Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[Buckminster]

A few bits of good news:

UK will stop all new Fracking projects going forwards. Significant step towards a greener future in UK - Greta is working her magic.

Tesla vehicles (not minibuses) to be used in Washington - Baltimore tunnel. Go-live 2021 I think.

UK's sixth biggest city, Bristol, might go diesel free (private vehicles) in 2021

 

[Dreadnought]

Be prepared for some fireworks when Telsa releases the first HW 3.0 optimized updates. Whether this leads to robotaxis in short order is anyone's guess but I'm betting it will put Tesla squarely in the driver's seat in the race to FSD. While my investment in TSLA is not predicated on FSD, it certainly provides a huge opportunity to excel in a manner completely unanticipated by the short-sighted and unimaginative naysayers.

This!
As much as analysts fail to see the progress in FSD capabilities today, once thousands of happy Tesla owners share their hands-off commuting experience, this should add a Waymo x 10 valuation on top of TSLA, even if L4 let alone robotaxi is still years out. Not betting the house on it, though.

OT but yet à propos: Had a lot of sudden EAP disconnects lately when driving towards a low sitting sun. Hazard lights went on until car recognised steering input. Hope that HW3 will not be blinded as easily.

 

[Swampgator]

Absolutely none of that is really correct. At this moment in time, Tesla has the best battery technology available. But in just a few years that can easily change. There is no reason to believe Tesla can ramp up car production while major companies 20 times their size can't ramp up battery production.

The major auto makers will spend HUGE amounts of money on designing the best batteries because they CAN. They have the money and will spend it because they are realizing they can make good EVs or they can become the 21st century Kodak.

You also need to realize that a small difference in battery technology won't make such a large difference in sales. Cars are sold through marketing. Small differences in technology are easily mitigated.

Nice try. Legacy OEMs are for sure caught in the Innovators dilemma. The real reason they do not try and scale to match TSLA is because they lose $$$ on every EV they make. Period. TSLA has fount a way to make EVs profitably. The legacy OEMs have bought their EV tech from LG and others. This is expensive as LG wants to make money too. How do they become better vertically integrated to lower costs? This is a problem because it's not in their culture. They have all been outsourcing for 40 years. Throwing $$$$ at the problem does not solve any of this.
And think about the billions of dollars of obsolete ICE equipment they all have. How do you get rid of that without suffering huge loses?
They are toast. Except maybe Porsche because they are small volume, high margin OEM. They only need to scale Taycan to 40,000 a year globally and will sell them at a profit. How does VW scale their ID-3 to 300k a year and make a profit? They can't. That is why they are not even bothering to sell that car in the US.

 

[Dreadnought]

The screenshot picture of the S cockpit during the race posted above by KarenRei (#98185) is pretty damning ... Range Mode, really ???

Oh, that is easily explained: After running out of charge testing the roadster, they just wanted to really be sure that the battery lasts for the entire quarter mile.

 

[kbM3]

Elon has tweeted that the new network only waits regulatory approval which should be ready in Q4.

In cars we see HW2 networks. In demos we have seen HW3 networks. HW3 network seems much more capable given the higher resolution, multiple frames, all cameras being used, camera agnostic layers(learn to see cars from all angles). They had it working in beta 3 years ago. They dared to do public demo with hundred of rides 1 year ago. They have had a fleet of 100-200k cars with HW3 gathering data in shadow mode for a year.

In conclusion: It seems pretty likely that HW3 networks will perform significantly better than HW2 networks, enough that FSD and Robotaxi might actually be viable 2020. Demos and Elon indicates this. Current HW2 performance might not give the full picture.

My likelihood estimation of this? 10%? 50%? 90%? It’s really hard... Just by Elon I would say 75%. By share price 0%. By wisdom of crowds 5%. From demos 50%. I will weight these together into my own estimate which is 20%. Yes that is high. But imo I don’t think most people have really grasped the magnitude of what I wrote above...

I do not recall Elon tweeting anything about regular regulatory approval in Q4?

Investment community should wake up when they realize with city NOA, Tesla is on par with Waymo but globally instead of in a few places. But who knows what the market will do ....

I hope you’re right but that’s not my guess.

I think they will release with an intervention rate of 1 per 100 miles, and many, many articles will state how far behind Waymo (1 per 14,000 miles?) they are. We all know Tesla is solving a FAR more difficult problem, but I believe that will go over the financial analyst’s heads.

I then believe they’ll improve rapidly, but i’d definitely be shocked if we get to 1 safety intervention per 4M miles (rough guess on what’s needed for no supervision) in Elon’s time frame. I fear exponential corner case explosion as you drive down the intervention rate. Hopefully Karpathy and team can solve increasing numbers of corner cases without greatly increasing labeling time, development hours...

 

[kbM3]

Twitter

Great video from @verygreen (who has hacked his Tesla to show AP feeds) showing AP detecting a pedestrian in black crossing the street at night who’s very difficult for the human eye to see.

 

[davepsilon]

I hope you’re right but that’s not my guess.

I think they will release with an intervention rate of 1 per 100 miles, and many, many articles will state how far behind Waymo (1 per 14,000 miles?) they are. We all know Tesla is solving a FAR more difficult problem, but I believe that will go over the financial analyst’s heads.

I then believe they’ll improve rapidly, but i’d definitely be shocked if we get to 1 safety intervention per 4M miles (rough guess on what’s needed for no supervision) in Elon’s time frame. I fear exponential corner case explosion as you drive down the intervention rate. Hopefully Karpathy and team can solve increasing numbers of corner cases without greatly increasing labeling time, development hours...

They are definitely behind waymo. But they do have a different approach in using vision only and a million car scale data gathering network. There's still a lot of risk as there are certainly development paths in front of them that they could choose that don't reach the end state. That's how difficult problems go, it's not a linear path to completion some effort will be highly productive some effort will be wasted.

Tesla does seem to be in the lead in terms of driver assist self driving, so supervised self driving. I wouldn't be surprised if the lidar equipped cars reach general unsupervised city driving first over the coming years. I also won't be surprised to see more decent supervised self driving options from other manufacturers. In short I think we're likely to see step progress and multiple competing offerings in FSD over the coming years.

I think for now there is too much uncertainty in the space for me to assign much investment value to teslas FSD efforts. But when I see city NoA that could change.

Yet, in the scheme of things, they aren't spending that much on the R&D so it's like an extra lottery ticket for the stock imo. Incidentally one of two since, from a business standpoint, as a lottery ticket is how I'm evaluating Tesla energy until we see more scale in that business. Both have immense potential but so far very limited evidence for what these pieces will be worth to the business.

 

[electracity]

Tesla dominates Auto Software by far... Thats more important than the manufactoring process which, at least for ev's, they dominate too.

Not just software but the specialized electronics controlled by that software. It's the continuum from application programming to electrical engineering that is so impressive. For Tesla it is just how they roll. For traditional automakers it's a whole new world. Senior execs in the auto industry don't have much experience managing programmers and EEs.

Conversely traditional automakers still do many important unglamorous tasks better than Tesla. Like painting a $17K Yaris better than a model X.

 

[jbcarioca]

you never know. Mobile payment took China by storm and nobody predicted it. It is very hard to believe that a country where people almost exclusively rely on cache, no debt/credit card, no personal checks, can leap to mobile payment in such an astonishing speed.

Factually an earlier and more instructive case probably is M-Pesa in Kenya:
Mobile Phones, Tablets, Mobile Money Solutions & Telecommunication Services - Safaricom
From essentially non-existent financial services for poor people, M-Pesa became the dominant payment mechanism is only a couple of years after the 2007 launch. Using prepaid mobile phone balances was the innovation. Thinking of disruption M-Pesa is beyond question the financial innovation of this century thus far. It even later became the method used to pay Afghan soldiers, displacing cash.

It seems to me that Tesla is to cars what M-Pesa is to banks. It redefines the terms in ways that are transparent, obvious and still ignored by the very antiquated entities that are being disintermediated. Recall back to pre-Model 3 pre-Model X times please. People who bought Model S back then often paid a multiple of any price they had ever paid for a vehicle. Even the higher end MB/BMW did not really notice what was happening, because the early Model S buyers were so very disparate that no specific loss of market share was evident. Once higher volumes and more models appeared the world suddenly say the threat.

The could copy but they simply don't understand how to do that. Just as in the case of M-Pesa it took almost a decade before banks understood they were losing business. earlier they only were not getting new customers, not losing existing ones. Never underestimate the myopia pf bureaucrats.

 

[Dan Detweiler]

They are definitely behind waymo. But they do have a different approach in using vision only and a million car scale data gathering network. There's still a lot of risk as there are certainly development paths in front of them that they could choose that don't reach the end state. That's how difficult problems go, it's not a linear path to completion some effort will be highly productive some effort will be wasted.

Tesla does seem to be in the lead in terms of driver assist self driving, so supervised self driving. I wouldn't be surprised if the lidar equipped cars reach general unsupervised city driving first over the coming years. I also won't be surprised to see more decent supervised self driving options from other manufacturers. In short I think we're likely to see step progress and multiple competing offerings in FSD over the coming years.

I think for now there is too much uncertainty in the space for me to assign much investment value to teslas FSD efforts. But when I see city NoA that could change.

Yet, in the scheme of things, they aren't spending that much on the R&D so it's like an extra lottery ticket for the stock imo. Incidentally one of two since, from a business standpoint, as a lottery ticket is how I'm evaluating Tesla energy until we see more scale in that business. Both have immense potential but so far very limited evidence for what these pieces will be worth to the business.

Behind Waymo? I disagree.

Waymo works in a very small environment requiring detailed mapping and they still need a driver to supervise. That's still driver assist. Tesla works (within its designed parameters) anywhere. Yes, they have significant hurdles left to clear but to say they are behind Waymo is disingenuous at best. Time will tell which approach unlocks the problem but Tesla has proven that it has a path to the goal. Waymo has proven that it will need to detail map the entire planet before it begins to come close. What happens when those maps are no longer accurate? Then they still have the fringe scenarios to deal with.

Dan

 

[electracity]

But they do have a different approach in using vision only and a million car scale data gathering network.

Yet auto wipers still isn't great in light rain. They presumably have access to millions of events where auto wipers is overridden, but they can't train the system to be as good as old school dedicated sensors. Light rain is a fairly trivial edge case and not remotely as complicated as the complex interaction of objects in the driving environment.

Even when FSD functions seem fantastically capable it still won't be ready for release. Running over only 1 in a 1000 pedestrians in the driving path isn't nearly good enough. Waymo must be much better than 1 in 1000 or they would not have removed the safety driver.

 

[davepsilon]

Behind Waymo? I disagree.

Waymo works in a very small environment requiring detailed mapping and they still need a driver to supervise. That's still driver assist. Tesla works (within its designed parameters) anywhere. Yes, they have significant hurdles left to clear but to say they are behind Waymo is disingenuous at best. Time will tell which approach unlocks the problem but Tesla has proven that it has a path to the goal. Waymo has proven that it will need to detail map the entire planet before it begins to come close. What happens when those maps are no longer accurate? Then they still have the fringe scenarios to deal with.

Dan

Waymo publishes disengagement reports (tesla doesn't and that increases my uncertainty for their FSD value) and waymo is crushing it. If robotaxis in cities are as valuable as some here theorize then the cost of mapping and the lidar system aren't show stoppers.

You say you disagree on the current state of affairs but only discuss a future capability. As of today Tesla has revealed plans to have supervised city driving. Waymo has driverless city driving in Arizona. (They imply unsupervised but it is possible there is remote supervision)

 

[jerry33]

Yet auto wipers still isn't great in light rain. They presumably have access to millions of events where auto wipers is overridden, but they can't train the system to be as good as old school dedicated sensors. Light rain is a fairly trivial edge case and not remotely as complicated as the complex interaction of objects in the driving environment.

Even when FSD functions seem fantastically capable it still won't be ready for release. Running over only 1 in a 1000 pedestrians in the driving path isn't nearly good enough. Waymo must be much better than 1 in 1000 or they would not have removed the safety driver.

This was true six years ago when the car was delivered. It's no longer true.

 

[electracity]

This was true six years ago when the car was delivered. It's no longer true.

model 3

 

[Dan Detweiler]

model 3

It's odd, I have never had a problem with the auto wipers in light or heavy rain. I think it is more a personal tolerance issue with individual drivers. Just my opinion though.

Dan

 

[mongo]

Yet auto wipers still isn't great in light rain. They presumably have access to millions of events where auto wipers is overridden, but they can't train the system to be as good as old school dedicated sensors. Light rain is a fairly trivial edge case and not remotely as complicated as the complex interaction of objects in the driving environment.

The cameras have their own glass heater and power dissipation. https://teslamotorsclub.com/tmc/threads/heated-forward-facing-camera-housing.141265/ In light mist, that can keep the windshield in front of them clearer than that of the annoyed driver. (See also: bug splat in front of passenger)

The real functional objective is not whether the humans consider the windshield clear, but whether the AP NN can function properly with the level of obscuration (see also dirty glasses). And if not, whether it detects that and triggers a wipe cycle. Next level operation will be wipe+wash to remove road salt.

 

[electracity]

It's odd, I have never had a problem with the auto wipers in light or heavy rain. I think it is more a personal tolerance issue with individual drivers. Just my opinion though.

Dan

Probably right. But it is still not as good as the dedicated infrared rain sensors IMO.

 

[CorneliusXX]

Yet auto wipers still isn't great in light rain. They presumably have access to millions of events where auto wipers is overridden, but they can't train the system to be as good as old school dedicated sensors. Light rain is a fairly trivial edge case and not remotely as complicated as the complex interaction of objects in the driving environment.

Even when FSD functions seem fantastically capable it still won't be ready for release. Running over only 1 in a 1000 pedestrians in the driving path isn't nearly good enough. Waymo must be much better than 1 in 1000 or they would not have removed the safety driver.

Do you have any evidence light rain is trivial. This goes against what the company is saying.

 

[Dan Detweiler]

Waymo publishes disengagement reports (tesla doesn't and that increases my uncertainty for their FSD value) and waymo is crushing it. If robotaxis in cities are as valuable as some here theorize then the cost of mapping and the lidar system aren't show stoppers.

You say you disagree on the current state of affairs but only discuss a future capability. As of today Tesla has revealed plans to have supervised city driving. Waymo has driverless city driving in Arizona. (They imply unsupervised but it is possible there is remote supervision)

I stand corrected...somewhat. Apparently they are offering fully driverless cars now.

Waymo tells riders that "completely driverless" vehicles are on the way

I am willing to give them credit on that front, however, you have to be "matched" to a specific route. So, while making progress it still is in very limited areas on very specific routes, even within the metro Phoenix area. Not really a map to full autonomy in my book, at least not in the foreseeable future. As I said, time will tell.

Dan

 

[MP3Mike]

Waymo publishes disengagement reports (tesla doesn't and that increases my uncertainty for their FSD value) and waymo is crushing it.

Of course the problem with the reports is that they are meaningless. There is no standard rule for what counts as a disengagement so you can't compare between companies. And if you know it won't be able to handle a situation you can disengage before it gets there and that doesn't have to be reported.