2017 Investor Roundtable:General Discussion

[EinSV]

Def bearish after that recap of upcoming news...

Why?

 

[TNEVol]

Some observations. I believe long distance drivers can run about 11 hours not just 8. Also I see no mention of International (known as "corn binders" to drivers) While i believe that medium to large trucking firms will jump into driver-less transportation, it will not happen quickly. Labor costs, down time, liability and maintenance cost? are major positives pushing adoption. In the meantime, big companies struggle to hire and keep good drivers who are very interested in the brand of the Tractor and lots of chrome. Peterbilt is their Harley. It is very important that Tesla's semi be gorgeous and powerful. For small companies and IOs the tractor is their symbol to the world and brand loyaly is supreme. Most drivers and small owners care nothing about efficiency and cost. They are likely to own a Ford Superduty to get back and forth to work. They aspire to own a Corvette some day. I realize this is sterotypical but it generally applies to the many drivers I know. They are proud of their skills and their truck. I think this is a great tactic by Tesla to approach semis first before the pickup because adoption by the OTR drivers would cause disruption in pickup owners who like Tesla and Corvette owners have incredible loyalty. However it would be a mistake to think disruption will be easy. Just consider how obstinate the Teamsters have been over history.
Opps wrong thread. Meant for General. Please excuse.

 

[geneclean55]

Why?

I believe @dw4ngg was indicating "bearish" but left out the scare quotes

 

[LargeHamCollider]

Also cost of Nvidia's computer will be substantial... I haven't seen estimates but my ignorant wild guess would be around $1k.

I doubt the nVidia board will cost Tesla $1k. nVidia sells the same GP102 chip to consumers at an MSRP of $699 in the GTX1080 Ti. Tesla's order for these boards was likely the biggest order, in dollar terms, in nVidia's history. I'd be surprised if Tesla is paying more than $800 per board if there are two GP102 chips on the board, if there's a single GP102 it should be well under $500.

Now we just need someone to do a teardown.

 

[D-egg-O]

I have a couple questions for somebody in the trucking industry...

Would Tesla be up against a legacy dealership model like in the auto side, or are trucking sales normally transacted directly with the OEM?

I'm also wondering if the existing logistics providers would embrace a small but perpetual "per mile" service revenue for Tesla - effectively sharing a small % of operating cost savings - based on autonomy as the service, like Tesla Network. This would avoid the capitalization problem of providing end to end logistical services, but still give Tesla a slice of the pie - without having to recreate the bakery.

One way to make this work could be to include the AP h/w and EAP/FSD as no charge standard features. Since Tesla Semi operating costs per mile are going to be significantly lower than ICE or other EV competitors (due to lowest battery costs), would the existing industry embrace a sharing of that per mile savings? This could give Tesla a perpetual profit stream far beyond what they would charge upfront for EAP/FSD. Would Tesla be in a position to drive this model as a new industry standard practice? (Does their unmatched leadership in data-miles collected leverage them into a first mover position in autonomous trucking?)

Same independent dealership model. Though I don't know if the same franchise laws apply. I do know Cummins (engine OEM) acquired all of their distributors.

I don't see why the Tesla Network Elon's explained for passenger cars could not be mirrored for the trucking business. You'll have fleets owning trucks in the network. As well as owner operators. And Tesla will provide trucks as well.

If the trucks have a cab, EAP/FSD will be an upgrade but the hardware should come standard similar as new S and X cars. If there is no cab, EAP/FSD has to be standard. Tesla is best positioned to drive this since they both own the software technology and are working on the manufacturing technology too.

 

[D-egg-O]

A trucker in the reddit thread telling me they can run non-stop with multiple drivers and the cost per mile goes down when they do so. The only way I can imagine that to be true is if the off-duty driver isn't paid, yet also is stuck in a truck instead of being at home free to enjoy one's non-paid hours.

It's called team drivers, and they literally drive non-stop. One drives as the other one sleeps. Not sure how much the one driver makes when he/she is sleeping.

 

[Fifield]

2. I can't imagine Tesla simply selling the semi truck to other companies, which would eliminate the added value of having full control. I think this was not a semi truck announcement, it was the announcement of Tesla Logistics, as others have mentioned. I think the service will include automated logistics solutions, which could mean purchasing a smaller company like project 44, project44 – connections that matter, if Tesla doesn't want to develop logistics solutions from the ground up, or it could mean they already have a logistics solution in line with Tesla Network development. Either way, this is just the beginning if they prove they can execute. It's a wide, wide net they're casting here.

 

[luvb2b]

@surfside, thanks for your hard work. after spending a few minutes trying to understand, i ran the numbers both ways. either way i get a sensible mix shift towards sales - which i did not have before your straightening me out.

the 2 line items that get moved for our discussion are almost identical except in q2:
period q3 16 / q2 16 / q1 16
Revenue from Solar Energy Systems Under Long-term Loan Arrangements 20,421 18,988 16,327
Revenue from Operating Lease Prepayments and Upfront Incentives 18,253 46,345 19,110

so my suggestion is let's just call it mystery solved, case closed and move on.

(i started writing all this stuff below before checking numbers both ways and realizing it probably doesn't matter at all)

one thing i learned as i was going through solarcity financials is that they are always moving the goalposts.

in this case i think it's due to reclassification: (scty 2016 10k, bottom pg 36)
Reclassifications ... Additionally, revenue from sales under MyPower contracts, which was previously presented within
revenue from solar energy systems and components sales, has been separately presented as revenue from solar energy systems under long-term loan arrangements; this together with periodic operating lease billings and incentives revenue and sales of solar renewable energy credits revenue comprise the total revenue from periodic billings.

i know the proforma numbers here only match if i use the categories as mentioned.
Operating leases and solar energy systems incentives
— 195,204
Solar energy systems and components sales
— 113,152

stopped working on it here =)

Been a busy day; just getting around to referencing my spreadsheet/notes.

Here is language from the 2016 10-K MD&A:

“Revenue from sales of solar energy systems and components increased by $201.9 million, or 190%, for the year ended December 31, 2016 as compared to the year ended December 31, 2015. This increase was primarily due to the $124.6 million increase in revenue from solar energy systems sold under Solar Loans and the $19.6 million increase in revenue from cash sales to residential customers.”

Given this comment, it is clear that solar loans are falling under the sales section.

Also, from the 2015 10-K (in the Quarterly Results of Operations section):

Revenue - Solar energy systems and components sales

1Q2015 = $12,708
2Q2015 = $24,520
3Q2015 = $28,798
Nine months ended 3Q2015 = $66,026

From the 3Q2016 10-Q:

For the nine months ended 3Q2015
Solar energy systems and components sales revenue = $47,641
Revenue from solar energy systems under long-term loan arrangements = $18,385
Total = $66,026


I haven't looked at the pro forms section you referenced, but this seems to make it pretty clear. Let me know if you still think otherwise.

I haven't had time to read all of the back and forth on the deferred revenues, but @neroden's comments seem to make sense -- they factor into the gain at the time of acquisition (decreasing the amount of goodwill created), and that is how those revenues "disappear". Let me know if there is more digging to do on this topic; I'm happy to look into it further if you don't feel like you've figured it out.

Also, while I'm posting, as for AP2 take rates, I would tend to agree with others that for the Model S and X, the take rate of EAP has got to be 75+%, whereas FSD must be significantly lower (maybe 30%?). For the Model 3, I would not be surprised if the prices are the not the same as for Model S and X -- we will see. One reason I believe this to be the case is because Elon commented that the increase in price of the Model 3 to go to dual motor would be less than what it is for Model S and X. Maybe my logic is faulty here, but I wouldn't put it past Tesla to reduce the price of the EAP and FSD software for the Model 3 to increase the take rate -- as you point out, there is no incremental cost to Tesla to provide access to the software; as a result Tesla needs to balance getting the price just right to maximize revenues through the combination of volume of takers and price paid. Just my $0.02 of course.

surfside

 

[dw4ngg]

I believe @dw4ngg was indicating "bearish" but left out the scare quotes

I was being facetious

 

[vgrinshpun]

I've been discussing this issue longer with vgrinschpun who is using terms as sole competitor etc. so I take that to mean a near monopoly.

This is just not accurate.

What I was saying is that Tesla is a sole *competitive* player as it has the lowest cost. It is clearly not the same as saying that Tesla is a sole competitor. There obviously are others, trying to get projects by emphasizing other (than cost) attributes, with varying degree of success.

 

[steve3]

Same independent dealership model. Though I don't know if the same franchise laws apply. I do know Cummins (engine OEM) acquired all of their distributors.

I don't see why the Tesla Network Elon's explained for passenger cars could not be mirrored for the trucking business. You'll have fleets owning trucks in the network. As well as owner operators. And Tesla will provide trucks as well.

If the trucks have a cab, EAP/FSD will be an upgrade but the hardware should come standard similar as new S and X cars. If there is no cab, EAP/FSD has to be standard. Tesla is best positioned to drive this since they both own the software technology and are working on the manufacturing technology too.

Thanks for your helpful reply! I'll post this here and then cross post in the dedicated Tesla Semi thread where this discussion belongs.

It seems likely to me that Tesla would mirror an existing/planned business model like Tesla Network, as that keeps things both simple and highly profitable - customers as silent business partners who happily generate pure profit for you. Does it make as much money for Tesla as starting up a new Tesla Logistics business unit? Probably not as much, but it would be nice for Tesla to profit from some breezy tailwinds rather than take on another business to create from scratch.

Buy selling semis rather than than producing only Tesla-owned trucks, they don't compete against fleet operators and independent operators. A great many of these operators would jump at the chance to access significant cost reductions in their existing business models: they pocket enhanced profits, they're customers are thrilled with significantly lower shipping costs, early birds grab market share from competitors slow to adopt, and the overall addressable trucking market expands further, probably into arenas traditionally served by rail freight haulers.

100% profit, royalty-type, burgeoning annuity - always a welcome line item on any high-growth enterprise's balance sheet!

 

[DaveT]

I doubt the nVidia board will cost Tesla $1k. nVidia sells the same GP102 chip to consumers at an MSRP of $699 in the GTX1080 Ti. Tesla's order for these boards was likely the biggest order, in dollar terms, in nVidia's history. I'd be surprised if Tesla is paying more than $800 per board if there are two GP102 chips on the board, if there's a single GP102 it should be well under $500.

Now we just need someone to do a teardown.

But Nvidia is selling Tesla the entire computer and not just the board.

 

[MP3Mike]

But Nvidia is selling Tesla the entire computer and not just the board.

Do we actually know that? We know that what Tesla has is not the NVidia reference design, it is custom. (And it was originally intended to have a Mobileye EyeQ3 chip in it as well as the NVidia parts.)

 

[zdriver]

Do we actually know that? We know that what Tesla has is not the NVidia reference design, it is custom. (And it was originally intended to have a Mobileye EyeQ3 chip in it as well as the NVidia parts.)

Agreed, I can't remember source off hand, but I seem to recall Elon saying they had to rework the board after Mobileye dumped them, implying it's a custom board that was originally designed to have both systems working alongside each other until they were ready to transition to AP2 only.

 

[luvb2b]

thanks. that makes sense, it did seem high at $5k. i think they get good value from those computers gathering all that data anyway. with so many cars collecting data they will have better resources to perfect self-driving than anyone else.

No sure Tesla will be able to charge $5k for enhanced AP to Model 3 users... maybe slightly lower. Also cost of Nvidia's computer will be substantial... I haven't seen estimates but my ignorant wild guess would be around $1k. Add sensors cost, maybe another few hundred. Then as an additional cost we need to add in the cost of users who don't activate AP. But overall I agree it's high margin, especially if more than 50% of buyers choose the option. And then fully self-driving option is even higher margin.

 

[DaveT]

Do we actually know that? We know that what Tesla has is not the NVidia reference design, it is custom. (And it was originally intended to have a Mobileye EyeQ3 chip in it as well as the NVidia parts.)

Agreed, I can't remember source off hand, but I seem to recall Elon saying they had to rework the board after Mobileye dumped them, implying it's a custom board that was originally designed to have both systems working alongside each other until they were ready to transition to AP2 only.

"Tesla Motors and NVIDIA have partnered since the early development of the revolutionary Model S. Today, all Tesla vehicles—Model S, Model X, and the upcoming Model 3—will be equipped with an NVIDIA-powered on-board “supercomputer” that can provide full self-driving capability." - The Tesla Motors and NVIDIA DRIVE Partnership

Nvidia announced Tesla is using their PX Drive liquid-cooled supercomputer, not just the "board" or GPU.

Regarding Mobileye, Elon never said they had to "rework the board" or computer after the relationship went sour. He said that he would have wanted and was planning for the systems to work together to aid in the transition. But from what I've heard, the relationship ended far before Tesla ordered the AP2 computer hardware from Nvidia.

 

[LargeHamCollider]

But Nvidia is selling Tesla the entire computer and not just the board.

The board is an entire computer, the enormous GP102 die will dominate cost, it will be fed by a handful of ARM processors that are on the same board, also a bunch of RAM and i/o will be on the same board. Main memory may be off board.
The ARM processors (and possibly main memory) is/are the only significant thing(s) the GTX 1080 Ti boards don't have.

GP102 really is huge, it'll probably have 5x the number of transistors that all the rest of the ICs on the board combined have (excluding memory).

Also calling it a supercomputer is a bit misleading, yes it sort of has supercomputer level compute performance but "actual" (I use quotes because one can always quibble on definitions) supercomputers usually run lots and lots of identical or similar chips in parallel.

 

[MP3Mike]

Nvidia announced Tesla is using their PX Drive liquid-cooled supercomputer, not just the "board" or GPU.

Tesla's version is not liquid cooled. (It has fans on top.) And Elon did say that they reworked the board to take the EyeQ3 chip off. (He said it during the ER call on 2/22/2017.)

 

[DaveT]

The board is an entire computer, the enormous GP102 die will dominate cost, it will be fed by a handful of ARM processors that are on the same board, also a bunch of RAM and i/o will be on the same board. Main memory may be off board.
The ARM processors (and possibly main memory) is/are the only significant thing(s) the GTX 1080 Ti boards don't have.

GP102 really is huge, it'll probably have 5x the number of transistors that all the rest of the ICs on the board combined have (excluding memory).

Also calling it a supercomputer is a bit misleading, yes it sort of has supercomputer level compute performance but "actual" (I use quotes because one can always quibble on definitions) supercomputers usually run lots and lots of identical or similar chips in parallel.

Tesla's version is not liquid cooled. (It has fans on top.) And Elon did say that they reworked the board to take the EyeQ3 chip off. (He said it during the ER call on 2/22/2017.)

Thanks @LargeHamCollider and @MP3Mike

 

[schonelucht]

What I was saying is that Tesla is a sole *competitive* player as it has the lowest cost. It is clearly not the same as saying that Tesla is a sole competitor. There obviously are others, trying to get projects by emphasizing other (than cost) attributes, with varying degree of success.

And I have been saying that Tesla has the likely the lowest battery cost, but that it may very well have higher other costs. And that cost is not the only metric that determines competitiveness. Therefore that your conclusion that Tesla is the sole competitive player is wrong. Maybe we should agree to disagree because you do not seem willing to acknowledge these two assertions.