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

[TheTalkingMule]

If the robot is for Mars, likely no fluids in it.

Yet.

 

[Krugerrand]

Loved the first part of the fsd/dojo presentation!
It increased my confidence that Tesla is on the right path making FSD happening!

I'm sorry... but the spandex guy dancing, humanoid idea... kinda freaked me out. Imho should've waited another year before revealing this until there is an actual prototype ready! Now that part had a big vaporware feeling for me. I know I'll get many dislikes for it... But that's my honest opinion.
I get the idea that Tesla is making the building blocks and infrastructure for the software that eventually will be applied to Optimus.
Elon is like: "hey, I have just invented the transistor. I can make/sell a Playstation 5 out of it" He's not wrong, but the timeframe is messed up.
At this stage, is it really necessary for Tesla to announce products that'll take years again before they ship anything to customers?
(Roadster 2, Semi announced in 2017. Solar roof 2016.... Cybertruck 2019... I get the argument that Tesla needed access to capital from investors, hence the look at future portfolio). I whish time between announcement and shipment of products was shorter.... But Optimus really enlarged that feeling


Maybe it's because I was watching this from Belgium and being tired... need to sleep on it... let it sink in, digest it further. Maybe tomorrow after reading other opinions, different views,..., I'll have a different opinion.

Yes a disagree from me. The TeslaBot is simply the natural progression. Let me give just one example;

GigaNevada where warm bodies are in short supply because not enough talented people want to live in the middle of nowhere.

Enter TeslaBots to work the lines and to handle logistics inside the factory (this is the natural progression of FSD). No more labor issues or housing issues at GigaNevada, free to grow to full dreamed size.

I think where people are getting stuck is imagining a humanoid robot in every home ala some sci-fi movie. Slow your roll. Oh, that’s totally happening, but you can breathe easy knowing that many of us here will be dead when that happens. It’s going to be your children and grandchildren who’ll need to worry about the robots taking over. If you’re not old like many of us - you’ve at least got some time to get used to the idea and to buy more TSLA.

Other applications:

Robot mining
Robot firefighting
Robot medical staff to help during pandemics
Robot in home care for the elderly
Robot anything you can imagine
Robot surgeons so there isn’t a 6 month wait list for elective surgeries
Robot jail guards
Robocops (yes, I went there)
Robot teachers
Robot dog walkers

I’m sad I’m not 20.

 

[CLK350]

D1 (400w TDP, 645mm^2, 7nm)
BF16: 362 Tflops
FP32: 22.6 Tflops
On Chip Bandwidth 10TBps(or 1250 Gb/s)
Off Chip bandwidth: 4TBps(or 500GB/s, 25 D1 per Tile)
Off Tile bandwidth is 36TB/s reported (I think 9TB/s is more like it for tile to tile communication), 3000 D1 chips connected together

AMD Radeon MI100 (300w TDP, 750mm^2, 7nm)
BF16: 92.3 Tflops
FP32: 23.1 Tflops
On Chip bandwidth : 1228.8 GB/s
Peak Infinity Fabric™ Link Bandwidth 92 GB/s (offchip)

Nvidia A100 (400w TDP, 826mm^2, 7nm)
BF16: 312 Tflops
FP32: 19.5 Tflop
On chip bandwidth: 2,039 Gb/s
Off Chip bandwidth: 600GB/s (up to 12GPUs)
Off Chip bandwidth PCI4: 64Gbs

My opinion, the actual D1 chip is pretty good, but not mind blowing given the size/power usage. It's inline with Nvidia's best. But dat scalability holy S balls with them interconnect bandwidth off tile. Mind blowing...

The fact that Tesla DID come out with a chip slightly surpassing the main players, AMD and Nvidia in a few years, PLUS the software stack for easy programming is mind blowing.
PS. Added correction from Singuy a few pages later " I would like to correct that the TBps from Tesla's presentation is actually Terabyte and not terabits, which make the bandwidth from on chip 5x faster, and off chip 7x faster than Nvidia. "

For a specific example: to get, say, a TeslaBot grape picker all you need to do is have actual grape pickers do their job with cameras and glove assist, and set up specialists at central base for the initial NN labeling.
High end organic grape growers use flocks of sheep to take care of weeds, instead of herbicides. TeslaBot could do that as well, same as tending the grapes as they grow (watching for leaf damage and growth, testing and measuring soil humidity, turning on /off watering).

If TeslaBot were the first big far out announcement, it wouldn't be as credible as it is. To put this in perspective: we've had the Model S, the Model 3 ramp, Giga Nevada, Giga Shanghai, AI day and Battery day - these have all become realities (well 4680 and Lithium mining are still at 90% and 10% only, Giga Casting is in production ). Not to mention the other EMSK (Elon Musk) initiatives, The Boring Co and LVCC, SpaceX and the multiple rocket programs, Starlink, Neuralink..

The dancing TeslaBot interlude will/ has already drawn some flack, but I think it was a very carefully planned demo, showing how dead serious Elon is at developing that new face of Tesla, the AI/ robotics one. The recruiting may not be immediate, or known immediately, but recall how as recounted in Liftoff the recent book by Eric Berger on the early days of SpaceX, Elon was intent on hiring the very best of the best.
I am sure that Elon/ Tesla has identified the top layer of the AI robotics talent, and they have received an invitation. Whether they attended or not, they certainly got the message, and the real issue is how many of these will Tesla eventually get on board. Regardless, they have enough as it is to already sell their tech, it is just that they figured, why leave 80% of the potential profits off the table?

From the investor point of view this is also huge, because so far and for the near future, Tesla's car production and so profits are battery constrained - since TeslaBot will cleary use much less batteries per unit (think car vs bot as 4,000 lb vs 150 lb so with a 1:27 ratio, 27 bots @ 20K ea, say that's 540K vs a 55K car) , so this is hugely positive.
On a side note, this is also the first time AFAIK Elon has clearly indicated they would be open to licensing the FSD tech to other carmakers. My bet would be, with better pricing for pure EV's. So here, another revenue stream opening up.

What could go wrong? Unions clamoring for a TeslaBot union hahah ha .. Lawsuits against damage or injuries caused by TeslaBots ..Apple /Google and Chinese competition .. detrimental legislation .. I don't see anything credible stopping Tesla's domination for the next 7 years at least.

 

[TheTalkingMule]

Loved the first part of the fsd/dojo presentation!
It increased my confidence that Tesla is on the right path making FSD happening!

I'm sorry... but the spandex guy dancing, humanoid idea... kinda freaked me out. Imho should've waited another year before revealing this until there is an actual prototype ready! Now that part had a big vaporware feeling for me. I know I'll get many dislikes for it... But that's my honest opinion.
I get the idea that Tesla is making the building blocks and infrastructure for the software that eventually will be applied to Optimus.
Elon is like: "hey, I have just invented the transistor. I can make/sell a Playstation 5 out of it" He's not wrong, but the timeframe is messed up.
At this stage, is it really necessary for Tesla to announce products that'll take years again before they ship anything to customers?
(Roadster 2, Semi announced in 2017. Solar roof 2016.... Cybertruck 2019... I get the argument that Tesla needed access to capital from investors, hence the look at future portfolio). I whish time between announcement and shipment of products was shorter.... But Optimus really enlarged that feeling


Maybe it's because I was watching this from Belgium and being tired... need to sleep on it... let it sink in, digest it further. Maybe tomorrow after reading other opinions, different views,..., I'll have a different opinion.

This day wasn't about marketing products, it was about showing what's bring worked on for the purposes of attracting top global talent.

 

[The Accountant]

the Q posed at the event is quite valid: is there a market for very expensive robots to replace low-waged labor?

I believe the answer is yes. Even low cost employees can be costly when you add in health benefits, 401k, etc.
In addition, robots:
- can work 24/7
- don't take vacation
- don't have sick days
- will work on Holidays
- won't take off 4 months on disability
- won't sue you
- won't demand annual salary increases
- won't be unionized
- won't work from home spending all day on TMC
- will provide consistent quality
- etc

Yes - there is maintenance and investments to upgrade but I can see the cost/benefit analysis working out just as we see automation in factories today taking work away from the low wage worker.

You refer to a high cost robot....so your point is valid; the cost of the robot can't be $1m. I think if would need to be $200k or less at scale.

 

[Iain]

The fact that Tesla DID come out with a chip slightly surpassing the main players, AMD and Nvidia in a few years, PLUS the software stack for easy programming is mind blowing.

For a specific example: to get, say, a TeslaBot grape picker all you need to do is have actual grape pickers do their job with cameras and glove assist, and set up specialists at central base for the initial NN labeling.
High end organic grape growers use flocks of sheep to take care of weeds, instead of herbicides. Well, TeslaBot could do that as well, same as tending the grapes as they grow (watching for leaf damage and growth, testing and measuring soil humidity, turning on /off watering).

If TeslaBot were the first big far out announcement, it wouldn't be as credible as it is. To put this in perspective: we've had the Model S, the Model 3 ramp, Giga Nevada, Giga Shanghai, AI day and Battery day - these have all become realities (well 4680 and Lithium mining are still at 90% only, Giga Casting is in production ). Not to mention the other EMSK (Elon Musk) initiatives, The Boring Co and LVCC, SpaceX and the multiple rocket programs, Starlink, Neuralink..

From the investor point of view this is also huge, because so far and for the near future, Tesla's car production and so profits are battery constrained - since TeslaBot will cleary use much less batteries per unit (think car vs bot as 4,000 lb vs 150 lb so with a 1:27 ratio, 27 bots @ 20K ea, say that's 540K vs a 55K car) , so this is hugely positive.

What could go wrong? Unions clamoring for a TeslaBot union hahah ha .. Lawsuits against damage or injuries caused by TeslaBots ..Apple /Google and Chinese competition .. detrimental legislation .. I don't see anything credible stopping Tesla's domination for the next 7 years at least.

I don’t think Tesla will be invited to the next White House EV event with the Bot announcement.

 

[JustMe]

Has anyone checked on Warren Redlich? I’m concerned his head may have exploded when Elon suggested Tesla Bots could enable unlimited economic expansion. If he survived the night, I predict he’s working on a new TSLA forecast that will breach $1 million per share.

 

[StealthP3D]

Ok my opinions. I understood like 99% of what they said.

4. HW4 introduced with CT. Not sure if it’s so great to buy a HW3 car atm, maybe a lot of benefit to wait for CT and then it should not take too many months for the rest of the vehicles to switch.

Newer versions of FSD hardware would be different only in terms of the main FSD compute board and 8 tiny cameras. When FSD is level 5 it will be a trivial matter to replace the cameras and processor, if necessary, so I fail to see how this is even a consideration with a new car purchase. A more valid perspective would be to wonder who wouldn't order their new car with FSD to ensure against the cost rising $20K or more when it's Level 5. A complete hardware retrofit would be under $2K and would be paid for by Tesla (but only if the vehicle was purchased with FSD).

In other words, all you have to do to "futureproof" your new vehicle purchase is order it with FSD.

 

[StealthP3D]

No people... it's an AI recruitment event, so just AIs

If that's true then I'm a little disappointed in the quality of most of the questions.

 

[Drezil]

I'm unsure how he still expects HW3 to be able to do legit FSD (ie above L2) given it can't run redundantly though, that bit was semi-asked but glossed over in the answer.

Destillation.

You take the large net that does everything correctly & train a smaller net with it, that fits time/space-requirements. Then you can transfern almost all of the knowledge of the teacher to the student.
Normally you can achieve an order of magnitude improvement from this while retaining all useful information.

Also you can plug this into "neural architecture search" to get an even more efficient student. With enough compute there are no limits. And the Architectures NAS comes up with are pretty wild & outperform anything that researchers can create. BUT to calculate them you need HUGE amounts of compute of NNs... like .. say .. an exaflop ..

 

[AperiodicCoder]

More specifically, how do they do all this and still show a profit each quarter?

This highlights how many people there must be at Ford and GM and drawing a nice salary and basically contributing nothing to the value offered to the customer.

Not to mention the marketing budget everywhere else!

Meanwhile, Tesla's marketing strategy: do great things for a great cause and you'll have millions of marketers (us).

 

[lascavarian]

GLJ started off his appearance by stating he dropped off the AI day presentation. He then made harsh accusations based on his partial information. Gene M will likely be fact checking him.

GLJ must be hoping to get another appearance. Quite the scheme he has…. Selling incomplete information to the many willing disbelievers with money to fund their wishful misinformation.

 

[JusRelax]

Newer versions of FSD hardware would be different only in terms of the main FSD compute board and 8 tiny cameras. When FSD is level 5 it will be a trivial matter to replace the cameras and processor, if necessary, so I fail to see how this is even a consideration with a new car purchase. A more valid perspective would be to wonder who wouldn't order their new car with FSD to ensure against the cost rising $20K or more when it's Level 5. A complete hardware retrofit would be under $2K and would be paid for by Tesla (but only if the vehicle was purchased with FSD).

In other words, all you have to do to "futureproof" your new vehicle purchase is order it with FSD.

Not sure I agree with this -- Elon made sure to emphasize that, even with HW3 + FSD computer, he is confident that Tesla cars will be able to run FSD, at up to 300% better than a human (He said HW4 with the upgraded cameras would be like 10x better than a human). I believe he is making sure that people understand that FSD is deliverable with HW3, just at a safety rate lower than HW4. This implies to me that cars with HW3 will have to pay if they want HW4 installed.

 

[H Mak]

I didn't stay up for the presentation but have followed the comments on here with increasing excitement throughout the morning. And now I can't wait to watch it. Thanks for the contributions, particularly from people who know what they're talking about on here, from computer scientists, games developers, software engineers etc, and not forgetting people who like cats.
This really is the best forum and Tesla really is the best investment.

 

[UncaNed]

weeding a raspberry field?

Weeding, totally. We've entered a new, troubling era of weed resistance to herbicides (NYT superweeds article); robot weeding tech arrives none too soon.

Way OT (sorry, can't help myself/PSA) - Blueberries and Raspberries grow like weeds requiring negligible pesticides while strawberries require pesticides and regular soil fumigations and replanting; the ecological damage done by strawberry growers are literally orders of magnitude apart.

 

[Dancing Lemur]

Weeding, totally. We've entered a new, troubling era of weed resistance to herbicides; robot weeding tech arrives none too soon.

NYT superweeds article.

We always knew the first Tesla-branded bot would be WeedBot.

“May I spark that for you, sir?’

 

[JohnnyEnglish]

Not sure I agree with this -- Elon made sure to emphasize that, even with HW3 + FSD computer, he is confident that Tesla cars will be able to run FSD, at up to 300% better than a human (He said HW4 with the upgraded cameras would be like 10x better than a human). I believe he is making sure that people understand that FSD is deliverable with HW3, just at a safety rate lower than HW4. This implies to me that cars with HW3 will have to pay if they want HW4 installed.

Fully agree with your statement (bolded by me).

 

[BornToFly]

As expected, most Headlines in my feed today are about the robot, and not about the incredible computing power and talent that is driving FSD. In the end, everyone is going to license FSD from Tesla, because nobody will be able to match it (at best they will be ten iterations behind). TSLA is the next TSLA.

 

[GOVA]

Once people have nothing to do a "basic pleasure models" need to be introduced especially in off-word colonies. This has been mentioned and shown in many scifi works. Same might be attempted via proper Neuralink implants and software.

This is robotics.

 

[Iain]

Fully agree with your statement (bolded by me).

I'm sure that's the business model and i wouldn't expect anything else. As long as it works and the older systems are still supported with software it makes complete sense.