TSLA Market Action: 2018 Investor Roundtable

[ThereAre4Cars]

That's wrong, shadow mode very likely runs and logs the vehicle control logic as well, except that it doesn't actively control the car. This offers a measurable differential between human reactions and (shadow mode) vehicle control reactions.

This is what @diatz was talking about.

Tesla has a massive NN training feedback advantage by having hundreds of thousands of vehicles capable of running their test NNs in shadow mode.

I suspect that's Tesla investor mythology.

 

[Thekiwi]

If so then there is a fair chance it is never made.

Accusations of bait'n switch.

How about 5% of production dedicated to $35k M3 when it achieves 15% GM?

Wonder if we will see SR + PUP ($40k?) before that

 

[sauce]

What we should care about is cash generated per week, not units per week.

Exactly. The 75M$/day figure is eye-opening. Any steep ramp in production like they've experienced means an incredible increase in inventory value they must support until every sales are in the books. The shorter the delivery time (implying improved efficiencies along the delivery chain), the more cash per week at comparable production levels.

 

[electracity]

I forget when we were expecting Model Y exactly but it has been internally approved for production prototype (forget his actual words) and we are still 5 months out from March Model Y unveiling. Elon also said we won't be seeing VOLUME production until 2020 which I found curious instead of just saying production. I wonder with his new found 'give reasonable estimates' and speak carefully if we are going to see the Y a bit sooner than we think. With GF1 producing it and if he's expecting a ramp somewhat like the 3 maybe we start to see the production trickle start in 19H2. That would be cool.

Where is Tesla making the model Y in the second half of next year?

 

[Ulmo]

Wrong! He stated that service, not repair after a crash, is neutral.

One way, I think the main way, to interpret "repair shop" is the non-crash variety of service.

 

[EngRyan]

Where is Tesla making the model Y in the second half of next year?

I'm saying if volume production by 2020 has to be GF1. Any other guesses?

 

[ThereAre4Cars]

I am thinking it's more that Tesla expects FSD to be competitive with whatever general purpose autonomous solutions Uber and Lyft want to field (via assumed clear superiority once FSD is ready), and on a longer timeline to be able to offer FSD autonomy vs human driver Uber/Lyft as well, though that will have some additional issues (such as dealing with drunk and disorderly passengers, cleanup, etc) beyond just the driving aspect. It seems probably they also go the route of letting Tesla owners participate in a ride hailing service like Uber/Lyft human vs human, and just take a smaller cut and/or some other benefit (perhaps free SC for every kWh spent in the Network).

I will admit I've always found the obsession (not just regarding Tesla) with autonomous taxis, and a future with a huge drop off in personal vehicle ownership, to be a bit cart-before-the-horse. "Blank-as-a-Service" is already out of control in the IT world, I'm really skeptical of it regarding physical things.

Not that I'm any kind of analyst, but if I was, I'd keep valuing the value of the Tesla Network at zero until it actually exists, because it's not just about self driving, but all the other issues that go with unattended people using you car and that some percentage of humans are terrible.

I'd value FSD as some kind of potential future thing, that I'm reasonably sure comes eventually, but not as an enabler for Tesla Network but to allow people to go on road trips while reading / watching movies / sleeping (hopefully), or having your car park itself and return to pick you up at the office/mall/etc (probably), etc - that has a value to the owners of the vehicles.

On the other hand, the oft ignored (outside of TMC) TE side of the business is clearly a huge thing just waiting to take off when given the chance.

But a lot of these uber-bull theses seem predicated on this future where "Millenials" never buy cars and just take ride hailing service everywhere, and thus see Tesla Network as some kind of golden goose. While I can see ownership decreasing among younger people become a trend, as you grow older and start a family you're going to want your own car. So I don't think the Tesla Network is going to be that big of a factor in the long run, compared to actual products (vehicles, TE, etc). It'll be some nice gravy on the side, but it's not the steak or the potatoes.

I think the automotive business is the driving valuation force up to about 2023-2025 at which point the maturity of FSD will start really enabling significant and carefully circumscribed TaaS type functionality that will then be clearly a trillion dollar growth space for 20 years. So my hope is that when automotive has basically maxed out its contribution to valuation the FSD side of the business takes up the reigns. And when I say automotive I include EAP because higher-than-usual margins from autopilot software is why Tesla can beat Toyota market cap and provide a great return even before the ARK narratives start to come into focus.

I personally don't think much of TE but it bores me to talk about. It apparently bores Elon and company as well since they never really discuss it. Just leaving that as an FYI.

 

[beachbum77]

Well, he said he expects profits from repairing out-of-warranty cars. I suppose you wouldn't expect them to repair out-of-warranty-cars at a loss. I hope he doesn't mean high, gouging-level profits, but I'm fine with a 1% profit margin on out-of-warranty repairs...

One percent profit on parts/service? Not a chance. Assume you have not seen a parts bill for body shop repairs?

This is where Tesla can start making serious money but hopefully balanced with what owners will find "reasonable" or they could kill resale value on out of warranty cars. That could bode very badly for Model 3 owners in 4 years/120k miles.

 

[EngRyan]

I suspect that's Tesla investor mythology.

You're from the bay area but don't follow in detail the latest V9 software neural network talk? While the new NNs running are impressive the real news is the one running in the background that was discovered by one of the guys on TMC (forget his name) To paraphrase it looks to be the largest commercially deployed NN by a decent margin and we know while it's on V9 cars it isn't running current AP because you need the power of the new chip to run the thing. If it's background deployed on V9 then clearly they are working it for the new chip.

 

[ThereAre4Cars]

I got the impression this cc plan was created by a PR team. Sounded like Elon had some notes prepared and even stuck to them for a couple minutes. They even dropped the Youtube section.

 

[Remus]

It's often (if not always) a sell on the news story. and in this market, it's take whatever you can get and get clear. We MAY get under 300$ tomorrow if the Asia/Europe selling catches up to us and we start lower.

Based on a 10 dollars annual earning that is a P/E of 30! If we get that price I can persuade my wife to liquidate all her holdings and all in.

 

[Mkorpal]

I'm saying if volume production by 2020 has to be GF1. Any other guesses?

Did they start construction on the next section that it looked like they were gearing up for a while back? I havent seen any new footage for a few months.

 

[Curt Renz]

CNBC - after market close:

 

[RobStark]

Wonder if we will see SR + PUP ($40k?) before that

I have advocated walking down the price of all Tesla vehicles in increments since Elon was a baby.

Start with fully loaded S,X and 3.

Then lower the content and corresponding price by $2k-$5k increments.

 

[elasalle]

How many Giga FUD Factories does the FUD machine need now?

 

[ThereAre4Cars]

You're from the bay area but don't follow in detail the latest V9 software neural network talk? While the new NNs running are impressive the real news is the one running in the background that was discovered by one of the guys on TMC (forget his name) To paraphrase it looks to be the largest commercially deployed NN by a decent margin and we know while it's on V9 cars it isn't running current AP because you need the power of the new chip to run the thing. If it's background deployed on V9 then clearly they are working it for the new chip.

Yeah I heard about it. It's really 'big'. Except that neural nets can be infinitely big but have asymptotic performance typically at fairly small sizes. Going 'big' is usually the realm of academics trying to beat other academics by .0001% on a digit recognition task.

Reinforcement learning uses a good/bad reward stimulus to propagate penalties back to decisions made earlier. If the NN is not controlling the car there is nothing to penalize. Further, the dimensionality of the problem is absurd. An NN might not know that a purple car is dangerous if it never saw one before because it doesn't generalize like we do and our intuitions are broken when we think about it. Heck even the visualizations that you see generally prove this isn't end to end training because those visualizations are of conceptualizations that are handed off from the perception layer and then used by the procedurally coded planning layer.

Just give me some actual evidence for end-to-end training.

 

[Ulmo]

You have to hire perfect drivers and then drive these perfect drivers through every possible situation.

I don't think so: most human drivers are near perfect and go through an infinite number of situations every day already. The tail risks are really bad (so the tail risks are much more amazing than the near perfect driving until one realizes all of this). The neural network can easily compare human performance and figure out when the humans are better vs when the humans are not as better, as well as some of the less good human driving. While absolute perfection is almost impossible, near perfection is easy. Tesla's neural network challenge will merely be to get very near perfection, and they're already ahead of the game when they do that. Elon keeps wanting to hint that the neural network will get very very near perfection, but has not actually said that in years.

I see two mistakes commonly made with people regarding brain-like computing:

1. People not thinking like brain-like.
2. Brain-like computing not being brain-like enough (or equivilent/similar/analogous).

 

[sauce]

I suspect that's Tesla investor mythology.

I'd like to point out that Tesla investor mythology (whatever that is) was spot on with Q3. Pretty much everybody except Value Analyst had underestimated Q3 results. No reason not to be bullish with TSLA going forward.

 

[X Fan]

One percent profit on parts/service? Not a chance. Assume you have not seen a parts bill for body shop repairs?

This is where Tesla can start making serious money but hopefully balanced with what owners will find "reasonable" or they could kill resale value on out of warranty cars. That could bode very badly for Model 3 owners in 4 years/120k miles.

Donn—

Under Elon's watch, I highly doubt they will ever fleece their owners through unreasonable service pricing.

But...you wouldn’t be you without throwing mud on the wall.

 

[BioSehnsucht]

Wonder if we will see SR + PUP ($40k?) before that

It's the obvious move. Pushes out developing a non-PUP interior until later as well. With the equipment to build SR packs supposedly going in this Q4, I wouldn't be surprised if they start offering SR w/ PUP in Q1 at some point.

Once non-PUP becomes an option, that opens a spread of new pricing options along nearly the entire Model 3 range (I wonder if they'll ever offer a non-PUP Performance ... there might be a market for it)