I'd like to think this is Elon sandbagging. I cannot believe they would invest in creating what they called the worlds most powerful supercomputer (or something like that) at AI day and then not go all-in on it once it's running on full steam.
Perhaps, but Elon has said that he tries to be very literal so people don't need to read between the lines. I have no doubt that they're going all in, but that doesn't mean it's guaranteed to succeed. In any case Dojo's advantage over GPU at this point does not seem to be a sure thing - because Elon literally said that.
The premise that less expensive equals lower margin is not in evidence in Tesla's case. Model Y is more profitable than is Model 3 because Model Y is less expensive to manufacture than is Model 3, as has been repeatedly said by several Tesla people including Elon. Exacerbating the situation is that Model y has higher ASP.
In the hypothetical case of the 'now debunked by Elon' small car, the new technologies and the Tesla Bot (now named Optimus) could combine with, say, BYD Blade structural packs to allow very high margins, even more so with Performance models.
Elon says this is not being planned. FWIW, the previously strongly implied 12 passenger van for the Boring Company etc has not even been discussed recently.
HOWEVER: There are design teams and design facilities being put in place in Shanghai and Grüneheide. What will they do if not new models.
In my very humble option there still will be new Models in both places. They just will follow Semi, Cybertruck and Roadster so will probably happen in late 2023 or 2024. Elon did say what we all know is true: There is no point in introducing new models when they are severely constrained by chips already. They did NOT say that a major reason for slow rollout of Model X refresh was lack of chips; what is the most chip-intensive model in Tesla's current offerings?
We should all be well aware that a major contribution to the TSLA 2021 results was optimization of chip deployment to the highest margin vehicles. For reference just check the wait times for orders during the last year. Faster deliveries were for the highest margin vehicles, with greatest emphasis on those that were easiest to deploy. Entirely in retrospect, Model Y has had inordinate attention and Model S Plaid has diffused that concentration to some degree.
It is no accident that both Berlin and Austin are producing Model Y, that the scarce 4680's are going to Austin until other 4680 producers come on line.
As we already know CATL, LG and Panasonic are all building 4680 lines to complement the Tesla ones in Kato Road, Shanghai, Austin and Grüneheide. Interestingly they said on the call that 4680's would NOT be LFP. Therefore CATL is nearly certain to produce both cell to pack LFP and also 4680s' in anther chemistry. Elon said he expected to have chip problems resolved in 2022 to early 2023 after which they would be cell constrained;
Thus we can infer that 2023 will have not only huge automotive growth but also truly massive TE growth, that also was explicit.
Now our problem is to quantify all that 4680 supply for ~2023. Here is what we seem to know plus a couple informed guesses:
CATL 80 gWh Adjacent to Tesla Shanghai (already producing and delivering CATL Cell to Pack LFP, appear to be planning for 4680 structural Cell to
Pack in that factory. My unconfirmed speculation is that they'll expand to produce the 4680's while also continuing the LFP.
Panasonic 10 gWh As announced to be built in Japan for 2023 production of ,10 gWh. If they cannot commit to more than that it will be surprising. I call it 10
LG 10 gWh As announced but...newly public they'll really want more of Tesla so I think this will grow before opening.
Kato Road. 10 gWh. Present stated capacity. Of course that will rise by 2023.
GF Berlin. ~ 50 gWh They are building a large footprint there with multi-story so we can expect very compact fabrication. My guess.
GF Austin. ~ 50 gWh
2023 total ~210gWh
Obviously this is mostly guesswork, but I think there is good justification for the assumptions. Each of these could very well rise substantially.
IN 2023 the vehicle demand for 4680's alone could be as much as 2,000,000 which at 80kWh per vehicles would consume 160 gWh alone. Of course Models S and X, Roadster, Cybertruck and Semi will use a multiple of that, and probably a majority of vehicles produced then will use LFP as will most TE products. Thus that calculation is only to illustrate the scale being planned.
My personal view is that the combination of Optimus, 4680, Gigapresses and new factories are about to combine to deliver a virtuous cycle unlike we have seen. The Elon perpetual bleating about Robotaxi may or may not happen. In any event the Optimus solution uses the same technology, but within carefully designed parameters and with known and limited edge cases.
We have been offered a clear statement of the near future, obscured by the distraction of FSD.
If I have suddenly missed amaro impediments please advise.
2023 logical possibility for 4680 use:
Cybertruck. 150 kWh x 200,000. 30 gWh
Semi 400 kWh x 15,000. 6 gWh
Model Y Lr+P. 90 kWh x 250,000. 22.5 gWh
Roadster. 250 kWh x 5,000. 1.2. gWh
Models S & X 100 kWh x. 20,000. 2 gWh
Total 61.7 gWh
That uses less than a third of the expected 2023 available production. First, the 4680 supply will still be ramping so there probably will only be around half that installed capacity actually available in 2023. Second, those production and kWh estimates may well be low. Third, and most important, these are all speculations so who actually knows? Specifically it si plausible to use more cell supply to produce much longer ranges than have been seen thus. That is an obvious advantage in mountains, incline weather, long haul trucking and...Germany!
I find it unlikely that the margin of a $25k car will produce the same margins as a $60k MY, especially when margins for MY are even better than M3. Are you saying that a $25k model can use fewer than half the chips (especially the supply constrained ones), while having less than half the production cost as a MY in the 2022 / 2023 timeframe?
