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

[Silent Ludicrosy]

I feel like this is a near perfect encapsulation of Mr. Spiegel:

Mark B. Spiegel on Twitter

FINALLY got $NFLX (we're short) via a free 1-yr coupon from Fios, so now I can be in touch with the "cultural zeitgeist," lol. 1st conclusions:
"Roma": beautiful but boring; lost interest in 15 mins
"Birdbox": stupid; lost interest in 15 mins
"Highwaymen": excellent, but too long

There’s a scene in bird box where they drive a car with the windows blacked out using GPS to stay on the road. Autopilot or FSD would have made their trip a lot easier. Someone should edit that scene into a Tesla commercial.

 

[MarcusMaximus]

Not really - it actually does a lot of things - like recognize stop signs/traffic lights, turn etc. Why would all those features be in production ?

Current firmware recognizes traffic lights for the red light running warning. As for the rest, it’s probably there to assist with data collection.

Green is really good at unlocking things and gathering this data, but he also seems to make a LOT of assumptions based on less than sufficient evidence.

 

[MarcusMaximus]

There’s a scene in bird box where they drive a car with the windows blacked out using GPS to stay on the road. Autopilot or FSD would have made their trip a lot easier. Someone should edit that scene into a Tesla commercial.

Idk, the little car widgets on the instrument panel probably would have turned them.

 

[dqd88]

This guy ever long anything? I honestly don't understand what's so great about shorting. Infinite losses and you have to pay interest. Best case you double your money. Out of all the 10 businesses I've invested in, zero of them went bankrupt. 4 have multibagged. Math is not on the short seller's side at all.

It has gotten harder for short sellers over time. With each "crisis" usually comes some new regulation (Dodd-Frank, Sarbanes, etc) which reduce chances of similar excesses, and resulting failures, occurring in the future.

 

[Carl Raymond]

There’s a scene in bird box where they drive a car with the windows blacked out using GPS to stay on the road. Autopilot or FSD would have made their trip a lot easier. Someone should edit that scene into a Tesla commercial.

FSD would prob refuse to drive over bodies on the road. Apocalypse mode req’d.

 

[⚡️ELECTROMAN⚡️]

Clown Chanos says Teslas poorly made.

Bob Lutz car industry expert says Model 3 highest quality.

Bob is a bit of a clown too.

 

[Singer3000]

Hmmm... the imported SR+ is gone from Model 3 order page in China
ie. the ¥ 377,000 version

Not sure if it will be back, probably not?


View attachment 420454

The estimated delivery date in China for a SR+ is now also listed as 6-10 months, versus 1-2 weeks for the LR models. Looks pretty bullish for Rest of World demand for the SR+ if you ask me.

Model S is still the one that's going to cause the results to be stickier than we'd like. If you play around on the order page you'll see there's still quite a lot of pre-Raven inventory Model S still to go before 30 June - 70 LR and 100+ Model S 75D for Norway alone. Conversely (if you believe the website) you're going to have a hard time picking up a Model X this time of 30 June anywhere in the world right now, outside of maybe a custom order if you're in the US or Canada.

 

[EVNow]

Model S is still the one that's going to cause the results to be stickier than we'd like. If you play around on the order page you'll see there's still quite a lot of pre-Raven inventory Model S still to go before 30 June - 70 LR and 100+ Model S 75D for Norway alone.

But there are only 300 world wide

 

[EVNow]

Anecdotes from people given rides at Autonomy day reported 1 or 0 interventions on that route.

Well trained route ?

All you have to do is to listen to Musk describe what is happening in his commute. Trust me, if Musk had only one intervention per trip he would have said so.

 

[SPadival]

This.

I bet you a signed dollar his limited partners are actual lobbyists or types of players who are less concerned about financial performance and more concerned about pushing an agenda with Spiegel as a fund manager.

Other notable mentions are Gordon Johnson of "Vertical Group", Gabe Hoffman of Accipter Capital and Josh Wolfe of Luxe Capital. These people are merely paid actors, hired by "behind-the-scene" big hedge funds to be the FUD talking heads on CNBC and twitter.

There is an interesting back story involving Gordon Johnson, plagiarism and the newspaper of (changing) record, New York Times

Wall Street Research: A New Low - Executive Suite Blog - NYTimes.com

In a lengthy report issued in March by a team of Lehman analysts on the subject of virtualization technology (don’t ask), language and complicated graphics were lifted directly from several reports by Toni Sacconaghi of Sanford Bernstein...
............
.......But after conducting an investigation, it sent out one of the more groveling client letters I’ve ever read, in which it admitted the plagiarism, withdrew the research report from its Web site, and apologized both to Bernstein and its clients. And according to people involved, Lehman also fired Gordon Johnson, the analyst the firm concluded had stolen Mr. Sacconaghi’s work.

And then several months later, when Gordon Johnson started recommending solar stock shorts en mass, the above post became rather inconvenient. NYT, in an unprecedented move, then tried to wipe the post off the internet .

I'll bet you a signed dollar that the NYT reporter received a call from a big hedge fund and/or someone from the top at NYT:

Clearing a Name at Lehman

I have now removed all references to Mr. Johnson from that post because, after a series of phone calls with him, I have become convinced that he was not the plagiarist. When he first became aware of the post, seven months after the fact, he called me to complain. He made a convincing case that he had never plagiarized other analysts, and that he had not, in fact, been fired. Rather, he said, he left Lehman because he was ready for the next step in his career. His new firm had given him the solar industry to cover, and he had initiated his coverage with “sells” across the board.I added his “strenuous denials” to the bottom of my original post.

But in this age of viral communication, the original post, even with Mr. Johnson’s denials, has been used by investors to discredit his stock picks — especially on message boards that follow his stocks. And recently he was told that he was no longer welcome to be on a panel at an upcoming solar conference. In the e-mail message disinviting him, my post was specifically mentioned as a reason....

You can follow the rest of Gordon's shape shifting saga in Jesse's post :

The second highlighted area is an incredible similarity to the Fairfax story. Gordon Johnson, a solar analyst of Axiom Capital initiated coverage of Solarcity that day with a scathing Sell rating. This seem to set off the "waterfall" of Solarcity's decline. Over the next year, Johnson would publish report after report reiterating his bearish stance. This in itself isn't necessarily incriminating, perhaps he was even right. Since Tesla acquired Solarcity, Chanos moved his short onto TSLA. Curiously, Johnson is now no longer with Axiom and has moved to Vertical group. He is also now no longer a solar analyst and instead "specializes in Basic materials sector", which apparently also includes Tesla. He initiated coverage on April 9, 2018 with a Sell rating and currently is the owner of the lowest price target on the street at $99. He is constantly seen on CNBC, every time reiterating the brilliant Bob Lutz-esque thesis of how much Tesla loses for every car sold.

Oh .. and before Lehman, Gordy was working at .. wait for it.. Bear Stearns... so 2 out of 2 of the poster child failures of the financial crisis.

 

[Carl Raymond]

Well trained route ?

Subtle difference, but I assume well selected route.

Tesla would know which situations remain on the “to do” list and chose a route containing none of them.

A route comprised of the same challenges anywhere in the world and the car would perform equally well. Musk steadfastly and wisely refuses to use exception programming, which is one reason Tesla will win the FSD race.

 

[BioSehnsucht]

Yes, pretty common scenario.

In the released "stable" branches, they have put features that we saw in autonomy day. Those also happen to be the basic features needed for City NOA, which would be needed for FC. There has to be a good reason why we saw on autonomy day is same as the one in production - and why a much better version, if it exists, wasn't shown during the demo. The simplest explanation is dev on those features are over, so they have been merged into master branch and then released to production for use in shadow mode (or for whatever reason).

Yes, as I wrote the actual "dev" branches will be a few weeks more advanced. Conveniently, we know the production branch is 20th week's (sometime in May).

Bottomline, there is no evidence that there is a secret super-duper FSD software we haven't seen. Afterall during annual day, Musk said his City NOA needs interventions to complete his daily commute to office. If it rarely needed intervention (or just once or twice), he would have said so. He is talking about teaching the car not to run on grass / recognize the curb - that doesn't sound like a very advanced version.

I should also say, it is this kind of optimistic thinking that didn't prepare us for the Q1 shocker. All the evidence points to fairly basic stage of city NOA. Personally, I look forward to city NOA being similar to (or somewhat worse than) Cruise/Waymo when released after FC (hopefully sometime early next year).

It's not just that the public releases are a weeks old. Entire feature sets may be omitted from them, or not updated to the current version of that feature, and those features only available / built into internal development releases. Another example is the Early Access Program releases which often get one specific thing to test but end up being actually quite a bit behind of the public releases in other aspects.

There is not a single timeline of software evolution, there are many in parallel, merging, diverging, getting ahead and behind, and sometimes never to be merged again or only merged a long time after a feature A is complete and validated because without some other feature B must be completed and validated before it makes sense to release it into the public.

 

[Fact Checking]

The estimated delivery date in China for a SR+ is now also listed as 6-10 months, versus 1-2 weeks for the LR models. Looks pretty bullish for Rest of World demand for the SR+ if you ask me.

This could also mean that all SR+ inventory that arrived in Q2 is sold, and SR+ won't be shipped to China in Q3 at all - the 6-10 months correspond to the ramp-up schedule of the Shanghai Gigafactory. This would create some pent-up demand for Shanghai units, and it would also push more buyers towards the higher margin U.S. configurations.

This could also be a pure Q2 demand lever to push more non-SR+ units in Q2, and they might re-add SR+ to the configurator in Q3.

 

[Tslynk67]

Like most folks here, I'm delighted with the apparent reversal on the SP and what seems to be a 180° change in sentiment overall.

But the $420k question is, what's changed? To me, nothing; it has just flipped, well, because.

Answers on a postcard, please, to the usual address.

 

[Mars ☰mperor]

The estimated delivery date in China for a SR+ is now also listed as 6-10 months, versus 1-2 weeks for the LR models. Looks pretty bullish for Rest of World demand for the SR+ if you ask me.

Model S is still the one that's going to cause the results to be stickier than we'd like. If you play around on the order page you'll see there's still quite a lot of pre-Raven inventory Model S still to go before 30 June - 70 LR and 100+ Model S 75D for Norway alone. Conversely (if you believe the website) you're going to have a hard time picking up a Model X this time of 30 June anywhere in the world right now, outside of maybe a custom order if you're in the US or Canada.

The 6-10 months has been there since the introduction of the Giga 3 Model 3 SR+.

The delivery of the ¥ 377,000 version Fremont SR+ had been "end of June" and was then changed to "August" after the announcement of Giga 3 SR+. And I noticed it was gone today (haven't checked the order page for several days, so not sure exactly when).

This could also mean that all SR+ inventory that arrived in Q2 is sold, and SR+ won't be shipped to China in Q3 at all - the 6-10 months correspond to the ramp-up schedule of the Shanghai Gigafactory. This would create some pent-up demand for Shanghai units, and it would also push more buyers towards the higher margin U.S. configurations.

This could also be a Q2 demand lever to push more non-SR+ units in Q2, and they might re-add SR+ to the configurator in Q3.

I don't think any Fremont made SR+ has reached China. Maybe some will be here in Q2, as the aforementioned "August delivery time".

I figure from now on, SR+ for China will only come out of Giga 3.

 

[Tslynk67]

its $420000/1500contracts = $280 per contract which makes 2.80 contract

And who says Jim doesn't have a sense of humour?

 

[Singer3000]

But there are only 300 world wide

Good news if that's the case. I got fed up running each page through Google Translate so only checked a few of the markets (the US and Canada are also quite annoying to check too as you need to enter zips for each major market).

 

[Mars ☰mperor]

Amazing. A New Yorker whose entire wealth is really poorly managing his parents' money.

Next thing you know he's our next President

Gosh, another idiot to run the US.
The Chinese Communist Party will be over the moon!

 

[Tslynk67]

I've found this site to be more reliable for Max Pain: Opricot Open Interest|Volume|Max Pain

Ah, 80k $50 puts and a load of other being flushed down Spiegel's toilet end of this week, how satisfying is that?

But there's also 6k $3 & $4 calls to be cashed-in, I wonder when they were bought and how much they cost...

 

[Fact Checking]

It's not just that the public releases are a weeks old. Entire feature sets may be omitted from them, or not updated to the current version of that feature, and those features only available / built into internal development releases.

Note that there's several other good reasons for Tesla not to upload the true development version of their FSD code fleet-wide:

• Development versions change a lot, by their nature. This means that for every public update there would be a lot of churn in 'deactivated' development code and data, running up Tesla's cellular bill when updates are pushed via LTE.
• While it's a widespread industry approach to embed development versions in production versions to simplify "single image" deployment, it also seems well established that Tesla's "Early Access" program uses a different firmware version. Why would Tesla embed deactivated development code in their fleet-wide release to support a few hundred Tesla employees in their FSD testing group, if they don't do it for the several thousand people strong Early Access program?
• Tesla has to be aware of industrial espionage, hacking and FSD revenue theft risks: if a commercial FSD robotaxi license is going to cost say $50,000 in three years, I'm sure there will be a lot of "demand" worldwide to not pay that sum to Tesla but to pay a fraction of it to a friendly neighborhood "chipper" ... While Tesla probably can't stop the capabilities leaking once it's deployed fleet-wide, limiting access to the development code physically does help: every month the competition is delayed from reverse engineering their FSD progress might be a crucial advantage in getting to market first. For example the large, camera-agnostic "AKNET" neural network was removed from subsequent firmware versions I believe (@verygreen might falsify this though).

I'm pretty certain that the full neural networks that utilize HW3 are only uploaded to the cars of Tesla employees participating in the FSD test group, which is just a few hundred employees out of tens of thousands of Tesla employees. So I disagree with @EVNow here for all these reasons.

HW3 might be activated for Early Access customers that have HW3 installed already later this summer, and I don't think the HW2 code @verygreen managed to activate recently is necessarily indicative of the HW3 FSD code they demonstrated on the Autonomy Investor Day.

I believe what is happening is that Tesla is trying to see how much they can push HW2 capabilities for EAP-alike and base-Autopilot functionality. Having to upgrade the hundreds of thousands of EAP/base-Autopilot customers to HW3 is a hundreds of millions of dollar expense, so I'm sure they'd rather avoid that and only upgrade them if they buy the FSD option.

I.e. running Autopilot on HW2 cars is an ongoing development effort that the release of HW3-only FSD features is not going to obsolete. If they can squeeze red-light and stop sign detection into HW2 processing capacity then they'll happily add it. If it doesn't work out then those features might stay HW3-only.

At this point they probably don't yet know how far they can push the HW2 platform, it's 80% utilized without all these extra features, so it's near the limits of its processing capacity.