-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable
- Thread starter AudubonB
- Start date
Any guess's about Monday for our favorite Stock?
Besides it might go up or down
Just like their cars, everyone who wanted shares of Tesla already has them. So there will be no shares traded tomorrow.
Elon liked this tweet:Any guess's about Monday for our favorite Stock?
Besides it might go up or down
NASA JPL on Twitter
But I’m not good at astrology...
Any guess's about Monday for our favorite Stock?
Besides it might go up or down
guessing the Hong Kong violence won’t be well received in the markets here, as it has affected Asian markets as well.
Flat Mon, down 15 Tues, then sharp rise to new 365 before leveling Fri at 360.Any guess's about Monday for our favorite Stock?
Besides it might go up or down
Just sayin
Last edited:
After my years of railing against short-selling, you're telling me to go massively short on Thursday?Flat Mon, down 15 Tues, then sharp rise to new 365 before leveling Fri at 260.
Just sayin
Buckminster
Well-Known Member
Fact Checking
Well-Known Member
Yes, which reminds me of this FSD business case related comment by @Doggydogworld:
This argument misses what LIDAR economics is about: Tesla is, as @ReflexFunds has pointed it out recently, betting on FSD driving being a "very hard problem", and has shaped their FSD program accordingly:
- Tesla has built a ~800,000 cars large fleet that has the role of a "planetwide neural network training computer", which fleet will grow to ~1,500,000 cars in about a year.
- Tesla has put their own AI chip, a ~150 trillion inference calculations per second NN supercomputer into half of that fleet - which percentage will increase to ~80% by the end of next year.
- Tesla is expanding backend side training compute capacity rapidly as well: they are expanding their current ~70,000 GPU hours per day capacity to what I expect to be around 1,000,000 GPU hours per day training capacity, via another custom ASIC (Project Dojo).
- Just to put this into perspective: if we estimate computing cost at $1/GPU-hour, then Tesla is scaling up their backend to an about $350m/year supercomputer - which would be around 30% of their R&D budget - at a fraction of the cost: less than $30m/year, or around $30 per fleet car per year.
- This gives Tesla the ability to scale up their backend as their training fleet grows, with no scaling limit and very good economies of scale.
- ID.3 entry level CoGs will certainly not be below $30k anytime soon:
- the first ID.3 deliveries have now slipped to October 2020 - a year from now.
- ID.3 initial prices will be €39k, €44k and €49k.
- Only the €49k version will have essential FSD hardware such as auto steer. Even these versions won't have LIDAR, nor any FSD computing capacity worth speaking of.
- So the ~50,000 ID.3's that might be delivered to first customers starting next October will increase Waymo's and VW "FSD training fleet" size by exactly zero units.
Tesla's approach is basically to mass-manufacture the nodes in the FSD training computer: every Tesla comes with full FSD hardware even in the lowest trim, and increases the fleet size.
Note how LIDAR cost isn't a cost constraint to a functioning robotaxi, but a constraint on the speed to reach unsupervised FSD capability.
The only chance Waymo has to win the FSD race at this point is if unsupervised self-driving is an "easy problem" that doesn't require a >1,000,000 units large training fleet - but nowhere does @Doggydogworld's argument even acknowledge this fundamental assumption and constraint.
Last edited:
Nooooo.... it can't go down below 273!Flat Mon, down 15 Tues, then sharp rise to new 365 before leveling Fri at 260.
Just sayin
EVNow
Well-Known Member
It’s not even the cost. It’s the aesthetics. Who will pay $100k for a car with an ugly rotating thingy on top ?
You could either have Lidar or large training fleet. Not both. Waymo and others chose lidar and Tesla chose large training fleet.
Now that's more like it... still a small block, but better. I'm still holding out for 426!
I'm believing in the cup and handle. That's why the quick dip and on up further to a new high for the year (and many more we hope).
Fact Checking
Well-Known Member
-0.30% macro drop so far - pessimistic but not severe yet.
TSLA price action this week might also be influenced by:
- The Starlink launch NET later today: failure would weigh negatively, success positively
- China GF3 "sales permit": no permit yet, there's rumors of news being imminent, but nothing firm yet. GF3 deliveries might be key to meeting Q4 guidance.
- There's a brutal number of TSLA options expiring this Friday, with a lot of them centered around $250 - I.e. big losses to realize and big gains to protect. Half a million contracts open interest - that's 50 million TSLA shares equivalent ... If most market makers are delta neutral already then this might not mean much. If this week is mostly flat or TSLA gains then this will be a powerful counterfactual to the MM driven "max pain" thesis.
- First HW3-only feature was activated last week (traffic cone visualization). It's unclear whether the HW3 chip got activated.
- A number of analysts following Tesla haven't updated their pre-Q3 price target yet. Since "Operation 10-Q gaslighting" clearly failed, the TSLAQ aligned ones are running out of time.
- VW ID.3 first customer deliveries got pushed out even more, to October next year. Weirdly no calls for a SEC investigation and no lawsuits for breaking their "start in 2019" promise.
Last edited:
Huh?
Umm? Cars with HW3 don't have a HW2.5 chip, so the HW3 chip has been active since the day they were installed. It may be that hey are just starting to roll out native HW3 NNs.
Causalien
Prime 8 ball Oracle
I have forgotten what it feels like to have my portfolio reaching all time highs.
This realization took away all the sympathies I have of the shorts. Two years of their childish drivel wasting my time.
Isn't Mobile Eye hooking up with Nio in China to have private vehicles with FSD by 2022? A Nio with Lidar could look odd.
Meanwhile, I just read that Mobile Eye was going to introduce FSD in Israel but it wouldn't be ready for the US until 2024 (or 5?).
Meanwhile (again), an older article states that Mobile Eye was planning to deploy FSD in VWs in Israel by 2019. How's that coming? Inside VW, Mobileye’s deal to launch an autonomous ride-hailing service in Israel – TechCrunch
Seems everyone (~40 companies) just keeps pushing the goal out. It's always 2 years out for anything, and never arrives, just new funding or a new partnership. What happens when the CEO doesn't understand the technology perhaps?
Meanwhile... I just got Rev 2019.32.2.1 on AP2.5 yesterday and it seems a bit more alive and alert in the parking lots. Scoots right along. I know where my money is (because I think I publish that everyday here).
Similar threads
- Locked
- Locked
- Locked
