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

[M|S|M MYP]

Wonder how much this dude is down now:

https://twitter.com/x/status/1808071689088954737

 

[Knightshade]

Stop trying to make me say things I didn't say sir. My point (not yours) is that these are inventory vehicles that were produced but not yet shipped.

I'm not dude.

It's just you keep saying untrue things.


My post is here:

what basis do you have for claiming those missing cars cost sales?

Your reply to my post- directly quoting said post is here:

South Shanghai Port. Specifically M3 Performance. This was posted days after Q1 P&D.

This is what you said- and it's factually untrue.

Those were NOT lost sales because none of those cars were for sale during Q1.

Production had JUST BEGIN and weren't even orderable until almost a month into Q2

There are zero lost Q1 sales there.


Moreover- the context you cut off in your original reply was about the " inventory being unexpectedly stuck on ships that got rerouted around the Cape of Good Hope" in Q1

Which ALSO aren't the M3Ps at that port.

 

[scaesare]

The C code and the NNs run on different chips though. There's SOME shared resources, but not nearly as much as it sounds from throwing that 300k lines of code bit out there.

Different parts of the same chip. The original FSD chip had 8 ARM cores right next to the NN accelerator. I assume the latest chip uses a similar design.

You are conflating "chip" with 'board'

The resources "freed up" by not using 300k lines of C code are mostly cycles on the ARM cores.

The end to end NNs run on the NN accelerators.

Again there's probably some "savings" from reduced traffic between them, and some memory savings, but it's not like the NNs themselves can suddenly run the stuff they were running on the NN chips on those ARM chips instead.



Given Tesla had ample inventory in the places those cars were going- what basis do you have for claiming those missing cars cost sales?

Can you cite any specific places there's any reports of cars not being available from inventory but there being expected deliveries to customers? I'm entirely open to the idea this is true but I've seen nothing in the numbers to actually support that, and plenty that contradicts it.

No, he wasn't. The ARM cores and the NN procesing units are on the same chip not board:

 

[EQC_]

The C code and the NNs run on different chips though. There's SOME shared resources, but not nearly as much as it sounds from throwing that 300k lines of code bit out there.

Understood. I know I don't know this stuff .

Follow up question -- I recall that Hardware 3 has 2 neural net chips. Does it also have two of the chips intended for C-code? And is each C- chip coupled/paired/partnered with one of the neural chips?

Depending on the answer to that...I wonder IF having one of the C-code chips bogged down with more than it was intended to do, such that other C-code processing had to spill over onto the other C-code chip...is there any sort of coupling that would then result in the neural chips either being under-utilized or work inefficiently? Or force both neural chips to be running different code even if, for example, neither ends up above 50% utilization?

So then, my likely-wrong thinking is that by removing the 300,000 lines of C++ code...and replacing it with more efficient neural processing...I could vaguely imagine that you could get down to only needing one of the C-chips to run everything C-code related, and then maybe all of the neural processing could be optimized to run on just one neural chip.

It wasn't dead- they still delivered 387k cars--- but it was factually lower than in the past, even for historically weak Q1.

Agreed.

Only because the massively cut production YoY. By nearly 70,000 cars.

And STILL ended up with sales also down YoY for the quarter.

Agreed.

Last quarter was terrible though.

This one is "only" down a little under 5% YoY (which is the more correct way to compare given the seasonal nature of car sales)

Agreed....although I'll point out that, depending on the media source, the readers shouldn't be expected to know about the seasonal nature of car sales...which can result in misleading headlines.

They really do- and you're engaging in exactly the same misrepresentative hyperbole you appear to be mad at others for.

Because again that's the most fair and accurate way to compare an industry that is seasonal in sales.

If you think QoQ is better then Q1 was really awful. Can't have it both ways.

You SHOULD ignore them-- because Tesla clearly had tons of inventory cars available and none of those things hurt sales.

Lack of higher demand hurt sales.

As proven by THIS quarter where they had to cut production almost 70,000 cars YoY to manage to show inventory drawdown and still had lower sales YoY despite record low promo rates on financing.



Again if you want unmitigated great news- it's over there under Tesla Energy.

The best spin you can put on the car business right now is "not as bad as feared and Tesla is being increasingly nimble in mitigating the situation"

Whoa whoa whoa...please don't read into or misinterpret what I said. I may be more optimistic than you (that's not supposed to be an insult), but I'm fact-driven as I believe you are.

If you read just my words, and not assume anything beyond that, I was really just complaining about the way news is reported misleadingly.

The example that got me today is THIS specific headline from Jalopnik (and it sounds like others have noticed similar headlines elsewhere):

Sales clearly dropped in Q1 2024. Obviously. That was heavily reported and brought up repeatedly over the past few months by the writers on Jalopnik to paint a picture of doom. Others here probably saw similar on other media sources. So, Q1's sales drop is what is in the readers' minds. To most readers (not thinking about seasonality), "continues to drop" would imply that sales are lower than Q1...but they are not. So, the headline is false based on the way MOST readers on a this type of site will see it.

Similarly, with seasonality as you mentioned, if my quick googling is accurate:
-- Q1 2024 deliveries were down 8.5% relative to Q1 2023.
-- Q2 2024 deliveries are down 4.7% relative to Q2 2023.

So, while it is TRUE that seasonal sales "continue to be down" year on year, that percentage deficit has shrunk. There has been some recovery, so it is NOT truthful to say that sales "continue to drop." The words "down" and "drop" have differnent meanings, and the authors made their choice. Among the minority of readers of a car blog or a main stream news site that are thinking about seasonality, seing the phrase "continues to drop" would probably have them thinking that Q2's percent deficit was even worse than Q1's...and that is not true.

So, to conclude:

I know last quarter was bad. We agree on that.

I suspect Q1's numbers were made somewhat (maybe very slightly) worse due to the shipping and production disruptions. We may disagree on that...but I suspect that there may have been a few extra custom-order cars that couldn't get to their owners before the end of the quarter because they were either stuck on a ship taking the long way around Africa, or because the production disruption at Berlin caused a misalignment of when some custom-order cars could be produced vs. shipped and delivered to their orderer before the end of the quarter. Yes, inventory went up by a huge number...but SOME (even just a few) of those may have just been cars that couldn't get to the right place at the right time and delivered before quarter end because of the unique issues.

Q2, for car deliveries, is worse than Q2 2023...but is better than Q1 2024. I think we agree on that?

I'll admit to being over-imaginative and hyperbolic in illustrating how biased *some* media sources are, I'll agree with you on that too.

BTW someone else mentioned it earlier but seems worth calling out again--- S/X sales appear to have been terrible....

21,551 total deliveries for "other" and most seemed to be guessing CT was around 15k.... meaning only ~6500 S/X for the entire quarter? What's going on there?

I get it's a tiny part of the business anymore, but was historically a nice margin pad...

Interesting. I think I would assign a smaller number for Cybertruck deliveries in Q2...but even if I go all the way down to like 8,000 cybertrucks delivered (which seems WAY too low), then the S/X sales are still much smaller than expected.

 

[Knightshade]

No, he wasn't. The ARM cores and the NN procesing units are on the same chip not board:

View attachment 1061618

View attachment 1061619

My apologies for not being specific enough.

Change it to "the ARM cores on the SoC are not where the NNs are running so freeing up space by reducing C code doesn't magically create a bunch more processing power the NNs (which run on the TPU cores on the SoC) can use."

As I suggested there's probably some overall savings (power budget for example, or some memory- though IIRC the NPUs have some fast sRam to buffer hits on going to slower main memory) but it's not like if they've reduced the ARM workload say 30% they suddenly have 30% more NN performance available.

So, to conclude:

I know last quarter was bad. We agree on that.

I suspect Q1's numbers were made somewhat (maybe very slightly) worse due to the shipping and production disruptions. We may disagree on that...but I suspect that there may have been a few extra custom-order cars that couldn't get to their owners before the end of the quarter because they were either stuck on a ship taking the long way around Africa, or because the production disruption at Berlin caused a misalignment of when some custom-order cars could be produced vs. shipped and delivered to their orderer before the end of the quarter. Yes, inventory went up by a huge number...but SOME (even just a few) of those may have just been cars that couldn't get to the right place at the right time and delivered before quarter end because of the unique issues.

Q2, for car deliveries, is worse than Q2 2023...but is better than Q1 2024. I think we agree on that?

I'll admit to being over-imaginative and hyperbolic in illustrating how biased *some* media sources are, I'll agree with you on that too.

Other than I still can't find (and nobody appears to have- Muskash especially) any evidence of, the shipping delays costing even a single sale- yeah I think we agree on this stuff.

Interesting. I think I would assign a smaller number for Cybertruck deliveries in Q2...but even if I go all the way down to like 8,000 cybertrucks delivered (which seems WAY too low), then the S/X sales are still much smaller than expected.

Yeah, not sure what's up with that... I do wonder if maybe the new 3P cost a few regular S sales since (the US version anyway) is running 10s now? Still doesn't seem like it explains the whole drop though.

 

[Musskiah]

Short interest increased about 4M shares (not including dark pools) in early June. Good setup for the upcoming squeeze.

 

[Todd Burch]

Accountant as a profession here. That's not how it works.
Simplified, gross profit = delivered units x (ASP - COGS).
Delivered units is known. 443k is a good number, but not a surprise anymore.
ASP and COGS have little room for surprise
So where's the massive profit is coming from?

Possibly low COGS. We've been seeing cost of goods dropping the last few quarters, and Tesla has mentioned it themselves.

Regardless, I'm not expecting anything too magical from the automotive side. Energy will be the interesting one this quarter.

 

[Mengy]

Your comment adds nothing to this investment thread so it got a disagree. If you insist on a reply, TSLA being at 230 doesn’t change the fact that Elon should be spending more of his time at Tesla, laying off the entire SuperCharger team was a mistake, and Tesla is dropping the ball not announcing/breaking ground on more Gigafactories/Megafactories.

"In your opinion", you forgot to add that. Not all of us agree with you on these things.

 

[Todd Burch]

I seem to recall that Tesla dropped Megapack prices sometime in the last 3-9 months. Does anyone recall when that was? Given the backlog of energy products, I wonder how that works. Do price drops affect a customer that's placed a large grid storage order years earlier? Or are they locked in at the earlier price?

Just trying to get an idea of how Megapack revenue will look for Q2.

 

[scaesare]

My apologies for not being specific enough.

Change it to "the ARM cores on the SoC are not where the NNs are running so freeing up space by reducing C code doesn't magically create a bunch more processing power the NNs (which run on the TPU cores on the SoC) can use."

I don't believe he was arguing that regarding C-code execution location.

He was merely pointing out those processing units are on the same chip die, not different chips, as you stated. You then doubled down by stating he was conflating "chip" with "board"... hence my providing the die layout clearly showing the processing units are indeed on the same chip as @Doggydogworld correctly asserted.

 

[Knightshade]

I don't believe he was arguing that regarding C-code execution location.

Then what do you think this meant, which was the original post that spawned this discussion?

now that they have cut out the 300,000 lines of C++ code, the hardware is now able to be used more efficiently and make the best use of its capabilities.

 

[M|S|M MYP]

As I’ve been monitoring my CNBC app more than usual this morning, I noticed a new article headlined:

Tesla shares rise on better-than-expected Q2 deliveries report​

When I read it I realized it was basically the same article I had read earlier by Lora Kolodny but I think the headline had previously featured the huge YOY drop in deliveries. The article no longer showed the poor lonely Tesla sitting on the lot anxiously awaiting a buyer, but a picture of a bustling scene of numerous Teslas loaded on trucks and awaiting delivery.

I hope Lora doesn’t get too upset about some editor changing her story. It would be such a loss for CNBC if she quit.

Eager intern must have hit the publish button before getting approval.

 

[scaesare]

Then what do you think this meant, which was the original post that spawned this discussion?

You are now quoting a different person. As I posted above, you attempted to correct @Doggydogworld, as this is the post you replied to when you tried to claim he was conflating "chip" with "board"

Different parts of the same chip. The original FSD chip had 8 ARM cores right next to the NN accelerator. I assume the latest chip uses a similar design.

No need to drag this out. You re welcome to debate with @EQC_ the merits of eliminating 300K lines of C code, and on what cores it was running.

@Doggydogworld and I are pointing out you were incorrect in where those cores were physically housed. There's no real debate here. I don't plan on responding further.

 

[FreqFlyer]

The whole purpose of this forum.............And the feeling is good.

Last | 2:51 PM EDT
229.25

+19.38 (+9.24%)

 

[Knightshade]

You are now quoting a different person.

Yes- the one who originally starting the topic and to whom I first replied to point out why removing the 300k lines of code doesn't free up any equivalent headroom because that code and the NN code are running on different cores.

You seem to be ignoring the actual thing originally raised in favor of continuing to focus on the difference between a core on an SoC and a standalone chip (a difference which is significant in many ways- but not at all to the fact you don't run the same type of code on an ARM core as on an NPU-- the actual topic)

This despite my already apologizing for my poor phrasing about that distinction and correcting it in a later post.


Point being we got rid of 300k lines of C does free up some power budget and maybe some memory bandwidth, but it doesn't suddenly give you a lot more core processing room for doing NN stuff.

Most of the benefit here I think you'd see in something like the design of HW5, which would presumably dedicate more of the SoC to NN related cores than traditional code ones...and I think Elon has hinted at that with the naming (AI5 rather than HW5)

 

[Featsbeyond50]

Your comment adds nothing to this investment thread so it got a disagree. If you insist on a reply, TSLA being at 230 doesn’t change the fact that Elon should be spending more of his time at Tesla, laying off the entire SuperCharger team was a mistake, and Tesla is dropping the ball not announcing/breaking ground on more Gigafactories/Megafactories.

Ah ha! Ross Gerber identified!

 

[Musskiah]

@Knightshade

I mean this in a genuinely sincere complimentary way: Did you compete on a Debate team at any point in your life?

 

[Robertj]

Accountant as a profession here. That's not how it works.
Simplified, gross profit = delivered units x (ASP - COGS).
Delivered units is known. 443k is a good number, but not a surprise anymore.
ASP and COGS have little room for surprise
So where's the massive profit is coming from?

Accountant professional here . Large jump in profits due to
1 Increase in GP $ from increased sales numbers with large fixed costs already covered
2 Large increase in energy deployed , 20 % plus margins , again with fixed costs already covered , sales up 130 %

 

[M|S|M MYP]

Too much positive news for Tesla today, someone had to do something about it.

'I'm On The List,' Says Elon Musk And Reveals Two Homicidal Maniacs Tried Killing Him And A Bunch Of Other People

At a recent Tesla shareholder meeting, Elon Musk, the CEO of Tesla, shared that two homicidal maniacs threatened to kill him and others in the past seven months. During the meeting, a shareholder asked Musk what he was doing to stay healthy and safe. Musk admitted he might need to start working...

finance.yahoo.com

 

[alexxs88]

I don't see how margins get squeezed much on the energy side. Demand is still quasi-infinite and battery cell prices will continue to fall.

Storage is even more of a commodity than cars. Tesla is just packaging cells with basically no moat.