General Discussion: 2018 Investor Roundtable

[scaesare]

If you're trying to train an AI system to identify objects in a video feed, yes. You train the AI with video feeds and training data.

Highways are the easiest target. While you're moving fast, they're also the most predictable. There are no stop signs on highways, no traffic lights, no pedestrian crossings, etc. Typically, lanes are much wider and the road condition much better than normal roads.

For Tesla to start moving towards full self driving or L4/L5, they're going to need to start training their cars on public roads. They're going to need to train it to see traffic signals and signs. They're going to need that video data.

Which is why it was concerning that Tesla recorded zero autonomous miles driven on California roads in 2017.

The answers as to what is actually going on are already on this forum.

Finding them are an exercise left to the reader.

As a bonus, the answer that should alleviate your oh-so-genuine concern in California is also lurking around here.

Happy hunting!

 

[adiggs]

The problem is this: myself, and to the best of my knowledge nobody else, is getting any younger. My ICE vehicle sucks and I want an EV. Myself, and increasingly more people, don't want a sedan--I want a sport utility vehicle. We can play the song and dance of when exactly they should and shouldn't start manufacturing the Y, but I'm not going to wait forever, and I suspect many others won't as well. The more years that go by before the model Y is released, the more market share & earnings they will give up to other automakers.

Specifically, I really don't want to wait much longer than 2020 (my car is a 2011 GMC Terrain). Will there be decent alternatives to the model Y available then? That's yet to be seen, but I wouldn't write that possibility off completely.

It may be of little help to you, but I can tell you that Model X is an amazing vehicle

What might actually be helpful - I was saying for years before the first Model 3 showed up in the wild, that the first Model 3's would actually be used Model S'. The early Model Y's will be some of the early Model X's that are being traded in or returned at lease end. And we're far enough in the Model X life cycle that I would expect to start seeing a modest number of used Model X's available now.

I particularly like Model X for towing a utility trailer - lets my Model X replace (for me) a pickup truck as well.

 

[RobStark]

So when did this transaction occur? Because the guy managing this fund is the one quoted in a bunch of FUD articles about being disappointed with Musk, with the implication funds would pull money out. lol

And didn't Fidelity reiterate support for management?

As they sold TSLA?

 

[adiggs]

AGreed. Not many things an SUV can do that a model s can’t do better. AWD, voluminous hatch plus frunk, third row for soccer team transport.

Towing my truck bed utility trailer

 

[Anstandswauwau]

You can hook up any Model S to a WiFi network and monitor the router usage. The payloads it's uploading are quite small (largest I saw was 20 MB), certainly not large enough to be either compressed or uncompressed video data, and especially not for 8 separate cameras.

Talking about flawed arguments

You do know that they ONLY send compressed stills (something in the realm of 2-8 x 1080p, essentially greyscale plus some metadata) of situations AP couldn't resolve properly, right? Since, you know, that's what they feed into their neural nets.

20MB should be good for some hours of shadowdriving.

Next.

 

[mongo]

Talking about flawed arguments

You do know that they ONLY send compressed stills (something in the realm of 2-8 x 1080p, essentially greyscale plus some metadata) of situations AP couldn't resolve properly, right? Since, you know, that's what they feed into their neural nets.

20MB should be good for some hours of shadowdriving.

Next.

Heck, I've done bulk sensor validation using only counters for events.

 

[UrsS]

@FirebirdAlpha

(quote) Which is why it was concerning that Tesla recorded zero autonomous miles driven on California roads in 2017. “DMV”) (endquote)
"Concerning" hmm...
The real reason however is that there are no Teslas driven without a driver in the car. Therefore there is nothing to report. After all its called "Driverless Testing of Autonomous Vehicles".

 

[Curt Renz]

Gene Munster - 23 minutes ago: An Open Letter to Elon Musk | Loup Ventures

Mod: leaving this here because it is informative and relevant, and not just about the tweets, but any discussion of the tweeting bit should probably be in that other thread. --ggr

 

[Lessmog]

Gene Munster - 23 minutes ago: An Open Letter to Elon Musk | Loup Ventures

As a rank outsider it seems to me that Loup Ventures is here trying to maintain an influence as investors while Elon has appeared to forswear of such trying to stand on own legs.
It may still be good advice, but it rings a bit tainted by the tone of threat I imagine to detect.

BTW, have you changed the avatar into deployment gear? Wonder why ...

 

[Curt Renz]

...BTW, have you changed the avatar into deployment gear? Wonder why ...

As motivation to diet and exercise.

 

[Esme Es Mejor]

As a rank outsider it seems to me that Loup Ventures is here trying to maintain an influence as investors while Elon has appeared to forswear of such trying to stand on own legs.
It may still be good advice, but it rings a bit tainted by the tone of threat I imagine to detect.

BTW, have you changed the avatar into deployment gear? Wonder why ...

All Musk needs is one intern to act as a Twitter gatekeeper with simple instructions:

Is it a fan/customer with a question to which you can provide helpful info? Go for it!

Is it a fan/customer saying something nice for which you can be grateful? Go for it!

Is it a positive article you can retweet? Go for it!

Is it a request for humanitarian help? Go for it!

Anything else? No! No! Hells to the no!

Haters gonna hate. Trolls gonna troll. Randos gonna rando. The only way to win is not to play.

 

[electracity]

This report has a really, really critical flaw:

The latter I refer to as “shadow mode” because the data streams in these vehicles (whether under manual or Autopilot control) is available to be used for training the neural networks that perform the various components of the perception-control task.
​

The problem is that the cars aren't transmitting data.

You can hook up any Model S/X/3 to a WiFi network and monitor the router usage. The payloads it's uploading are quite small (largest I saw was 20 MB), certainly not large enough to be either compressed or uncompressed video data, and especially not for 8 separate cameras. I can't test for 4G connections, but I doubt that Tesla would be using this to stream video, as it would get quite expensive.

Since the most challenging part of autonomous driving is rapid object identification from video feeds, it doesn't seem that Tesla cars are currently "uploading data" in any meaningful way.

Production cars are probably used to test software, not accumulate primary training images. The car could accumulate images of exceptions. For examples, most traffic control signs are know via the map. Can the visual perception system traverse the road and read the signs under all conditions?

Another example is emergency braking. The car brakes, but the driver does not (low hanging branch or whatever). The car sends a photo of the exception.

 

[Reciprocity]

Really? Samsung has a comparable 2170 cell? What specifically do you know about this cell besides it's packaged as a 2170 cylinder to conclude it is comparable? Do you know if it is NCA? Does it have the same energy density as Tesla's 2170? Does it use as little costly cobalt as Tesla's? Do you know Samsung says it won't produce these in BEV quantities until 2021? Even if Samsung's were just as high performing, they won't be available for real use for three more years. A three year lead at minimum? Almost sounds like a part of Tesla's moat!

From Tesla Q1 2018 Shareholder Letter:
"Cells used in Model 3 are the highest energy density cells used in any electric vehicle. We have achieved this by significantly reducing cobalt content per battery pack while increasing nickel content and still maintaining superior thermal stability. The cobalt content of our Nickel-Cobalt-Aluminum cathode chemistry is already lower than next-generation cathodes that will be made by other cell producers with a Nickel-Manganese-Cobalt ratio of 8:1:1. As a result, even with its battery, the gross weight of Model 3 is on par with its gasoline-powered counterparts. "

Battery cell energy density improvements aren't driving costs down? The 2170 cell chemistry improvements allow them to store about 30% greater energy per Kg. No one outside Tesla/Panasonic knows just how much cost reduction over previous generation accrues from lower manufacturing costs at GF and how much from M3 needing fewer cells for the same pack KWh.

Your original assertion was battery chemistry and performance isn't a significant component of Tesla's total moat. Neither Elon or J.B. would agree.

Elon says moats are lame. Never heard JB speak on the subject.

Here is a thought exercise for you. Where would Tesla be with no gigafactory? Sourcing cells from Samsung and LG like everyone else. They already consumed all of Panasonic's cells. Panasonic doesn't even use the 2170 chemistry in the 18650 cells.

I'm done arguing this point. Cell chemistry is not a moat. At least not one that is not easily crossed. Building the largest factory on the planet and fully automating it is a moat. It's one Tesla has a 5 year headstart on.

Edit:

Elektroautos: Tesla Model 3 kann mit Gewinn gebaut werden
This german M3 analysis has bombshell news:
Battery manufacturers and their suppliers are therefore working hard to develop new cathode materials that use less cobalt. Tesla and his partner Panasonic appear to be ahead of the competition, as previously known: According to the laboratory analyzes leaked to WirtschaftsWoche, the cathodes of the Panasonic cells used in the new Tesla Model 3 consist of only 2.8 percent cobalt. The current state of the art is currently eight percent cobalt. "That would be a significant competitive advantage for Tesla, Kobalt is currently very difficult to get on the world market," says Sven Bauer, head of Germany's largest independent battery producer BMZ.

 

[Sancho]

Munster suggests Musk issue an apology, go on a Twitter sabbatical, and ignore the short sellers.

This advice is too simplistic. What’s more important than refraining to tweet, is for Musk to learn how to tweet more effectively. He needs to pick his battles more wisely, be more strategic and less reactive, and recognize that his position carries responsibilities of demeanor within one or two standard deviations of CEO-like demeanor.

Most importantly, Musk needs to view his enormous Twitter following as a company PR asset that needs to be better coordinated with the other company PR assets and strategies. He should be collaborating with a hand-picked PR person that has a style Musk is comfortable with, to maximize the effectiveness of his tweets.

That’s not as fun and liberating as being able to just sound off on Twitter like an average Joe can, but who said that being a CEO of a major public company is fun and liberating?

 

[mongo]

Elon says moats are lame. Never heard JB speak on the subject.

Here is a thought exercise for you. Where would Tesla be with no gigafactory? Sourcing cells from Samsung and LG like everyone else. They already consumed all of Panasonic's cells. Panasonic doesn't even use the 2170 chemistry in the 18650 cells.

I'm done arguing this point. Cell chemistry is not a moat. At least not one that is not easily crossed. Building the largest factory on the planet and fully automating it is a moat. It's one Tesla has a 5 year headstart on.

To extend this thought:

A moat keeps people out. GF1 doesn't keep anyone out of making electric cars. Tesla's cell chemistry does not keep anyone from making e cars. The supercharger network doesn't either.

No one is restricted from developing their own factories, cars, and chargers due to the existence of Tesla.

The true moat is the one OEMs have dug around themselves. The moat of short term profitability.

Look at Tesla, intial vehicle was low volume, high price, zero charging infrastructure. But it worked because the company was small and lean and didn't try to serve the mass market.

Then the S, original projection was maybe 20k cars a year. First ones had no supercharging. Then they started rolling out supercharging, starting where the cars were and extending out. S was also more popular than expected. Over time, Tesla, along with capacity and coverage (if not net profits) grew.

Flip to OEMs, they can do all the same things, if they are willing to be worse than Tesla at the beginning. They can't just turn on a supercharging network, it has to start somewhere and grow. They can't just find battery supplier availability, it must be developed. All that is required is to have a long term plan that is expected to be profitable at the end. But they want to keep the dividends flowing, and that is their moat.

 

[ValueAnalyst]

How about this: Model 3 ASP in the second half of 2018 may be more than $60,000.

Since Standard Range is now a 2019 story, and P (10% take rate) and D (50%) are around the corner, as well as already popular EAP (80% take rate) and possible FSD-exclusive features in the coming months with the new-and-improved Karpathy-magic neural networks (50% take rate in 2H18 but possibly as high as 80% with a killer feature like even highway-only Level 3 at $3,000), including wheels and premium paint and delivery:

OPTIONS
Enhanced Autopilot: $5,000; 80% take rate
Full Self-Driving: $3,000; 80% take rate as exclusive features roll out
Premium Upgrades Package: $5,000; 50% take rate
Premium Wheels: $1,500 to $4,500; 50% take rate with $2,000 average
Paint: $1,000 to $1,500; 80% take rate with $1,200 average

MODELS
Long Range: $44,000 - 45%
Long Range with Dual: $49,000 - 45%
Performance with Dual: $75,000 - 10%

ASP without options: $49,350
ASP of options: $10,860
ASP: $60,210 excluding $1,000 delivery

Model 3 ASP may be more than $60,000 in 2H18.

Now consider that the ultra-bearish UBS analyst had estimated Model 3 breakeven (not COGS but EBIT-level so including R&D and SG&A expenses) at $41,000 (page 45 of 95) even using $160/kWh battery cost estimate, which is just dumb. His other dumb assumptions include D&A per unit and Warranty Provision per unit, which should be combined $3,000 per unit lower, but to be conservative, let's use his bear-level dumb assumptions.

$60,000 ASP in 2H18 - $41,000 COGS, R&D, SG&A = $19,000 EBIT per Model 3

31.66% EBIT margin in 2H18.

That's better than Apple.

I posted this analysis here on March 24, and concluded 31.66% margin for Model 3 in 2H18.

Yesterday: Tesla Model 3 teardown shows over 30% profitability as Munro research firm ‘eats crow’
Today: Tesla Model 3 is most profitable electric car: consultant

It turns out that Munro also was the source for the UBS analysis that I references and specifically pointed out the items where it was wrong.

 

[UnknownSoldier]

To extend this thought:

A moat keeps people out. GF1 doesn't keep anyone out of making electric cars. Tesla's cell chemistry does not keep anyone from making e cars. The supercharger network doesn't either.

No one is restricted from developing their own factories, cars, and chargers due to the existence of Tesla.

The true moat is the one OEMs have dug around themselves. The moat of short term profitability.

Look at Tesla, intial vehicle was low volume, high price, zero charging infrastructure. But it worked because the company was small and lean and didn't try to serve the mass market.

Then the S, original projection was maybe 20k cars a year. First ones had no supercharging. Then they started rolling out supercharging, starting where the cars were and extending out. S was also more popular than expected. Over time, Tesla, along with capacity and coverage (if not net profits) grew.

Flip to OEMs, they can do all the same things, if they are willing to be worse than Tesla at the beginning. They can't just turn on a supercharging network, it has to start somewhere and grow. They can't just find battery supplier availability, it must be developed. All that is required is to have a long term plan that is expected to be profitable at the end. But they want to keep the dividends flowing, and that is their moat.

Just so you know, you just described the Innovator's Dilemma

The Innovator's Dilemma - Wikipedia

 

[TradingInvest]

Gene Munster - 23 minutes ago: An Open Letter to Elon Musk | Loup Ventures

I highly respect Gene Munster. But in this case, I think Elon knows what he is doing.

Tesla shorts have been making an unprecedented effort to break Tesla. They have too much on the line, they can either make 10 billion or lose 20~30 billion. I would not be surprised if they spend 10% to actively destroy the target. At this stage they don't have a choice, there is no easy way out. I am talking about a few large shorts who likely own the majority of the 12 billion short position.

If Shorts get their way and cause the stock to crash badly, lots of Tesla supporters would lose all their money and may not have the means to buy Tesla products down the road; shorts would win big and continue have the resource to hurt Tesla in the long run. Remember many shorts loaded Puts try to gain many fold this summer. So when the known bear analyst asked the conversion rate, the bear was looking for ways to break the Tesla stock. If you were Elon, how would you answer? Tell them only 15% converted, the rest are waiting for other configurations? If Elon said that, the next day every news media would have a bold title "Only 15% Tesla reservers are real buyers, Tesla is in big trouble". Most of those media are controlled by money. They will intentionally omit the second part that many people are waiting for the performance, AWD, base, etc. As a result, part suppliers would start to demand cash payment. Then the stock goes into free fall, all the leveraged longs get wiped out. Shorts get big profit and get ready to attack again.

Shorts attack Tesla in many ways. One of them is to smear the Tesla products. In the past, this has not been too big a problem, those diehard Tesla supporters would buy Tesla cars anyway. But now Tesla is in a new phase, it's selling a lot more cars to the general public. If Elon let the shorts controlled media to freely smear Tesla's reputation, Tesla might actually start to have demand problem. Just imagine this, if everyday every where you go, you see negative news about GM cars, would that affect your buying decision?

Shorts also try to paint a picture that Tesla treats employees badly, if they succeed on this smear campaign, Tesla would have trouble to recruit high quality applicants. Right now, the best of the best join Tesla, which is big deal for the long run. Tesla's reputation is very important for long term success.

It's not a viable solution to just ignore the short attacks. If there is a better way to address it, I would love to hear it, and I'm sure Elon would be interested to know too.

 

[mongo]

It's not a viable solution to just ignore the short attacks. If there is a better way to address it, I would love to hear it, and I'm sure Elon would be interested to know too

That might be the plan around Pravduh.

 

[brian45011]

And didn't Fidelity reiterate support for management?

As they sold TSLA?

Gavin Baker got me-too ed. His replacement lightened up on the allocation to TSLA.