2017 Investor Roundtable:General Discussion

[techmaven]

I thought Elon said that 100k was the production capability combined for S & X? That from this point forward, they were just trying to increase gross margin.

There is a time expiration for that comment. BIW production line #1 is still in operation building the S and BIW production line #2 (high speed line) is building the X. The gating factor right now is general assembly and that is mostly a labor issue. This can be resolved by re-designing the product and applying some of the Model 3 lessons in GA to the S/X production. So the numbers are light for 2017, but likely really light for 2019 through 2021. The cost to build the Model X should drop in the next few years... Tesla can take that with either increasing gross margin or dropping the price or increasing content.

 

[Reciprocity]

Just a suggestion: I think your points would be clearer if you used EAP and FSD to refer to the software, not AP2.

Both FSD and EAP use AP2/AP2.5 hardware, which is the source of the confusion.

I have always used HW1 and HW2 to differentiate them, because it stands for Hardware v1 and Hardware v2. I dont really talk about Hardware v2.5 because I dont really know much about it. To me AP1 is software and so is AP2 and the version is really what every software version your car is updated with. From my point of view, there is no AP2.5 software, its the same software as AP2, with extra functions that may or may not be using a camera inside the car.

 

[Reciprocity]

The 'vision' part of Mobileye (2+years ago) and the current hardware in the Tesla AP 2.0/2.5 is perfectly capable of reading stop signs/other traffic signs/stop lights. Just not implemented currently.

Pull up the Mobileye 'Roadshow' for their IPO. Clearly they could do it, just cautious to actually implement it.

Part of the reason for the Mobileye/Tesla split was because EM wanted to push the envelop faster than Mobileye was comfortable doing.

Yes HW2 is totally capable of reading the signs with the right software, but AP2 software will not go that far. At some point AP2 software will be replaced by EAP which will be based on FSD. It makes no sense to continue to build out AP2 software when Tesla's original intent was to never create it at all and just leverage what they built for AP1 and Mobileye hardware. They where forced into creating AP2 software to work with HW2, which was purpose built for FSD and never originally meant to be a bandaid for the lack of Mobileye/AP1. HW2 is overkill if that was the AP1 level functionality was the original intent.

 

[MitchJi]

On theother hand, Tesla and Panasonic's figures are low if one was using the same metrics. Panasonic's 2016 automotive cell shipments was more than 50% higher than BYD's LiFePO4's at over 6 GWh and more than double LG's. Tesla's commissioned amount should be 16-18 GWh, Panasonic's is something like 10-12 GWh. And under construction/announced/available by 2021 is vastly larger. At least 105 GWh, if not hundreds of GWh.

Counting announc?

By the end of the year Tesla has announced that they will announce plans for several more Gigafactories. That's not even counting Elon's statements about seven or eight Gigafactories.

 

[Bobfitz1]

MOD Delete:

IF you have something to say, say it in the right thread. And, in fact, please do.

No, I cannot be 100% consistent 100% of the time. But when I do interject into a thread, ALL had better be on the qui vive for more than five minutes.

 

[Ovulation]

So yeah, more on the Jeffries report.
* low S+X, falling into the trap of over-extending Q1/Q2 2017 figures as a trend
* low 2018 sales, but really low 2019 numbers.
* very low growth past 2019
* still thinks 35 GWh is the cell production capacity of the Gigafactory 1. Really? Really? Are these people that incompetent?
* estimate of $175/kWh pack costs is high for Model 3 at volume in 2017

How came they up with this kind of estimates. I mean UBS took a Bolt apart, that was great work. I assume the other big boys and girls work the same, they research. I talk off analysts of course, not the shorties who claim the GF is a potemkin village.

BTW. I don't think it's unreasonable to assume to S could get into trouble, the interior of the 3 seems to be better thought out (cupholder etc.), the range is similar, in short the S is not differentiate enough, the car needs an upgrade. There are the truther and the birther and yes I'm a HUDster.

 

[Waiting4M3]

So yeah, more on the Jeffries report.

I'm only about 2/3rds of the way through looking at it, but the sales volumes projection by model chart is really, really off.

They have 91k S+X for 2017. With 47k delivered thus far, they are thinking 44k for the rest of the year, or 22k per quarter. Yeah, that's likely to be debunked pretty soon. The low-ish S+X estimates goes all the way through 2021, with 94k in 2018, and 107k in 2019... this seemingly smallish miss is a big deal because of the gross margin. They are actually pretty bullish on Model 3 numbers for 2017... 45,000. A higher M3 versus S+X mix impacts overall gross margins in their model. For FY18, they have 300,000 Model 3's. That's not a bad conservative estimate of 6,200 to 6,500 a week. Could be low by 10%. The big issue is FY2019. They have only 380,000. Even if the entire year was 10% less than the guided 10k/week for exiting 2018, the number would be 432,000. At guidance with no additional expansion for the year, that would be 100,000 low. 100k! They do have some Model Y's and maybe they think that it would impact Fremont production, but I would expect Model Y to be on a new production line somewhere. In addition, I expect changes to the Model X as well as production to increase that significantly. Beyond 2019, they have basically tepid growth. Have they met Mr. Musk?

The report actually has quite a bit that is bullish. But they fall down on a very key parts, and one fixes those, the report would be extremely bullish. The upside case in their report is $525. Thus far, here are the key misses:

* low S+X, falling into the trap of over-extending Q1/Q2 2017 figures as a trend
* low 2018 sales, but really low 2019 numbers.
* very low growth past 2019
* still thinks 35 GWh is the cell production capacity of the Gigafactory 1. Really? Really? Are these people that incompetent?
* estimate of $175/kWh pack costs is high for Model 3 at volume in 2017

I see the same old bear logic behind the guidance for low MS/MX, and low growth after 2018

• Demand will taper off once the initial "enthusiasts" buyers are done, and tax incentive goes away
  
• M3 will osbourn MS/X because limited # of enthusiasts
  
• MY will osbourn M3/S/X, ditto as above
  
• Big guys will easily build better EVs than Tesla once they get serious

 

[Yggdrasill]

If Tesla continues to produce Li batteries with Cobalt for their EVs, in a few years they may account for a significant share of world demand. 40% of world cobalt is used to make rechargeable batteries (Electrik 11/1/2016). I haven't seen a breakdown, but there are so many Li cells produced for use by phones, laptops and other EV producers, I'd guess Tesla to account for maybe 5 or 10% of the 40%. Thus perhaps 4% of world demand. Probably less.

More. You need around 125 grams per kWh for NCA and 350 grams per kWh for NMC. That means that once Tesla hits 150 GWh/year, at a 50/50 mix of NCA/NMC, Tesla will consume 35,600 metric tons of Cobalt per year.

At a global production of around 120,000 metric tons per year, Tesla will consume ~30% of the world supply of Cobalt. (Of course, supply will likely increase, as demand for cobalt rises.)

 

[racer26]

Deleted.

 

[Bobfitz1]

huh... there are no absolutes in life expect death and taxes. The point is that AP2 is not FSD1.0. The reason is that the amount of shared functionality is very small if you look at a Venn diagram, they would have a very tiny overlap and the AP2 circle would be very small and the FSD circle would be very large. AP is basically TACC with the ability to use Vision to see the lane lines and road edges (which is a tiny overlap with FSD). FSD will use HD3D maps more then lane lines, but uses lane lines as a backup or to confirm the HD maps. Imagine you are on a train and you are firmly on the tracks, that is HD maps. But you confirm that fact by looking with your eyes at the tracks as they stretch out in front of you (small overlap). HD maps will including things like stop signs, stop lights and exits where AP2 has no clue about any of those things and often darts into and out of exit lanes for no apparent reason. This is one way you will know when HD maps are being used as it will stop doing that in places that it has done it a million times. I know for me, I have a few roads I drive on where I can predict it will momentarily dart into a turn lane. No matter how many updates we have, this is a constant. Wasting anymore tech capital on that abomination is futile and I am all but certain Tesla has moved on.

I'm assuming you've seen Tesla released info that EAP and FSD will use a new code base currently under development and which takes advantage of AP 2.0 and 2.5 hardware. Has Tesla said when EAP will replace AP 2.0 software (on cars with requisite hardware)?
Architecturally will FSD when released work on top of EAP or replace it? It seems a bit odd Elon would make much of a coast to coast
no driver demonstration by end of year, if not even EAP has been released by now.

 

[SBenson]

I hope you are kidding. Do think the system has paper maps? Very confused. Are you even old enough to know what maps look like on paper? I guess im am having troubles knowing what a hardware map is if not for paper. So, yes, I am referring to software. You might be confusing AP2 with HW2. HW2 is hardware, AP2 is software. 3D maps are .... uhm.. software. Or possible tiny people with paper maps.

Ahhh yes.. there it is.. Sorry for the confusion. I assumed everyone knew the HW2 was the hardware part.

Well, you could very well be referring to a specialized hardware unit just meant for 3d-maps. A separate processor/disk/ram, who knows.... I didn't know what you are talking about so I asked.

If AP2 means software as you call it, then which software are you referring to? Tesla is rolling out software periodically with weekly versioning number. So if Tesla were to fundamentally revamp their software stack and push out over the air, are you expecting that Tesla will call it AP3? AFAIK Tesla doesn't do that. Apple does that with iOS, but Tesla only labels hardware with major versions, not software AFAIK.

Leaving all that aside, knowing that you are talking about software, all your commentary boils down to: the current software is not capable of x, y, z. The subsequent versions of software should be able to do these X, Y, Z... duh!

The real relevant question for investors and consumers alike is: Can the current HW2, HW2.5, etc which are being marketed and sold with the promise of EAP and FSD, really accomplish EAD/FSD (even including possible upgrades (on existing fleet)).

If the answer to that is 'no', then Tesla (and thus us) are in for some real s... That would be a massive issue. Tesla will lose a lot of good-will/brand image. Maybe there will be class action law suits. Could lemon law be invoked asking Tesla to refund the entire car purchase price? I don't know, I am guessing that's possible.

So when you say AP2 is done, if you are referring to only software then that is utterly irrelevant. If it is hardware, then that is some very serious alarm bells... Hence I was drilling in.

 

[Alketi]

So yeah, more on the Jeffries report.

I'm only about 2/3rds of the way through looking at it, but the sales volumes projection by model chart is really, really off.

They have 91k S+X for 2017. With 47k delivered thus far, they are thinking 44k for the rest of the year, or 22k per quarter. Yeah, that's likely to be debunked pretty soon. The low-ish S+X estimates goes all the way through 2021, with 94k in 2018, and 107k in 2019... this seemingly smallish miss is a big deal because of the gross margin. They are actually pretty bullish on Model 3 numbers for 2017... 45,000. A higher M3 versus S+X mix impacts overall gross margins in their model. For FY18, they have 300,000 Model 3's. That's not a bad conservative estimate of 6,200 to 6,500 a week. Could be low by 10%. The big issue is FY2019. They have only 380,000.

Well, my read is that Jeffries is more bullish than many for Model 3 numbers in 2017/2018.

It's unlikely that Tesla hits 45K this year and I would argue that given a normal "Elon discount" and any number of (un)foreseen production and logistics issues, that 300K units in 2018 is achievable, but a stretch.

However, I totally agree that 380K in 2019 is low. By then I expect issues to be sorted out and Tesla to be bringing another location online.

I also agree that the Model 3 demand is incalculable at this point and likely to be a very very high number. To be conservative, if the car performs as advertised, the demand should be more than 500K/year. Double that is a possibility.

But, what Jeffries seems to be missing, like so many other analysts, is the "trend" of what's happening. When the ICE obituary is written next year the perception funeral will be fully underway. Thus, Tesla won't be valued anywhere near a typical car company that (pre-funeral) enjoyed stagnant growth with non-existent innovation, and to just maintain this flatline state offered an endless sea of discounts and stupidly annoying advertisements (Happy Honda Days, Summerbrations, Decembers to Remember). That's an absolute comedy show and will be valued as such.

And lets also remember, the Model 3 will have no real competition before 2020.

 

[Bobfitz1]

More. You need around 125 grams per kWh for NCA and 350 grams per kWh for NMC. That means that once Tesla hits 150 GWh/year, at a 50/50 mix of NCA/NMC, Tesla will consume 35,600 metric tons of Cobalt per year.
At a global production of around 120,000 metric tons per year, Tesla will consume ~30% of the world supply of Cobalt. (Of course, supply will likely increase, as demand for cobalt rises.)

Thanks for the future stats. However my post was pointing out that Tesla does not currently account for a significant % of world Co demand,
not that it won't be significant once they hit 150 GWh/year. I gather the target for GF1 to reach 150 GWh is around 2020.
I posted earlier today that Li cathodes are being developed that don't need Cobalt at all. See NanoOne website. Not a done deal yet but promising. As are the battery cost reductions/performance increases their main process innovations promise.

 

[larmor]

SpaceX hints at satellite internet... (ignore if previously posted)
SpaceX might name its satellite broadband internet service 'Starlink'

 

[Ovulation]

Deleted.

 

[Waiting4M3]

SpaceX hints at satellite internet... (ignore if previously posted)
SpaceX might name its satellite broadband internet service 'Starlink'

not Skynet?

 

[Yggdrasill]

Deleted.

 

[Bobfitz1]

The real relevant question for investors and consumers alike is: Can the current HW2, HW2.5, etc which are being marketed and sold with the promise of EAP and FSD, really accomplish EAD/FSD (even including possible upgrades (on existing fleet)).

Hang on. As far as 'sold' aren't HW2 and now HW2 included as standard equipment in every Tesla? I thought it was the EAP and FSD promised features that are options costing thousands each. If I understand Reciprocity's post, he believes that there will be no further features added to AP2. Those features will not be released until EAP is ready and is pushed out to replace AP2 in cars with HW2 and/or HW 2.5..

 

[scaesare]

not Skynet?

I was surprised too, based on their datacenter:

 

[Bobfitz1]

Deleted.