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

[Singer3000]

Regarding Tesla Semi
April 2019 Q1 call, Jerome Guillen:

“Next year we will start production. We are very happy, we are driving the trucks extensively with so far, I think, quite amazing success, yes.”

If this timeline is to hold, shouldn’t we be hearing quite soon about the where and the how?

We know some of the long lead Model Y machinery is already on order and it will be made at Fremont (Shanghai to follow). What of the Semi? Hope we get a proper update on the Q3 call.

Has felt like a very long wait between the Nov 17 unveil and any further detail. Perhaps the Q3 Battery Day I wonder?

Model Y will be the cash cow that when combined with the 3 finally tips Tesla into sustained profitability. But it’s the Semi that’s likely to make the biggest impact on emissions.

 

[Artful Dodger]

The worst is SC. Service Centre, Supercharger, Securities Commission?

Worse still in South Carolina.

 

[Intl Professor]

I do agree with this. If you have a fringe gaggle of activists opposing a given project and they feel justified in performing some act of civil disobedience, then they will go to jail, and rightfully so.

On the other hand, if you have non-violent activists acting with substantial (though possibly latent) public support and sympathy, then their actions may well help to galvanize public opinion and thereby influence the democratic process. This was certainly the case in the US civil rights movement, which has been judged favorably by history.

I do think there are ample legal avenues to address our current climate crisis, so please don’t misunderstand me; I am not advocating that people go to pipeline construction sites and block bulldozers, etc. I would rather that people buy Tesla stock, drive EVs, support sustainable energy, invite others to do the same, and lobby for better government policies. Those actions will, I expect, send a much better message than Greenpeace-like disruptive actions.

I bolded the part at end that this applies to. But there's got to be some outlet for eco friendly jocks and other crazies!

 

[MP3Mike]

Has felt like a very long wait between the Nov 17 unveil and any further detail. Perhaps the Q3 Battery Day I wonder?

I thought battery day was next year. But maybe they will give more Semi details during the truck release later this year.

 

[FreqFlyer]

Regarding Tesla Semi
April 2019 Q1 call, Jerome Guillen:

“Next year we will start production. We are very happy, we are driving the trucks extensively with so far, I think, quite amazing success, yes.”

If this timeline is to hold, shouldn’t we be hearing quite soon about the where and the how?

We know some of the long lead Model Y machinery is already on order and it will be made at Fremont (Shanghai to follow). What of the Semi? Hope we get a proper update on the Q3 call.

Has felt like a very long wait between the Nov 17 unveil and any further detail. Perhaps the Q3 Battery Day I wonder?

Model Y will be the cash cow that when combined with the 3 finally tips Tesla into sustained profitability. But it’s the Semi that’s likely to make the biggest impact on emissions.

I will stick my neck out there and state the new Tesla PU has the makings of being a cash cow too with higher margins if accepted by the market.

 

[Singer3000]

I thought battery day was next year. But maybe they will give more Semi details during the truck release later this year.

Yes you are right, it’s been formally pushed back from q3 (nearly over!) to c. q1 2020. Which makes me doubt any meaningful Semi production until 2021 to be honest.

 

[Fact Checking]

Kinda ...still all sorts of unproven stuff ....like the Tesla was stripped inside....while the Porsche was not.

Actually, it was proven that the Taycan was stripped:

TLDR version: "No, we published a time from our slow car, without telling anyone at the time that it was our slow car, because that's a totally normal thing to do! What, did we race our fast car, and what time did we get? Um.... no further questions."

And of course, the car they raced was neither Turbo nor Turbo S; it was a stripped pre-production prototype:

Stock my arse.

Oh, and remember that dispute earlier over whether Turbo S could be gotten with summer tires on the Mission E wheels at all (let alone the specific ones they raced on)? There wasn't any disagreement at all that the Turbo can't.

In the first picture you can see that the Taycan back seats were removed and a (lightweight) roll cage was installed. The second picture shows a stripped interior.

 

[MarcusMaximus]

Worse still in South Carolina.

Or Southern California.

 

[pyromatter]

 

[avoigt]

Construction worker leaving and production workers arrive!

 

[Krugerrand]

Construction worker leaving and production workers arrive!

Phew! I can sleep at night now knowing the damaged fence has been removed.

Edit: Scratch that. Some of the damaged fencing.

 

[jerry33]

I'm shaking my head reading that. They're talking about right-brain brute force optimization without any creativity.

I once made a program to fit geocoded maps of every single road in Iceland (including lots of road metadata, altitudes, etc) - gigs of raw data - into just a couple megs that could be rapidly loaded by a website and processed by Javascript. How? Rather than storing all of the points that define each road, I took advantage of the fact that the points on roads are not just random, but relative to the points that came before, and furthermore, curves on roads tend to just continue along. So I wrote a program to calculate optimal splines that could represent dozens or even hundreds of points that define a road, only switching splines whenever the deviation got too great to be represented with a single spline. Representing data as splines not only made it far smaller, but it made it easier for the client side to interpolate the roads to arbitrary resolution and with much greater speed than having to draw every single point.

In Cruise's case, it's not exactly the same thing (1d vs. 2d data), but the same sort of logic applies. Your dataset is too large to cache? Don't just try to come up with a way to crunch an awkwardly large dataset; look for a better way to represent your data.

More succinctly, "The fastest way to do something is not to do it".

 

[Fact Checking]

I'm wondering what this big installation on one end of the factory is:

Based on the large diameter ducting I'd guess air treatment - but aren't those HVAC units on the roof?

The new foundation works in the "Phase 2" area are also interesting:

Am I interpreting this correctly, that the deep foundations on the left side suggest foundations for heavy machinery and multi-story industrial buildings?

I.e. some of the more benign explanations for the new foundations like worker living space or sewage treatment plant can probably be excluded, right?

Whatever "Phase 2" is, it's on a big area about half the size of the rest of the factory:

​

 

[Drax7]

Knowing Elon, he will sleep on the floor if required to get it going.

 

[CorneliusXX]

I'm wondering what this big installation on one end of the factory is:

View attachment 457711

View attachment 457713​

Based on the large diameter ducting I'd guess air treatment - but aren't those HVAC units on the roof?

The new foundation works in the "Phase 2" area are also interesting:

View attachment 457709​

Am I interpreting this correctly, that the deep foundations on the left side suggest foundations for heavy machinery and multi-story industrial buildings?

I.e. some of the more benign explanations for the new foundations like worker living space or sewage treatment plant can probably be excluded, right?

Whatever "Phase 2" is, it's on a big area about half the size of the rest of the factory:

View attachment 457710
​

The latest rumours I've heard is that phase 2 is for module production. Which makes sense at a high level. The vehicle manufacturing plant should go first to ensure demand for cells and alternative supply is available in the interim. The battery day is early next year which should roughly be the time phase 2 is finished.

 

[AlMc]

Best guess: 'Paint' ventilation/air exchange

 

[Troy]

Honest question - what good does estimating, crowdsourcing quarterly production estimates do for the greater good of TSLA stock price? Whatever gets polled , usually a high number - the Wall Street consensus just rises the bar towards report day and makes it impossible for tsla to look good. Guys like Troy don’t really add much value to increasing share value - just nonsense and noise to spin people’s wheels on. This is just my opinion of course and I’d love to understand and would appreciate hearing why I’m wrong here.

I think you are confusing two different projects of mine. The project you are criticizing is this one: Tesla Production & Delivery Estimates Survey. This is a survey where people enter their estimates and I measure the accuracy. I participate in this survey myself. The aim is to find out people who are consistently good at estimating. However, my main project related to Tesla estimates is what I do on Twitter. I've been releasing my quarterly estimates there for many quarters. Check out my Q3 2019 thread here.

If we look at the last few quarters, in Q4 2018, my estimate was pretty bad. Actual deliveries were 90,966 units. My estimate was 79,509 units. I was off by 11,457 units. My error margin was 12.6%. These things happen sometimes. Bloomberg was off by 26% once. So my worst was ~13% off. My other estimates were better. I have a project called Model 3 Order Tracker here, and a similar project for Model S/X here. I use those data points when calculating my estimates.

For example, in Q1 2019, most people were estimating more than 60K Model 3 deliveries. I estimated 52,000 units based on my survey. The actual number was 50,900 units. I tweeted that here as usual on the last day of the quarter. Here is my accuracy in the last few quarters.

Total Tesla deliveries:
Q2 2018: I underestimated by 4,011 units
Q3 2018: I underestimated by 3,169 units
Q4 2018: I underestimated by 11,457 units
Q1 2019: I overestimated by 4,800 units
Q2 2019: I underestimated by 4,028 units

My current estimate for Q4 is 97K. I'll update this again on 30 Sep. You can follow this here on Twitter.

 

[Webeevdrivers]

I think you are confusing two different projects of mine. The project you are criticizing is this one: Tesla Production & Delivery Estimates Survey. This is a survey where people enter their estimates and I measure the accuracy. I participate in this survey myself. The aim is to find out people who are consistently good at estimating. However, my main project related to Tesla estimates is what I do on Twitter. Check out my Q3 2019 thread here.

Looking at the last few quarters, in Q4 2018, my estimate was pretty bad. Actual deliveries were 90,966 units. My estimate was 79,509 units. I was off by 11,457 units. My error margin was 12.6%. These things happen sometimes. Bloomberg was off by 26% once. So my worst was ~13% off. My other estimates were better. I have a project called Model 3 Order Tracker here, and a similar project for Model S/X here. I use those data points when calculating my estimates. Here are the last few quarters.

Total Tesla deliveries:
Q2 2018: I underestimated by 4,011 units
Q3 2018: I underestimated by 3,169 units
Q4 2018: I underestimated by 11,457 units
Q1 2019: I overestimated by 4,800 units
Q2 2019: I underestimated by 4,028 units

My current estimate for Q4 is 97K. I'll update this again on 30 Sep. You can follow this here on Twitter.

Your Canadian numbers interest me. It seems Canada is much lower than originally thought. What’s happening in Canada. It seems last quarter Canada deliveries were more robust.

 

[KarenRei]

I think you are confusing two different projects of mine. The project you are criticizing is this one: Tesla Production & Delivery Estimates Survey. This is a survey where people enter their estimates and I measure the accuracy. I participate in this survey myself. The aim is to find out people who are consistently good at estimating. However, my main project related to Tesla estimates is what I do on Twitter. Check out my Q3 2019 thread here.

Looking at the last few quarters, in Q4 2018, my estimate was pretty bad. Actual deliveries were 90,966 units. My estimate was 79,509 units. I was off by 11,457 units. My error margin was 12.6%. These things happen sometimes. Bloomberg was off by 26% once. So my worst was ~13% off. My other estimates were better. I have a project called Model 3 Order Tracker here, and a similar project for Model S/X here. I use those data points when calculating my estimates.

For example, in Q1 2019, most people were estimating more than 60K Model 3 deliveries. I estimated 52,000 units based on my survey. The actual number was 50,900 units. I tweeted that here as usual on the last day of the quarter. Here is my accuracy in the last few quarters.

Sorry, but there's no bonus point for fixing a bad prediction when actual data shows that it was wrong.

Total Tesla deliveries:
Q2 2018: I underestimated by 4,011 units
Q3 2018: I underestimated by 3,169 units
Q4 2018: I underestimated by 11,457 units
Q1 2019: I overestimated by 4,800 units
Q2 2019: I underestimated by 4,028 units

My current estimate for Q4 is 97K. I'll update this again on 30 Sep. You can follow this here on Twitter.

Oh please. In both Q2 and Q3 you've forecast a delivery collapse based on a bogus "demand problem" hypothesis, and predicted way low numbers, then corrected your numbers as it neared quarter end and the actual facts came out, and now repeatedly pat yourself on the back for "getting it right", as though you hadn't each time falsely predicted a delivery collapse.

Sorry, but there's no bonus points for fixing a bad prediction once data comes out showing that it was wrong.

 

[Ulmo]

It goes like this: A market maker wants the price to be much lower. They submit short sales of say 250,000 shares using a waterfall approach: They will execute against all of the order books, gobbling up all the highest 250,000 shares of standing buy limit orders, but doing it over a period of a minute or two. The market makers generally also have insight into most of these buy limit orders, so they can estimate just how big a drop they will cause. Say it causes a $6 dollar drop in about two minutes. Now this huge drop has the potential to spook retail market participants, many of whom will be scrambling to figure out what news do big insiders have that they don't? Some of them will decide to sell too. Over the next several hours (or several days sometimes - market makers are allowed to have FTDs for many days too!) they buy back these naked shorts with many small ~500 share orders at the "new" lower prices they caused with their artificial mini-crash.

The buying back does have an upward effect on price, but it is smaller because it is drawn out over a long period of time. If the market maker sees actual strong buying interest before they have bought back their shares, then they can cap it with limit sell (naked short again) orders such that large buying volume does not raise the price much. If they do that, it will increase the size of their short so it's likely they will also extend the time over which they cover. The point is that this is done to control market psychology - that this stock is being dumped by big players, so now is not the time to buy and maybe it is the time to sell. Coordinating it with some bogus news, downgrade or such just adds to the psychological force.

I saw that many times back when I was watching the stock price all day (most of 2016, and very little in 2017). I wish I had screen recordings, because it happened exactly as you described, but whenever I looked into the tape (recorded trades), it never showed up on the tape. I think whoever was doing it, and I do suspect and think most likely it would have been those with insight and access like the market makers, had and have ways of doing it that don't show up on tape.

The SEC rules allow them to do this for the supposed reason of providing orderly markets. I'm sure the MM would say that they are "adding liquidity" to the market, but there is no need for them to say anything because the SEC does not care what they do.

$TSLA is a stock that doesn't need additional liquidity. What a bogus excuse! Sure, if we turn off all Market Makers, the system might start looking really different, but I am guessing it would be much cleaner and maybe more honest.