2017 Investor Roundtable:General Discussion

[sundaymorning]

Shorts running out of hope, no shutdown:

White House vows to continue funding Obamacare, averting showdown with Democrats

 

[neroden]

I've visited Porsche factory last summer and couple of things stuck with me. Final assembly was the same line for all sport cars boxsters/caymans/911, and line was sequential; speed of the line was equal to slowest operation on the line (2 minutes something) and there were many stations were people were quicker and just waiting for line to move. Also, line had 123 stops (~), only one model needing them all, 911 Turbo Cab model. Simpler model apparently needed only 90 something operations. So tremendous waste...

I have a strong suspicion that Tesla is avoiding this particular problem, at least for now, by eliminating options. ("You can get it in any color as long as it's black" and all that.)

 

[neroden]

I hope this is helpful. Please keep challenging me with thoughtful reasoning. I very much appreciate it.

I think it probably qualifies as nitpicking for me to say that I don't think Tesla will be at 4 million per year in 2020, but rather in 2021 or 2022. I gave my reasons already (basically, unanticipated delays always happen). I realize this isn't a big difference in timeline!

 

[NicoV]

Those pictures trouble me. I honestly thought they were much further ahead. Nearly 500 robots is not some trivial part of the line that they need to fill in. It's a major undertaking. Especially programming the robots may not be trivial (depends on the complexity of that particular station and how much two or three robots need to be synced). Another big risk is that they don't have enough time to optimize the production line and hence will produce for a longer than planned period at lower capacity. Or, worse, have to shut down the line for a week to optimize everything like they had to do for S/X once or twice a year. At least on two separate occasions it created major production lapses.

Progamming the robots in this stage is probably just uploading the program to the robot. Developing the program has most likely already been finished, and you just need one robot to test the all the programs for all those robots.

 

[Ulmo]

I've visited Porsche factory last summer and couple of things stuck with me. Final assembly was the same line for all sport cars boxsters/caymans/911, and line was sequential; speed of the line was equal to slowest operation on the line (2 minutes something) and there were many stations were people were quicker and just waiting for line to move. Also, line had 123 stops (~), only one model needing them all, 911 Turbo Cab model. Simpler model apparently needed only 90 something operations. So tremendous waste...

Couple of conclusions from that experience. Improving some of the operations to be much quicker is not necessarily helping a whole lot. You eliminate few steps, you automate few operations, your bottleneck is still the slowest work station.

Can you create few parallel stations for the slowest operations? Maybe, I don't know if anyone is doing it. I don't know if Tesla wants to handle this complexity, as your cars can change the sequence of arriving(or could they not?), and your supply line software need to understand it. You want to avoid just in time parts delivery system bringing you red parts for the blue car, just because blue car overtook red car at the previous station.

Can you attack 10-15 most complex stations and optimize/automate them, so that other 90 stations are all sub 1 min? I don't know, but only then we'll start seeing 2x improvements. Road to 10x is mind boggingly complex, unless you have almost full automation and few tasks that people need to do, you parallelize to the highest extent possible.

Ok, to summarize, partial automation probably brings only incremental, maybe even only marginal improvements. To capture true benefits, the whole line needs to be considered at the same time. So scaling the line is probably step change, but considering Tesla has been thinking about this for awhile, they're probably starting with the line that is proper subsystem of the final solution (0.5 solution though), i.e. part that is built first will be forward compatible

I never understand why we must always have an assembly line with the product moving around rather than a building pad with the parts flowing around and general purpose robots coming to stationary built cars* when needed. Some robots would be generally used often such as place gasket, glue, weld, feed wires, solder, clip bolt and screw, etc., but other specialty robots could move around to give room for other specialty robots to also move around. This way, space could be used more efficiently, by having the entire 3D to 4D space used up by flowing parts. The entire floor would be nothing but parallel build pads, with each build pad being one whole vehicle. Because of the spaces involved, it would be easy to allow large size vehicles in this format. In this way, the assembly line never slows down because of something else in the chain of product assembly; the only slowdowns would be when parts become unavailable, and that just means vehicles being built that don't need that part will be the ones to complete first and the factory will output less, not stop.

* The cars would probably be mounted on robots that could move it around as needed within its build pad, but wouldn't take the car away until it was finished (or delayed).

 

[neroden]

I think s&p inclusion would significantly increase the liquidity (which would limit constant manipulation by shorts) but it wouldn't necessarily increase the valuation of the company.

I disagree. There have been studies: inclusion in the S&P causes an immediate stock price bump, which sticks. This is because there's *so* much money in S&P funds -- something like 4% of the market cap of the S&P -- that the result is that 4% of TSLA stock will be bought by S&P funds and held forever. This drop in supply raises the price.

Unfortunately, the other effect is that general inflows and outflows into the S&P will now affect TSLA, so TSLA will become more responsive to macro trends.

 

[neroden]

There also are a lot of funds who can only invest in companies included in major indices. All I'm saying is the bid/ask spread should decline dramatically if TSLA gets included in S&P500.

Ah, that's actually quite different from having more liquidity. I've seen illiquid stocks with narrow spreads and liuqid stocks with wide spreads...

 

[neroden]

Such inclusions/exclusions don't happen very often with companies that have 50B+ market caps.

When TSLA gets added to the S&P, all of the S&P tracking funds (of which there are many) have to buy it. That will put buying pressure on it. The size of those funds means that they will have to acquire something around 5-10% of TSLAs outstanding shares in very short order, and essentially remove it from circulation.

TSLA typically only trades something like an average of about 4% of its outstanding shares per day. Trying to acquire 5-10% of TSLAs shares in a short time span will drive the price up significantly - especially when so many of the shares are already in the hands of long-term holders. This is essentially the same concept as 'days to cover' for shorts - someone is forced to buy a lot of shares (like more than the daily average volume) in a short time span, and the result is that its very difficult to do that without dramatically driving the price up.

That should also cause a number of the shorts to want to head for the exits since the effect of those shares being bought by the S&P trackers is removing them from the float, which will make it harder for the shorts to exit in an orderly fashion when someone cries wolf.

Someone with more experience than me might be able to say how it works - I'm not sure how long those index tracker funds will have to acquire their shares, or how long before it is added to the S&P we will know about it.

I can't think of an example of a stock with such high short interest being added to the S&P, but the last example of a *really damn huge* company being added to the S&P was Berkshire Hathaway. (It had been excluded due to low liquidity and a monumentally high per-share price; it was added after the BNSF merger lowered the price of B class shares, IIRC)

It was... disruptive to the stock price. It went up.

 

[neroden]

unless they've changed it recently, the funds that want to track the index have to get filled on the closing print the day of the add. it's literally as simple as an order like "buy 10 million tesla market on close."

Soooo, there are "strict tracker" funds which try to do this.

But since that quite consistently means the brokers are unable to fill the orders and the funds suffer underperformance, an awful lot of index funds try to front-run the addition themselves and get the stock a little *below* the closing print, which means a little early. So you usually have huge buy orders all day long, both from the front-runners andf rom the "looser" index funds.

And then you have massive after-hours activity by the funds which didn't get filled at close. It's considered embarassing to not have managed to add it to your index tracker fund the day after it was added to the index...

because it's known roughly the size of the order and the time it will arrive, everyone who can reasonably frontrun the trade does, and then they flip shares back to the index funds.

if too much of this happens you can get a short term top in the stock (or even a long term top) as happened with jds uniphase years ago. on the other hand if a number of funds say, "we'll gamble on being able to buy it for less later" and don't pay that closing print... and the stock starts to move, why then you can have some further sustained movement as the stock forces a long squeeze (longs have to get in or risk tracking error/underperformance).

Mmmm. The only thing which prevents massive upward price movement at that point is massive shorting activity that day!

 

[Yggdrasill]

Here comes the Cobalt spike already.
Electric Bar Boom Sets Cobalt Up For Massive Price Spike

Also, when it comes to Cobalt, NCA uses around 125 grams per kWh.

That means a 60 kWh vehicle needs 7.5 kg. At 26 USD/lb, that's 430 USD (or 7 USD/kWh). Sure, that's a meaningful part of the cost of a vehicle, but it's not the factor that will make or break the BEV revolution. Cobalt prices need to go vertical for the cost of Cobalt to be an issue.

And with every price increase, more effort will be put into using less Cobalt. A number of chemistries don't use Cobalt, including LFP, LTO, Li-S, Li-Air. LFP is probably the most relevant right now - BYD is producing over 10 GWh per year, and a lot of this is going into cars and buses.

Edit: If Cobalt prices do go vertical, LG Chem is pretty screwed. NMC uses around 350 grams per kWh, or 21 kg for each GM Bolt. Lets say Cobalt prices quadruple, that means:

GM Bolt:
Cobalt cost goes from 1,200 USD to 4,800 USD
Cost of car goes from 37,500 USD to 41,100 USD

60 kWh Tesla Model 3:
Cobalt cost goes from 430 USD to 1,720 USD
Cost of car goes from 35,000 USD to 36,290 USD

How long before LG Chem is forced to go to NCA or LFP for vehicles?

 

[GoTslaGo]

China's Didi Said Near Deal to Become Most Valuable Asia Startup

quote:

The company hopes that driverless technology could help it overcome these hurdles in the future. Didi wants to take advantage of data on 300 million users across some 400 cities. It opened an artificial intelligence lab in Mountain View, California last month, called Didi Labs. It’s already lured dozens of stalwarts in the field including former Uber auto-security expert Charlie Miller, known for remotely hacking into a Jeep Cherokee in 2015.

Didi counts more than 100 investors as backers including Tencent Holdings Ltd., Alibaba, Tiger Global Management and China’s sovereign wealth fund China Investment Corp.

End quote.

Looks like they are trying to develop Autonomous driving too.

 

[Ovulation]

I recently stumbled into an very good made and insightful Tesla podcast (no I have nothing to do with it). In their last episode they echoed my line of thinking of why Tesla value is justifiably, considering the scope of the car market and Teslas margins. Check it out, do you find any flaws in the arguments?

54 – Tesla's Valuation

 

[schonelucht]

It's interesting that this quote is from Fanuc, not Kuka, who is installing the 400+ robots for Model 3. Both are current suppliers to Tesla. Does anyone know if one supplier does the current BIW lines and the other does the S/X final assembly? It might give us a clue as to the purpose of the robots that we saw pictured this week.

Thanks for digging that up. I think I have the same questions as you have. How do you know both are current suppliers? Maybe Fanuc was not able to deliver and Kuka stepped in?

 

[schonelucht]

Progamming the robots in this stage is probably just uploading the program to the robot. Developing the program has most likely already been finished, and you just need one robot to test the all the programs for all those robots.

I have witnessed over a dozen industrial robot installations and this was literally never the case. All of these installations were planned, drawn in CAD and simulated down to the last details. Yet, it was never as simple as uploading a program at install time and you're good to go. For one, you need an exact replica down both in weight and size to realistically simulate the timing of your line movements, which influences the order in which two robots working together must make their movements etc etc.

 

[ValueAnalyst]

I disagree. There have been studies: inclusion in the S&P causes an immediate stock price bump, which sticks. This is because there's *so* much money in S&P funds -- something like 4% of the market cap of the S&P -- that the result is that 4% of TSLA stock will be bought by S&P funds and held forever. This drop in supply raises the price.

Unfortunately, the other effect is that general inflows and outflows into the S&P will now affect TSLA, so TSLA will become more responsive to macro trends.

You're assuming the following:

1. There are no hedge funds that may have bought TSLA anticipating that it will be included in the S&P500, only to sell right after inclusion.

2. The higher SP due to inclusion will not entice any existing shareholders to sell.

3. The higher SP due to inclusion will not bring in any new short sellers.

So yes, massive index funds will flow, but this is an easily predictable event, for which many money managers prepare on the other side.

 

[ValueAnalyst]

I expect Ford to file for bankruptcy in 2019.

Ford’s Profit Falls 35% on Recall Costs, Weaker U.S. Sales

 

[jhm]

I never understand why we must always have an assembly line with the product moving around rather than a building pad with the parts flowing around and general purpose robots coming to stationary built cars* when needed. Some robots would be generally used often such as place gasket, glue, weld, feed wires, solder, clip bolt and screw, etc., but other specialty robots could move around to give room for other specialty robots to also move around. This way, space could be used more efficiently, by having the entire 3D to 4D space used up by flowing parts. The entire floor would be nothing but parallel build pads, with each build pad being one whole vehicle. Because of the spaces involved, it would be easy to allow large size vehicles in this format. In this way, the assembly line never slows down because of something else in the chain of product assembly; the only slowdowns would be when parts become unavailable, and that just means vehicles being built that don't need that part will be the ones to complete first and the factory will output less, not stop.

* The cars would probably be mounted on robots that could move it around as needed within its build pad, but wouldn't take the car away until it was finished (or delayed).

Interesting idea. I would think that Tesla would be interested in maximizing the output per volume of manufacturing space. I suspect it largely comes down to minimizing work of moving the product, robots and supplies around. Perhaps in the assembly of semi trucks you get to a scale of product where moving robots and supplies is less work than moving the product. I do likenthe idea a few generalist robots that can do most of the assembly. It seems these could be supported with assistant robots that always assure that tools and supplies are within reach of the general. These assistants can be more optimized to fetch the things they must carry. They could also travel along tracks mounted from above to optimize use of vertical space.

Fun to imagine what this would look like.

 

[jhm]

I expect Ford to file for bankruptcy in 2019.

Ford’s Profit Falls 35% on Recall Costs, Weaker U.S. Sales

But, but,...their market cap per vehicle is so much lower than it is for Tesla.

 

[D-egg-O]

I recently stumbled into an very good made and insightful Tesla podcast (no I have nothing to do with it). In their last episode they echoed my line of thinking of why Tesla value is justifiably, considering the scope of the car market and Teslas margins. Check it out, do you find any flaws in the arguments?

54 – Tesla's Valuation

I love their podcast.

 

[Bobfitz1]

Thank you for challenging me with a reasonable argument. I highly appreciate it as this helps me tremendously.

The following are the reasons why I see Tesla moving forward simultaneously and quickly on "Gigafactories 3, 4, and possibly 5."

3. You raise a good point about capex needs to achieve such a goal, but if one actually thinks about how much capital is needed, and compares that to the benefits, one quickly realizes that (given that demand is there) Tesla should be building as many Gigafactories as possible. Basically, starting a Gigafactory costs a few hundred million dollars and a few billion over the full course of the project with majority of the cost being in tooling (so towards the end of the construction project). So the cost of starting all three simultaneously would be maybe a billion dollars. If the Model 3 ramp-up goes smoothly, as it seems to be so far, and Tesla produces 1 million cars over the next ~20 months, which I think is a strong possibility, that's $50B+ of revenue. Even if the gross margin is ~15% in earlier days of the Model 3 ramp up, that's many more billions than what it would take to start the Gigafactories 3, 4, and possibly 5. Further, I expect Tesla to now be able to take on non-dilutive corporate debt as its revenue multiplies in the next 12 months. This simple back-of-the-envelope calculation looks even brighter for Tesla beyond 2018 as the Model 3 production line (and gross margin) reaches maturity. Much much brighter.

Excellent post. All your points make sense. On point 3, capex needed to start multiple Gigafactories, as well as continue them into the high capex mid and end phases, would be reduced a good deal if Tesla and Panasonic continue partnering as they have to date.
Panasonic has put real skin into the game with both GF1 and GF2. I can't think of any reason why they would not want to continue doing so or why Tesla would not want them to continue providing some of the necessary GF capex for 3 - 5.