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

[deffu]

I'm not a TSLA super bull and I strongly disagree with this bear guy. Tesla is not demand constrained at all. In fact, Tesla is gaining and will gain more organic demand over time. EVs become more and more enticing by the month and year.

To say that Tesla is demand constrained is to say that the tens of billions being invested by other oems right now will go to waste. Tesla makes the best EVs. If Tesla ever has a problem with demand, the industry will implode.

I would not call it production constraint or demand constraint. Clearly tesla lost the pricing power they had last year when they could command high premium on fullfilling backlog orders. If it is production constraint, the price woudnt go down. At the same time, the production is still increasing. If it is demand constraint, that needs not to happen. As dumb as Musk is, he is not in a business to lose money. tesla has entered a phase where an equilibrium needs to be reached by carefully allocating production resources and deliberately pulling demand triggers. The overall trend is still going up. production increase is steadfast, the addition of GF3 is imminent. On the other hand, there is a huge potential demand to tap, but a good portion of it are less hardcore or more price sensitive. The priority is to increase production as fast as possible, working carefully to match production to current demand till we reach a new equilibrium where production not only translating to sales but also profits. To say Model 3 is production limited is blind to lot of facts. When we needed to wait for months for iphone 4 to ship, when we could flip a wii bought at $249 at bestbuy for $500 on ebay, when Nikon D800 was on backorder for half a year on amazon after its release, this is called production constraint. To say Model 3 is demand constraint is also ignorant. The demand has not plateaued and will not be for quite a while. However, the low hanging fruits style demands have mostly gone is indeed true to me. In summary, model 3 is still in growth mode trying to break through on both fronts: production and sales. To put a cap onto either of these fronts has no point for the advancement of the mission or for the interests of its shareholders for the long term. It's not constraint either way. It's growing at a pace in a much needed balanced way.

 

[Bet TSLA]

AI is about big data. You need tons of data to train it. Then you need tons more to validate it. Then you need tons more to verify it for approval.

Sometimes. AlphaZero whupped human ass with no data at all. Seemed a lot like AI to me. You can make the assertion that autonomous driving is completely different from Go, but you would be hard pressed to prove it. Sure, different game, different rules, but is it different enough that you can be sure nobody else can some at it from a different direction and win?

 

[SpaceCash]

Just got a notification, etrade is jumping on the "no commission" bandwagon starting Oct 7

 

[KSilver2000]

Regarding Model 3 production rate.

-----------------------
2018 Q4
Production: 61,394
Weekly average: 4,723
Seq Growth: 15%

Letter (20 Jan 2019):
In our Fremont facility, we are now past the steep portion of the production S-curve, and we expect our production rate to continue to gradually improve. Every part of the Model 3 production process has demonstrated over a 24-hour period the ability to produce at an extrapolated rate of 7,000 vehicles per week. By the end of this year, we expect to be able to produce Model 3 at this rate on a sustained basis.

2019 Q1
Production: 62,975
Weekly average: 4,844
Seq Growth: 3%

Letter (24 Apr 2019):
We produced roughly 63,000 Model 3 vehicles in Q1, which was approximately 3% more than the previous quarter. This improvement in production rate was modest mainly due to changes to the production process for the introduction of new variants of Model 3, fewer working days and a supplier limitation. We started production and deliveries of Model 3 vehicles for overseas markets during Q1.

2019 Q2
Production: 72,531
Weekly average: 5,579
Seq Growth: 15%

Letter (24 Jul 2019):
"The production rate of Model 3 continued to improve gradually throughout the quarter, breaking a monthly record in May and then again in June. All manufacturing equipment in Fremont has demonstrated capability of a 7,000 Model 3 vehicles per week run rate, which we continue to work to increase. We aim to produce 10,000 total vehicles of all models per week by the end of 2019."

2019 Q3
Production: 79,837
Weekly average: 6,141
Seq Growth: 10%
YoY Growth: 50%
---------------------

All seems on track to me. Could very well be that Fremont exited Q3 producing 7k Model 3 per week.

It's the sluggish demand for S&X that has been the main problem, hotly followed by gross margins for the 3 stubbornly stuck in the mid teens.

With luck margins should before long nudge towards the target, with softer depreciation impact at Fremont once it's sustainably +7k per week and with further improvements in the the production process. The overall margin mix will get another kick once Shanghai Giga is up to speed with its lower COGS and depreciation per unit (though will be horrible during the ramp of course). And once the Y is in full swing, S&X demand won't matter a jot. Patience young padawans.

The black swan remains a 2020 global downturn and/or US-EU trade war, to which I don't have a good answer for you, other than to diversify into govies and bullion and have trust that the best companies tend to come out the other side stronger.

Good info.

Would also be curious if anyone knew what the general shift schedule (if any different from Q to Q) was like for each quarter.

 

[Carl Raymond]

Sometimes. AlphaZero whupped human ass with no data at all. Seemed a lot like AI to me. You can make the assertion that autonomous driving is completely different from Go, but you would be hard pressed to prove it. Sure, different game, different rules, but is it different enough that you can be sure nobody else can some at it from a different direction and win?

You can’t compare chess and ‘go’ to driving. The data for leaning games can be generated by the computer itself, and translate perfectly to the real world task of beating human opponents.

If you computer generate driving situations, your data is lacking, because if the computer thought of it, it ain’t an edge case.

 

[MC3OZ]

The production constraint is the main issue, higher volumes of production will increase margins and allow slight price reductions which can spur demand. Tesla can continue to improve the product, specifically the software, which can also help spur demand..
IMO even a 1K-3K price drop on a SR Model 3 can unlock significant additional demand, that will be true as long the price is above 35K.
Increasing experience of friends/neighbours Model 3s, should also increase demand...
Model Y will have strong initial demand...
With the exception of Model S/X, demand is not a significant problem,

 

[ggr]

Me too, at least 10. No crashes but 2 times it tried to go straight over the curb it was front parked against. I was using “come to me”. Every person I showed did a video or said “get the f out, no way!” It is by far from perfect but like AP I can’t wait to see what the next version does. Also, this is the path any other FSD company is going to have to do. This is a very heavy lift and they are all years behind if not a decade.

Just two weeks ago my Model 3 drove over a curb in a parking lot. Of course I didn't have V10 yet, I was driving...

 

[LN1_Casey]

Just got a notification, etrade is jumping on the "no commission" bandwagon starting Oct 7

I'm with e-Trade too, what exactly does this mean? I know they have to make money somehow, so there must be something else done to keep this business afloat.

 

[Bet TSLA]

If you computer generate driving situations, your data is lacking, because if the computer thought of it, it ain’t an edge case.

I don't think that's as simple as you make it sound. I can sit here at my computer and generate more edge cases than are in any database of real world driving because I have imagination. Oh, you saw a car with a bicycle on the back? I can imagine three bicycles on the back and an elephant on top. And there's a kid riding the elephant. In pajamas. And people keep darting out into traffic to throw the elephant peanuts. If I can imagine things, then we can train an AI to imagine things. You can feed in real-world data as hints, but it will riff off that in myriad ways.

The real world is not particularly rich compared to the worlds I can imagine. And you can feed in all the real world data you like in the form of "Do you have this yet?" and after a short time the answer will always be "Yes, and a few million similar things." Remember, it took AlphaZero just a few days to become the best Go player in the world by far. With technology that was relatively primitive. We don't need to create anywhere near the best driver in the world, just a reasonably adequate one.

In short, I think there are quite possibly alternative approaches that don't need tons of data. Of course they don't need lidar either.

 

[SpaceCash]

I'm with e-Trade too, what exactly does this mean? I know they have to make money somehow, so there must be something else done to keep this business afloat.

They make money on the bid/ask. By frontrunning your orders like they have always done

 

[adiggs]

Sometimes. AlphaZero whupped human ass with no data at all. Seemed a lot like AI to me. You can make the assertion that autonomous driving is completely different from Go, but you would be hard pressed to prove it. Sure, different game, different rules, but is it different enough that you can be sure nobody else can some at it from a different direction and win?

AlphaGo worked by CREATING the data it needed. It created that data by playing random games using as guidance the rules of the game. By playing enough games, the model started learning plays that improved win probability and other moves that reduced win probability.

So if you think you can design a simulator that can create random detailed traffic situations and randomly learn success from "playing" those random traffic situations, then I suppose yeah, you've created something comparable to AlphaGo that can learn to drive.


Very big difference between the two. There's a pretty limited set of legal moves at any point in Go - nothing like the range of moves when driving.

 

[Duffer]

I'm with e-Trade too, what exactly does this mean? I know they have to make money somehow, so there must be something else done to keep this business afloat.

Someone earlier (I think it was @Hock1, apologies if I'm incorrect), who is a former trader, said that the companies make the identical trade seconds before they make yours. That way you pay slightly more and they pocket the difference.

 

[LN1_Casey]

They make money on the bid/ask. By frontrunning your orders like they have always done

Someone earlier (I think it was @Hock1, apologies if I'm incorrect), who is a former trader, said that the companies make the identical trade seconds before they make yours. That way you pay slightly more and they pocket the difference.

I see. I had been under the impression that they didn't do this, since they charged the commission.

 

[deffu]

The production constraint is the main issue, higher volumes of production will increase margins and allow slight price reductions which can spur demand. Tesla can continue to improve the product, specifically the software, which can also help spur demand..
IMO even a 1K-3K price drop on a SR Model 3 can unlock significant additional demand, that will be true as long the price is above 35K.
Increasing experience of friends/neighbours Model 3s, should also increase demand...
Model Y will have strong initial demand...
With the exception of Model S/X, demand is not a significant problem,

This is true. But I would not call it production constraint. It's more like margin constraint or scale constraint. While I dont deem Musk's words misleading, i dont take it at face value either. Production/demand constraints are all about who have more sayings in price negotiation. I dont see it all in favor of tesla in this game currently. While I dont see tesla has short term demand problem like what we had in Q1 and long term demand is very healthy, i don't buy into the production constraint theory. needs to scale the business to become profitable is an operational issue tesla is working on. We really dont need to follow the the rhetoric Musk had towards those ws analyists. We are not production constraint or demand constraint. We are growing both production and demand at a healthy pace.

 

[MP3Mike]

They only managed 6.141 average so something must have happened that they did not manage to achieve the 7.000+ average production rate

Are the people capable? Suppliers? Logistics? Etc.

 

[wipster]

You should edit so as not to come off as hostile towards Fred. I 100% agree with you but have had multiple comments deleted for being mean to him

Me too.

 

[RobStark]

Q3: Model S/X produced 16,318, delivered 17,400.

They sold MORE than they made.

@KarenRei

Tesla, Inc. (TSLA) Q1 2019 Earnings Call Transcript | The Motley Fool



Tesla, Inc. (TSLA) Q1 2019 Earnings Call Transcript


Pierre Ferragu -- New Street Research -- Analyst

Hey, thanks for the call. My first question is really on the Model S and Model X. And Elon, you said, you're comfortable with them. You see -- based on what you saw in April, do you think that the 25,000 units per quarter is the level of demand that is where you see the market coming back already or are we not there yet? And more specifically, in the US, the pull forward in Q4 probably hurt a lot of demand for S and X. Is it that something that we still see in the numbers today in recent weeks or is that behind us? And I'll have a follow-up on Q2.

Elon R. Musk -- Chief Executive Officer

Yeah. I mean, I think something like the -- returning to the 100,000 a year annualized demand for S and X is what we anticipate. That's to the best to my knowledge. We don't have a crystal ball, but that's probably our best guess. And sorry, what was the other point?

 

[Carl Raymond]

nothing like the range of moves when driving.

...multiplied by the range of ‘playing surfaces’.

 

[Thekiwi]

From the Q2 letter



They only managed 6.141 average so something must have happened that they did not manage to achieve the 7.000+ average production rate

That 6,141 Q3 average weekly production number assumes zero downtime. It’s highly unlikely there was zero downtime in Q3, so the actual weekly run rate is higher than that number, and since logic says they undoubtedly finished the quarter at a higher rate than they started at, then I think it’s likely the current weekly output number is fairly close to a 7,000 rate, at the very least over 6500.

 

[kengchang]

That 6,141 Q3 average weekly production number assumes zero downtime. It’s highly unlikely there was zero downtime in Q3, so the actual weekly run rate is higher than that number, and since logic says they undoubtedly finished the quarter at a higher rate than they started at, then I think it’s likely the current weekly output number is fairly close to a 7,000 rate, at the very least over 6500.

At least July 4th and Labor day are national holiday in Q3.