2017 Investor Roundtable:General Discussion

[kbM3]

This is a choice. You can run them much more frequently than that, and China does run their HSR more frequently than that. And the reason for this is demand. It's possible -- nay, easy -- to build a train line whose capacity vastly exceeds demand. (The secondary reason is labor costs, which is why all trains should be fully automated like Vancouver Skytrain already is.)


Again, demand.

Unlikely. Loading takes time -- a lot of time for cars, though less time for humans. The deceleration profile is a critical issue. You have to be able to stop clear of the next pod (in case it suddenly fails hard, in place), and clear of a mislined junction. But the safe deceleration rate is limited by *passenger safety* (not merely by technical issues) -- no good to safely stop the pod and break all the passengers' necks. The deceleration distance therefore increases with speed.

You can do 2.5 minute spacing on a (passenger!) train system with a functioning safety system: Vancouver Skytrain is an example of a system which does exactly that. (And by the way, it's fully automated.) 2 minutes seems to be the practical limit. If you're going at higher speeds you will have to increase your spacing, though. If you're increasing the loading time (and yes, loading cars will increase your loading time) you will have to increase your spacing further.



Yes, it is easy. However, I'm pretty sure they didn't do it, because Musk would have put it up when he posted the rest of his napkin scribbles.

You have to remember that this is a guy who is pirating most of the software on the cars he makes because he couldn't be bothered to comply with the trivial conditions of the very short license agreement. You can't assume that he did something just because it's easy and straightforward.

There is no technical reason to load one pod at a time. I am sure you can envision a system where many pods are loaded simultaneously.

 

[NicoV]

There is no technical reason to load one pod at a time. I am sure you can envision a system where many pods are loaded simultaneously.

Loading time is irrelevant. You can take as long as you like, as long as you don't do the loading in the high speed part. The relevant metric is the time time it takes to insert a loaded pod in the high speed part safely. E.g. the Chunnel (the tunnel from France to the UK) takes like half an hour to load a train with cars (I forgot how long, it's too long ago that I went through the Chunnel), and probably only needs a minute or so to insert itself into high speed track/tunnel (shared by Eurostar high speed passenger trains, trains carrying cars, and trains carrying trucks).

 

[Menifeer]

OK, there are two reasons not to do this.

1. You're in the US. Short term trading like this converts long-term capital gains (taxed at 0%, 15%, or 20% depending on your tax bracket) to short-term capital gains (effectively taxed at 15% to 43.4%+ depending on your tax bracket). Including state taxes in NY, the top short-term bracket can be around 50%, while the top long-term bracket is more like 25%. In addition, short-term trading means you pay tax now rather than years and years in the future. That's stupid. Buy and hold gives you better tax treatment.

I had a share of TSLA purchased at $180.
If I hold it until it goes long-term and then sell it at $380, I *eventually* get a $200 profit, pay $50 in taxes, and realize a net of $150.
If I sell it at $380 while it's short-term, and then buy back at $320, and later sell again at $380 (while it's still short-term), then
-- today I get a $200 profit, but pay $100 in taxes now, for a net of $100
-- later I get a $60 profit but pay $30 in taxes for a net of $30
So I'm down $30 versus buy-and-hold, and in addition, I have to pay $100 now rather than years later (I could invest that $100 and make money off it).

2. Market timing is hard. If you sell and the stock never dips, you just lost the long-term appreciation. There's a story someone told here of a friend who sold his TSLA at $35 just before the big runup. Ouch.


Those options are *unqualified*, so all your stock holding periods are short-term.


Short term capital gains. If the stock isn't taken away, still short-term capital gains.


The option prices are high due to high expected volatility (puts are high-priced too).

You have to figure out whether the loss of long-term-capital-gains treatment and the loss of deferred taxation is outweighed by the option premiums. For me, it wasn't, which is why I sell puts against cash but I don't sell calls against the stock.

About taxes...

The short-term vs long-term tax treatment is irrelevant. I make money on short-term trades (due to the high call option values) when the stock price may or may not change. Why should I care if the tax rate is higher than the long-term rate when, if I didn't do the transaction, there would be no profit to tax?

About market timing...

I'm not trying to time the market. I simply buy the stock and sell the calls early in the week. On Friday I usually have the shares taken from me at a profit (the short-term tax on which, I'm happy to pay).

 

[RobStark]

Parking shouldn't be an issue once full self driving is rolled out. Tesla would just have a few cars running around picking up people at home.

Tesla has ~6500 employees at Fremont to make 100k cars per year.

How many employees will it take to make 800k-1,000k cars per year?

How many FSD Model 3 will be required to make the parking at Fremont manageable?

And how long before we have level 5 Model 3 that can go to pick up an employee without a driver?

Assuming it is legal for a car to drive itself when Tesla has achieved a geofenced Level 5 in greater Fremont.

 

[Yggdrasill]

Tesla has ~6500 employees at Fremont to make 100k cars per year.

How many employees will it take to make 800k-1,000k cars per year?

10k, maybe?

How many FSD Model 3 will be required to make the parking at Fremont manageable?

Say you decide to transport the 5000 employees living closest to the factory, each car does 5 trips carrying four people each time. That would require 250 autonomous Model 3.

And how long before we have level 5 Model 3 that can go to pick up an employee without a driver?

Next year, maybe?

Assuming it is legal for a car to drive itself when Tesla has achieved a geofenced Level 5 in greater Fremont.

That might take more time. Maybe 2019/2020?

 

[kbM3]

Loading time is irrelevant. You can take as long as you like, as long as you don't do the loading in the high speed part. The relevant metric is the time time it takes to insert a loaded pod in the high speed part safely. E.g. the Chunnel (the tunnel from France to the UK) takes like half an hour to load a train with cars (I forgot how long, it's too long ago that I went through the Chunnel), and probably only needs a minute or so to insert itself into high speed track/tunnel (shared by Eurostar high speed passenger trains, trains carrying cars, and trains carrying trucks).

I am not sure you are disagreeing with me or not. The pertinent fact is that pod loading will not be a bottleneck. Nor will pod insertion. I assume with automation, they can have very short distance/time between pods.

 

[ValueAnalyst]

10k, maybe?

Say you decide to transport the 5000 employees living closest to the factory, each car does 5 trips carrying four people each time. That would require 250 autonomous Model 3.

Next year, maybe?

That might take more time. Maybe 2019/2020?

Again with overestimating time for regulation.

Didn't we all just read that the congress is pushing trough a safety exemption for 100,000 autonomous cars per manufacturer by the end of the year?

 

[Rarity]

If you have 8-9 minutes this is a great video, She gets it and articulates the whole Tesla ecosystem very well.

This a good, level-headed analysis, but isn't Tesla well on its way to solving the charging infrastructure issue without government involvement?

 

[NicoV]

I am not sure you are disagreeing with me or not. The pertinent fact is that pod loading will not be a bottleneck. Nor will pod insertion. I assume with automation, they can have very short distance/time between pods.

No disagreement, should have replied to the same post of Neroden you commented on.

 

[MitchJi]

I rolled my eyes at this when I realized that his "smaller diameter tunnels" were the standard diameter of London Underground tunnels.

Reducing the tunnel diameter by half means that you are only required to remove one fourth of the material.

In order to guarantee that anything which slows the first slows them all, they must be mechanically connected. (Consider a catastrophic physical failure in the first one. The second one must be able to stop before plowing through the wreckage. Nobody has come close to designing a "virtual linkage" which is sufficiently failsafe.) If you have mechanically connected them, guess what you have. A train.

I partially agree with you to the extent that it don't think it makes sense to transport cars with people via hyper loop from NYC to DC.

OTOH you just made the case that hyperpod cars, which could be connected as demand requires would make a feasible system. I believe that a few weeks ago that you were stating that that was absurd?

 

[winfield100]

right now the Elon Musk interview is being redone on CSPAN radio, that plays in Washington DC, from the National Govenors Association conference. @ 5:45pm DC time. it is getting press a lot. (yay!!)(the one where he says 3 more gigafactories, in the US, lots of jobs)

 

[MitchJi]

Say you decide to transport the 5000 employees living closest to the factory, each car does 5 trips carrying four people each time. That would require 250 autonomous Model 3.

Tesla could provide parking spaces at a remote site and provide shuttles from there.

 

[Navin]

Why I remain optimistic about Tesla

 

[Bet TSLA]

Why I remain optimistic about Tesla

From that link: "To me, other electric vehicles, such as the BMW i3, the Mercedes B-Class, the Nissan Leaf and the Chevy Bolt, all of which I have driven, seem by comparison to be a clumsy repackaging of old technologies."

I just test drove the soon to be released electric Honda Clarity. That accurately describes my impression. My first car was a Honda. I love Honda. But if this is their best effort, they're doomed.

 

[Reciprocity]

When you say climate alarmists, I assume you mean the following:
Every respected university.
Every scientific institute.
Every well respected scientific publication (scientific American, popular science, MIT technology review…)
Every famous scientist.
Department of defense.
NASA.
JPL.
And 97% of climate scientists.

I think you miss the point, not everyone is the groups you mentioned. Not everyone shares the same beliefs or values. And Tesla wants to reach the other 7 billion people that are not in those groups as a ways to a means. Instead of just trying to inspire people to save the world, they can also inspire them by selling them a S3XY vehicle or a fancy roof that makes there home look better. It's a great way to expand the market share.

 

[Turing]

@elonmusk hit 10.5M followers, didn't he just hit 10M last week?

 

[Navin]

Vehicle technology: Tesla and Elon Musk’s moment of truth

Free version

Moment of truth for Elon Musk as Model 3 hits the market

 

[Waiting4M3]

Same, im out of money. Hoping things stay fairly constrained until Jan. 1 when I can drop more into for my IRA as a 2019 contribution as I have already made my 2018 contribution. Long and strong all the way!

sorry buddy, but I would like to cash some $ out to buy my M3, I hope it goes way up before Dec.

 

[Waiting4M3]

I agree with Mitch. It would make NO sense from a software maintenance or code standpoint to have two entirely separate systems, duplicating work, fixing two different sets of bugs, etc. It would be both highly inefficient and ultimately dangerous.

It's much more likely that the lane-keeping behavior you see now is the same that will be used in FSD. The differences are that FSD will incorporate street/traffic sign awareness, pedestrians, parking lots, etc. etc.

Musk is clearly counting on a step-change behavior in AP to hit his end of year deadline. Tesla, to note, did just turn on the data spigot in May, and this last release, if the version number change is an indication, appears to be Karpathy's first. It made some noticeable improvements, but it still has a long way to go.

While discussing the recent Audi autonomous level 3 announcement, a coworker of mine brought up an interesting thought, that an autonomous car may need a backup system that is completely separate from the primary system, in order to be fault tolerant. If the primary system encounters a scenario that it fails at, if the backup system runs on the exact same logic, it may fail the same way. From this perspective, there may be some merit to thinking that a separate code base may be of use.

 

[Reciprocity]

Tesla has ~6500 employees at Fremont to make 100k cars per year.

How many employees will it take to make 800k-1,000k cars per year?

How many FSD Model 3 will be required to make the parking at Fremont manageable?

And how long before we have level 5 Model 3 that can go to pick up an employee without a driver?

Assuming it is legal for a car to drive itself when Tesla has achieved a geofenced Level 5 in greater Fremont.

You could have a level 5 model 3 today as long as it's geofenced to the factory. They could literally load themselves into the trains or drive to where they will be delivered in the Fremont facility. It's much easier to train a machine for a single case and not allow humans along those routes.