JB and other executives selling their shares in droves. And possibly there would not have been as many executive departures. (Not to say that they don't believe in the story, but what you're talking about is a sliver bullet.)

 

[powertoold]

Well, you were saying that the execs were dumping shares in droves. But based on JB's selling habits, it seems like he's just gradually reducing his stake, regardless of share price. And it only really became a story once the share price was really low.

FSD is just one reason why Tesla shares may skyrocket in the next 3 to 4 years. Even if Tesla does achieve FSD, it doesn't mean the Tesla Network would be successful. There could be many other challenges to implementing a driverless ride-sharing network that are outside of Tesla's control or anticipation.

 

[Antares Nebula]

There is no evidence that JB is reducing his shares for an exit. As far as I can tell he still has around 350000 shares. Every month around the 27th, he exercises options to buy 15k shares at around $31 and sells some or all at the current price. He has been doing this for years. I think it is just a set plan to cash his shares for income.
${Instrument_CompanyName} ${Instrument_Ric} Insider Trading | Reuters.com

I agree. I never said that.

 

[Antares Nebula]

UPDATE: Tesla Announces Date for Second Quarter 2019 Financial Results and Webcast | Tesla, Inc.

"UPDATE: Please note the call is being moved by one hour to 3:30pm Pacific Time (6:30pm Eastern Time).

"Tesla will post its financial results for the second quarter of 2019 after market close on Wednesday, July 24, 2019. At that time, Tesla will issue a brief advisory containing a link to the Q2 2019 Update Letter, which will be available on Tesla’s Investor Relations website. Tesla will hold a live question and answer webcast that day at 3:30pm Pacific Time (6:30pm Eastern Time) to discuss the Company’s financial and business results and outlook."

 

[kengchang]

UPDATE: Tesla Announces Date for Second Quarter 2019 Financial Results and Webcast | Tesla, Inc.

"UPDATE: Please note the call is being moved by one hour to 3:30pm Pacific Time (6:30pm Eastern Time).

"Tesla will post its financial results for the second quarter of 2019 after market close on Wednesday, July 24, 2019. At that time, Tesla will issue a brief advisory containing a link to the Q2 2019 Update Letter, which will be available on Tesla’s Investor Relations website. Tesla will hold a live question and answer webcast that day at 3:30pm Pacific Time (6:30pm Eastern Time) to discuss the Company’s financial and business results and outlook."

Short Short : Tesla is using that extra hour to cook the book! FRAUDDDDDDDDDDD

 

[mrmage]

Perhaps you've found something. I can't say without knowing more. But can I ask WHEN you came to know this info and WHEN you think Musk became aware of this.

The reason I ask is that money talks (and BS walks). If Musk was aware of this fact by the recent capital raise (around May 5th or so), then Tesla could have raised equity at a much higher share price by revealing this info to a select group of investors with an NDA.
...

It came together recently for me, but I have no idea when it did for EM. I assume that he and his team must have found the same or better a long time ago, because they are are a sharp bunch immersed in autonomy.

 

[CorneliusXX]

Yes but not everyone wants to use it as such. It's like forcing a high mortgage because you can rent out your spare bedroom to make it up.

No, it's more like a house that's for sale goes to the highest bidder - and the highest bidder is the one who knows they can make extra money from Airbnb. If Tesla doesn't sell their cars at the market rate it's like selling houses at a below market price. We know what happens when homes are sold at undervalue - they get flipped by smart investors for excess profits.

Do we want Tesla to be selling their house for less than they can get? Probably not.

 

[jeewee3000]

It came together recently for me, but I have no idea when it did for EM. I assume that he and his team must have found the same or better a long time ago, because they are are a sharp bunch immersed in autonomy.

If you're not going to spill the beans, then don't post at all.

Now you're just making vague statements that benefit no one. Vague statements don't net you bragging points.

If you want to tell us "I told you so" by the end of 2020, then tell us at least one of the following:

1) What have you figured out that makes you change your mind about the near term possibility of FSD?
2) When will Tesla's FSD package be feature complete at te earliest/latest?
3) When will the Tesla Network be active?
4) Have any talks between Tesla and regulators (*cough California cough*) leaked?
...

If your insight only consists of: rise in FSD pricing + enhanced summon leaks + enhanced summon release in August, then I call Bravo Sierra.

Be that as it may, I wish you well in your investments in TSLA.

 

[Antares Nebula]

If you're not going to spill the beans, then don't post at all.

Now you're just making vague statements that benefit no one. Vague statements don't net you bragging points.

If you want to tell us "I told you so" by the end of 2020, then tell us at least one of the following:

1) What have you figured out that makes you change your mind about the near term possibility of FSD?
2) When will Tesla's FSD package be feature complete at te earliest/latest?
3) When will the Tesla Network be active?
4) Have any talks between Tesla and regulators (*cough California cough*) leaked?
...

If your insight only consists of: rise in FSD pricing + enhanced summon leaks + enhanced summon release in August, then I call Bravo Sierra.

Be that as it may, I wish you well in your investments in TSLA.

I kinda have to agree with this. Perhaps the intentions were good, but what useful info did @mrmage provide? At worst, it is BS. At best, he/she has discovered a valuable insight but is unwilling or unable to share it (but how fortunate for him or her).

Who is this person, and what do I know about him/her? Not much. So the gist of it is that some random internet poster claims to know something quite valuable about FSD, without revealing exactly what that is. What is an investor to do with this? Absolutely nothing. Anybody else here could claim the same thing (or the opposite, that FSD will fail), but without facts or details (or prior credible history) it is completely unactionable. I can only file this post in my "ignore" drawer.

 

[sparcs]

I kinda have to agree with this. Perhaps the intentions were good, but what useful info did @mrmage provide? At worst, it is BS.

Waste of time. Just set to ignore and move on. Anyone can come in and make grand proclamations that they have some secret source of information about TSLA. Are they giving away free bitcoin like Elon on twitter?

 

[KarenRei]

Sounds like sensible diversification to me. An “Act of God” event could hit Tesla and going from having $100 million+ To essentially nothing would be horrible for anyone.

One could also bother mentioning that, like Elon, JB is involved in multiple companies. He's been funding the development of a battery recycling company in Nevada (Redwood Materials). He's not an executive there, but he is its bankroller.

Believe it or not, suppliers don't want to be paid in "stock options in some other company". They want to be paid in cash.

I do think JB is setting himself up for conflict with Tesla in the future. He probably envisions Redwood working as a partner to Tesla at some point in the future, the contracts grow huge, and eventually they get acquired for a lot of money. Musk however likes to do everything in-house, and I have trouble picturing him supporting a policy platform that amounts to giving away so much to JB. Right now Redwood says they have no plans to work with Tesla, but... I mean, it's the elephant in the room; they're both even in Nevada, for crying out loud.

 

[Antares Nebula]

Elon Musk tweeted about Graham's number the other day. (Elon Musk on Twitter)

But just this snippet from Wikipedia is quite revealing:

"Graham's number is.... so large that the observable universe is far too small to contain an ordinary digital representation of Graham's number, assuming that each digit occupies one Planck volume, possibly the smallest measurable space. But even the number of digits in this digital representation of Graham's number would itself be a number so large that its digital representation cannot be represented in the observable universe. Nor even can the number of digits of that number—and so forth, for a number of times far exceeding the total number of Planck volumes in the observable universe."
Graham's number - Wikipedia

That's really large folks! What business do mortals have tinkering with such numbers? They are meant for the realms of gods and angels and other immortals.

 

[NakedLong]

And my reply is that your assumption that "FSD car = being able to sleep in it unattended" is a very complex milestone that is not the real economic "FSD car" milestone. The real milestone is "can the car drive itself safely without passengers, to drive to the next customers or back home."

Defining milestones and gating conditions that unlock 'professional Tesla Network FSD revenue' is essential if we want to talk about the probabilities of these developments.



Actually, the "car can drive itself unattended with no passengers" is what Tesla effectively demonstrated to select investors at the Autonomy Day in a limited but complex urban environment, and with safety drivers.



And I say that "what we need for FSD" has to be precisely defined, and I did that, and my definition is a much more achievable milestone than the "sleep in the car" FSD milestone you mentioned.

You have to agree that the difference between "sleep in the car FSD" and "no humans in the car FSD" is significant, right?

To me this idea of sleeping in the car or no humans in the car are imprecise ways of looking at the problem of FSD.

Discussions about FSD need to talk about the reliability. It is quite easy to get to feature complete FSD in 1 year, but that might have a reliability of only 99% which is far below an acceptable level. Increasing the reliability is exponentially harder. The progress to something like 99.99 will probably be rapid, but those last few 9s are going to take huge amounts of work. A better way of thinking about it is in terms of error rate. For example, to go from 99.99 to 99.9999, the error rate will go from 0.01 to 0.0001, a 100 time reduction in error rate. Each 9 is a tenfold reduction in error rate. So it becomes very important to know what reliability is targeted for public release.

Elon Musk said this (from BI): Elon Musk says a 'massive effort' is required to get Tesla driverless cars to '99.9999%' safety

I have read a lot of different number of 9s, and don't actually know what is a realistic cut off error rate for a robo-taxi network, so I will do some calculations here:

From Tesla's latest safety report:

> In the 2nd quarter, we registered one accident for every 3.27 million miles driven in which drivers had Autopilot engaged. For those driving without Autopilot but with our active safety features, we registered one accident for every 2.19 million miles driven. For those driving without Autopilot and without our active safety features, we registered one accident for every 1.41 million miles driven. By comparison, NHTSA’s most recent data shows that in the United States there is an automobile crash every 498,000 miles.*

US drivers have one crash about every 500,000 miles, so their reliability = 1 - 1/500,000 * 100 = 99.9998%.

So if Tesla were targeting 99.9999% reliability this would be a halving of the number of accidents going from an error rate of 0.0002% to 0.0001%. To me this 99.9999% reliability, or one crash every 1 million miles seems like the minimum reliability to release FSD, and it is inline with Elon's figure of 99.9999%. Note that the figure of 1 accident every 3.27 million miles is not really relevant to the discussion of FSD because this is a combination of human+AI which can perform better than just human or just AI.

Taking all this into account, I can only assume when people talk about the binary condition of having FSD or not, they are talking about passing a reliability level of 99.9999%. Of course even getting to 99.99% reliability will seem amazing, but until you reach the 99.9999% level you will still need a driver. My general feeling after reading this thread is the people don't appreciate the difference between 99.99% and 99.9999%. So when Elon said feature complete in 2019 I guess that will have a reliability of anywhere between 99.9 and 99.999%. And when he said robotaxis in 2020, this would imply he thinks in 2020 they can pass the 99.9999% reliability level.

Using this 99.9999% definition of FSD is still very rough. It does not take into account many factors, and is just a ballpark figure, but it is still much more accurate than 'sleeping in the car' or 'no humans in the car', which imo do not convey much.

TLDR; FSD needs to be defined as >99.9999% reliability or 1 million miles per accident.

 

[Fact Checking]

(Note to moderators: this is not off topic in the investor thread I think, defining when FSD can be deployed is key to when and in which jurisdictions it starts generating revenue for Tesla. This is a major, dominant factor to TSLA valuation in ARK's thesis too for example. )

Will allow this discussion for now. But let's NOT turn it into the umptieth discussion about the technical feasabilty of FSD - mod

TLDR; FSD needs to be defined as >99.9999% reliability or 1 million miles per accident.

That's probably still too high an accident rate though: it's only twice as good as humans, and with a current fleet of 600,000 FSD-capable cars and 15,000 miles driven annually that's 9 billion miles driven per year, or 9,000 million miles driven with 9,000 FSD accidents in a single year, or 25 FSD accidents per day - half of which I assume are at-fault accidents.

That is quite likely way too large even if the large majority are fender-benders and not high energy collisions or accidents with personal injury, and half are due to some other driver being at fault.

My minimum threshold guess is one at-fault FSD accident every 10 million miles, or ~10 times better than the human driver.

But the type of at-fault traffic accidents matters too: fender benders versus accidents resulting in personal injury matter a lot to liability. @neroden made the argument that the primary metric is going to be legal exposure, as probably no insurer is going to be willing to underwrite the exposure.

Regulation matters too, not just in terms of allowing FSD operation, but in terms of setting liability limits for FSD cars, the FSD implementation was approved by a government agency and there's no gross negligence by the manufacturer. No such laws exist currently in the western world that I'm aware of, but they might exist in the future as larger and larger segments of the economy move to FSD. (Such liability limits might also exist in China already.)

 

[Chocochip]

US drivers have one crash about every 500,000 miles, so their reliability = 1 - 1/500,000 * 100 = 99.9998%.

Sorry, but that’s not how it works. Your calculation — and the related discussion — assumes the average driver can drive safely for 500,000 miles and then have one crash… for one entire mile? Like, one mile-long crash? Because that is the necessary assumption when you use distance as your statistical parameter.

Rather, the reliability of a driver, whether human or AI-based, should be estimated as the number of correct driving decisions (meaning decisions not resulting in a crash) divided by the total number of driving decisions. It’s an absolute value rather than a relative one (that is, human vs. machine).

However, given how granular that estimation becomes (there may be thousands of small wheel and speed adjustments over a mile driven in a fairly complex urban environment, or even on highways), it is much easier to simply compare the accident rate of the average driver vs. an AI-based system, and for that you can use miles driven without incidents and do a direct comparison. For instance, if the AP system (assuming it’s running without any human imput) drives on average for 1 million miles before an accident occurs, you can say it is twice as safe as the average human driver. Assuming the comparison is made for the same type of driving conditions, including type of road, weather, road condition, traffic, etc. But none of that 99.999% reliability stuff.

 

[Fact Checking]

Sorry, but that’s not how it works. Your calculation — and the related discussion — assumes the average driver can drive safely for 500,000 miles and then have one crash… for one entire mile? Like, one mile-long crash? Because that is the necessary assumption when you use distance as your statistical parameter.

No, traffic accident statistics generally use "miles driven", and the number simply means that if a driver is driving 500,000 miles then the average number of traffic accidents he will be involved in during those 500,000 miles driven will be around ~1.

Edit: as you later point out I misunderstood your argument and only quoted a small part of your post out of context.

 

[Chocochip]

No, traffic accident statistics generally use "miles driven", and the number simply means that if a driver is driving 500,000 miles then the average number of traffic accidents he will be involved in during those 500,000 miles driven will be around ~1.

Does it make mathematical or statistical sense to you to describe that as 99,9998% reliability? Because that’s what my point was. If you read the rest of my post you discover that I agree this can be used as a comparison for relative incidence of traffic accidents, but it shouldn’t be expressed as that type of percentage. You correctly point out that an accident rate can be expressed as 1 accident / x miles driven. See how the unit is “accident/miles”. A percentage has no units.

 

[PaulusdB]

Does it make mathematical or statistical sense to you to describe that as 99,9998% reliability? Because that’s what my point was. If you read the rest of my post you discover that I agree this can be used as a comparison for relative incidence of traffic accidents, but it shouldn’t be expressed as that type of percentage. You correctly point out that an accident rate can be expressed as 1 accident / x miles driven. See how the unit is “accident/miles”. A percentage has no units.

But what about % of accidents/kilometer then? The same 1 accident per 0.5M miles would be 1 accident per 0,8M kilometers...

 

[dww12]

quarterly stats are in for safety.
Tesla Vehicle Safety Report
again, I'm not entirely pleased with the statistical meaningfulness here, as it certainly does not yet capture apples-to-apples comparisons (all cars vs. new cars, highway vs. all around driving, etc.).
but it's something. and I consider it investor relevant as they are building a long-term anti-FUD set of data that hopefully improves (not just the data, but the appropriateness of the data) over time.

I wish they would define Autopilot Engaged as Autopilot Engaged within the last 3 seconds of a crash. Many times you dis-engage AP right before it does something stupid. Those should count as on AP, if they already don’t.

 

[ReflexFunds]

(Note to moderators: this is not off topic in the investor thread I think, defining when FSD can be deployed is key to when and in which jurisdictions it starts generating revenue for Tesla. This is a major, dominant factor to TSLA valuation in ARK's thesis too for example. )



That's probably still too high an accident rate though: it's only twice as good as humans, and with a current fleet of 600,000 FSD-capable cars and 15,000 miles driven annually that's 9 billion miles driven per year, or 9,000 million miles driven with 9,000 FSD accidents in a single year, or 25 FSD accidents per day - half of which I assume are at-fault accidents.

That is quite likely way too large even if the large majority are fender-benders and not high energy collisions or accidents with personal injury, and half are due to some other driver being at fault.

My minimum threshold guess is one at-fault FSD accident every 10 million miles, or ~10 times better than the human driver.

But the type of at-fault traffic accidents matters too: fender benders versus accidents resulting in personal injury matter a lot to liability. @neroden made the argument that the primary metric is going to be legal exposure, as probably no insurer is going to be willing to underwrite the exposure.

Regulation matters too, not just in terms of allowing FSD operation, but in terms of setting liability limits for FSD cars, the FSD implementation was approved by a government agency and there's no gross negligence by the manufacturer. No such laws exist currently in the western world that I'm aware of, but they might exist in the future as larger and larger segments of the economy move to FSD. (Such liability limits might also exist in China already.)

I agree, exactly what crash rate Elon considers "safe enough" for Robotaxis is a very important consideration for Tesla's valuation.

It's important to note the NHTSA data actually tells us there are 1.8 cars per crash on average so the actual NHTSA reported crash rate per car per mile is 498/1.8 = 1 per 277k miles. This is also a significant underestimate because many minor crashes are not reported for insurance reasons - the actual US crash rate per car is closer to 1 per 150k miles.

Elon said earlier this year he is assuming he can launch unsupervised Robotaxi's when he gets FSD to 3-4x the average human. But does this mean one crash every 150*3 - 450k miles, 277*3=831k miles or 498*3 = 1,494k miles?

I would guess they are aiming for around 1 per million miles which would in fact be 6.7x safer than the average human in the US. Of these crashes, only a very small % are likely to be the Tesla's fault.

Most likely Tesla's real goal is based on probability of injury per mile. In 2016 (i haven't checked updated data) in the US there was an accident with an injury every 794k miles and a fatal accident every 60.8 million miles.
There may be driving policy prioritisation decisions which make a minor crash more likely and a serious crash less likely for example.

Perhaps Tesla is aiming for a 10x improvement in accidents with injury as the threshold for Robotaxi launch (this will have added help from the safer design of Teslas & EVs relative to the US fleet). This would amount to one accident with injury per 8 million miles travelled per car.