Tesla Weekly (#3): The Cult Melts Down Over FSD
The path to "unsupervised" FSD likely means that the current hardware is obsolete.
Trading Thoughts
After rising 3.3% to $223 through Wednesday, Tesla sold off yesterday on higher-than-average volumes and dropped by 5.7% to $210 versus the Nasdaq 100’s decline of 1.7%.
What’s interesting is that this 5.7% drop in Tesla’s stock came amid call option volumes of 1.081 million, which was up 61% from the previous day. Put volumes were 0.68 million or 48% more than the previous day.
Overall call buying through Thursday this week was slightly down by 0.2% WoW and put volumes were down by 6.8% WoW. A pattern I’ve noticed is that Tesla’s options trading volume rises on Thursdays and Fridays.
As mentioned in last week’s Tesla Weekly (here), I’ve done well on my Tesla shorts this year because of being disciplined, i.e. only shorting when the stock is overbought at an RSI of 70 and covering when it is oversold at an RSI of 30. This week again, I was undisciplined and added the 9/27 puts at 200 strike. After yesterday, I’m breaking even on these puts.
I’m still not at my 100% “fully loaded” short position pre-Q3 earnings results on October 18th, but I plan to be there the day before the October 10th “Robotaxi Day”.
These “Something Something Days” at Tesla are superb “sell-the-news” events and I expect disappointing Q3 results to provide further downside in the stock (as of now, I’m seeing $0.44 in Q3 non-GAAP EPS versus consensus of $0.61).
China Weekly Sales Are Up 3% YoY and 36% QoQ—This is Not Positive for Q3 Profits
While Tesla bulls are excited about the strong China sales numbers this quarter, it should be noted that Tesla’s old lineup in China only continues to sell due to its “0% for 5 years” loan program.
This began as a “limited-time offer” in April but is still around, which tells you how weak real demand is for Tesla in China. 0% is roughly 500 to 600 bps lower than local lending rates, which is as bad, if not worse, than the 1.99% loans in the US.
These aggressive loans led to Tesla Q2 deliveries rising by 10.3% QoQ, while revenues were roughly flat at 1% above Q1 results. Profit margins on local sales in China are likely at 0% at this point.
Tesla’s Shanghai factory provided around 70% of global pretax profits last year, but this was mainly due to huge exports to Europe (22% of output), where the Model 3 and Y are roughly 44% higher than local prices in China.
Total exports from Tesla Shanghai were down 15% YoY in July and 18% YTD. This is a huge hit to profits given that domestic prices for Teslas in China are the lowest in the world (the entry-level Model 3 in China is priced at $32,491 including taxes, while the Model Y is $35,013. By contrast, the Model 3 in Germany is $47,266 and the Y is $50,047 including taxes).
The good news for Tesla this week was that the EU Commission lowered the tariff on Tesla’s imports from China to 9% from 20.9% previously. This news didn’t move the stock on Tuesday most likely because Tesla only raised the Model 3 price by 3% last month, which only covers 1/3 of the tariff hike and is effectively a price cut.
As has been the trend this year, the Model Y is being supplied by Tesla’s German factory (which has been running below 50% capacity utilization since its March 2022 start of production).
The fact that the Model 3 is still being shipped from China despite the 9% tariff (on top of 10% import duties), proves that it’s more profitable for Tesla to ship it from Shanghai, rather than Fremont.
It should also be noted that the year-on-year growth in Tesla China sales have a low hurdle as Q3 2023 saw a shutdown of the Model 3 line in Shanghai in order to retool for the “refreshed” version.
And finally, the more that Tesla’s output goes to domestic China rather than exports to higher-priced countries, the worse the profitability becomes. Figure 1 shows how high Tesla’s domestic sales as a % of output is rising since mid-2022, which is a bad sign for global country mix and overall profitability (the higher the percentage of output delivered locally, the lower overall profitability of Tesla’s Shanghai factory).
Figure 1: Tesla Domestic China Sales as a % of China Production
Source: CPCA and CAAM
This Week’s Main Topic: Tesla Hardware for FSD is Set to Become Obsolete; Epic Lawsuits Ahead
The Tesla Cult had a meltdown this week over Tesla’s floundering Full Self-Driving (FSD) driver-assist product, which Musk repeatedly says is the deciding factor between Tesla being a company worth trillions of dollars or zero. Every time Tesla releases a more advanced version of FSD, YouTube is filled with Tesla drivers barely escaping traffic accidents while using it.
The below tweet from Fred Lambert perfectly depicts this week’s struggle within the cult. In essence, it’s becoming clear that the hardware running current versions of FSD needs more computing power. The Cult realized this after a recent FSD update was only beamed to Tesla cars with Hardware 4 (HW4), which is more advanced than HW3. The problem is that most Teslas on the road only have HW3 (2.6 million worldwide and 1.6 million in the US). HW4 was released in Q2 2023.
The radical Cult members tried to defend Tesla and Musk by saying that Tesla never promised FSD on old hardware, which immediately received a Twitter “Community Notes” clarifying that Musk declared in Q3 2016 that all Teslas made as of that time would have all the hardware needed for full autonomy. It’s surprising that Tesla didn’t sell off when this news trickled out last week, as the implications are serious (main points below).
Tesla bulls were excited to point out that the latest version of FSD was released for Tesla cars running on HW3 this week. The problem is that it was available 3 weeks later than Tesla FSD drivers with HW4 and it has yet to be widely distributed to all HW3 cars.
This erupted into a war between the radical cultists (who protect any mistake by Musk or Tesla) versus the moderate cultists (who already feel duped by still not having a self-driving car despite having paid up to $15,000 for FSD years ago).
Figure 2 below shows a comment from an FSD user who’s sick and tired of getting “Musked”. It’s a glimpse of how many other HW3 owners may feel the same.
It is becoming clear that FSD will need more computing power on its way to full autonomy. The fact that the latest version of FSD had to be tinkered with for 3 weeks before it could be released for Teslas with HW3 is proof of that. What happens to HW3 when HW5 is released?
Figure 3 shows Musk’s claims from January 2023 that HW3 would be too expensive to retrofit. In 2022, Tesla paid $1,000 to upgrade any HW2 or HW2.5 owners to HW3.
Figure 4 shows a stunning legal precedent for Tesla owners with FSD running on HW3. It shows that a US court ruled in favor of a Tesla owner in Washington state to get upgraded to HW3 for free.
Tesla’s HW3 cars were sold from April 2019 to October 2023, when HW4 was released.
Based on TroyTeslike’s FSD take-rate estimates (which he stopped publishing after Q3 2022, from which point I use my own estimates), there were roughly 440,915 Tesla cars sold with FSD (mostly in North America, as FSD is not street-legal in any other countries).
Based on the prices per quarter ranging from $6,000 to $15,000, Tesla would have to refund its customers with roughly $4.7 billion (43% of 2023’s non-GAAP net profit and 7% of last quarter’s Shareholders Equity).
If anyone who bought a Tesla (even if they didn’t buy FSD) could lay claims in such a lawsuit, the number of cars sold between April 2019 and October 2023 globally is 1.68 million vehicles.
There are currently over 90 lawsuits in the US related to Tesla’s Autopilot and FSD. This HW3 problem is about to add fuel to the fire.
Figure 2: Tesla FSD Owner Who’s Had Enough
Source: Electrek (link here)
Figure 3: Musk Admits in January 2023 That HW3 is Not as Good as HW4
Source: Tesla Q4 2022 earnings call on January 26, 2023
Figure 4: Legal Precedence for Lawsuits from HW3 Owners
Source: Wikipedia (link here)
Lucid is This Week’s Top Performer; Rivian is the Worst
Lucid Motor rose by 22% this week (through Thursday) on very high volumes in the last two days. My gut feeling is that there is continued short-covering going on (Lucid rose 7.5% on high volumes and no news yesterday as the Nasdaq 100 fell by 1.7%).
Figure 5: Auto Stock Price Action (Sorted by Weekly Returns)
Source: Bloomberg
After Tesla, Lucid would be the next strong-conviction short on my list. Based on consensus estimates, it has the second-highest EV/Sales ratio after Tesla (EV/Sales is a key metric for loss-making companies). On my estimates, Tesla has the highest EV/Sales ratio, but the reason I’m not short Lucid is twofold:
Short interest is high: If all the shorts in Lucid were to cover at once, it would require 7.1 days of average daily trading volume for shorts to fully cover. My rule of thumb is to avoid shorting any stock that has more than 4 to 5 days of short interest. Also, because the Saudi Arabian sovereign wealth fund (PIF) owns over 60% of Lucid, the free float is only around 38%. This is why I stopped covering Lucid after my last report in February 2023 (here). If there ever was a need for an “Avoid” stock rating, it’s here for Lucid now.
The Saudis Will Likely Take Lucid Private: While I believe that Lucid has the top powertrain technology in the EV industry, its cars are too expensive for the current market, where Toyota and Honda can’t make enough Corollas and Civics due to lower and lower affordability among car buyers. Given that the Saudis already own over 60% of Lucid, it wouldn’t take much to take it private when the real cash crunch comes.
Rivian is the worst-performing auto stock this week, down 3% through yesterday. There was no specific news aside from its production chief jumping ship for Chrysler’s Jeep division. To me, it simply seems like Rivian is slowly coming back to earth after having spiked by 51% since June 25 when VW reported that they were taking a $5 billion stake in the US EV start-up.
A Peak at the Tesla Bull’s View
Figure 6 shows a list of catalysts for Tesla tweeted by a Tesla bull this morning. In yellow shade, I highlighted the points which I feel are highly unlikely. They go as follows:
EPS can’t bottom unless Tesla unveils another “homerun” hit model that can offset the decline of the Model Y, which makes up 67% of global sales and 62% of installed capacity worldwide.
FSD take rates will not rise, given all the gripes outlined above. Why didn’t Musk disclose the take rate improvement after the free trial for all users in May? Because they were likely horrible.
Tesla has put the $25,000 “Model 2” on the back burner. The only cheap models coming out next year will be “tweaked” Model 3s & Ys.
Tesla’s Megapack business is growing, but profitability is actually down to 25% gross margins versus a high of 29% in 1H 2017. While this segment might grow, it’s highly doubtful that marginal profits will increase: Tesla is simply sourcing battery cells and packaging them (cell production is 85% of the work, while packaging is only 15%. This is a low value-added business.
Robotaxis next year is highly unlikely given how buggy Tesla’s FSD is these days: it can only drive around 133 miles between critical disengagements versus nearly 20,000 miles at Waymo and 26,000 miles at Zoox.
Tesla’s humanoid bot, Optimus, is a pipedream. Boston Dynamics is way ahead. Tesla can barely make cars, let alone humanoid robots.
And finally, Judge McCormick’s ruling on the Plaintiff’s comp package which caused so much controversy among the lemmings, is more likely ruled upon shortly after the US elections, not 2025.
Figure 6: A Tesla Bull’s Outlook on the Stock
Source: Twitter
Figure 7 shows Tesa’s valuations and returns based on 2025 estimates. It makes no sense, but this Tesla behemoth continues to lurk around at its mammoth valuation, despite deteriorating fundamentals.
Figure 8 shows how Tesla—which has no discernable AI—trades at higher valuations than leading AI stocks like Microsoft, Meta, Google, etc.
Figure 7: Tesla’s Valuation & Returns vs Auto Rivals (Ranked by 2025 Pre-Tax Margins)
Source: MH research & Bloomberg
Figure 8: Tesla’s Valuations & Returns vs The Magnificent 6 (Ranked by Cheapest PEG Ratio)
Source: MH research & Bloomberg
Does HW4 include lidars and radars?
No amount of computer power will make a computer vision only solution miraculously work.
random thoughts if interested:
(if you've not seen the article on Seeking Alpha entitled "Options Trading Drives Tesla's Stock Price", cannot recommend it enough)
this year we've seen 3 discrete massive moves in the stock: 1) late Apr after earnings, 2) week 1 in July, and 3) Aug 6-10 (or thereabouts).
all 3 have the hallmarks of a gamma squeeze. and when you do some back of the envelope math (in a sec), it seems *very* likely gamma squeezes are pushing this thing up.
* current open interest in tesla options is ~ 7-8MM. notionally, thats 700-800MM shs
* while puts and calls would cancel out if the P/C ratio were 1, with Tsla its .6 P/C ( a notable distinction as most stocks its > 1)---which is to say the option market is structurally long calls.
(700M* .4 = 280MM shs notional length via calls in the option market)
* delta adjust the aforementioned 280M by a delta factor (lets be conservative as deferred options have deltas congregating ~ .5), so lets just .4, or 112MM shs delta adjusted length via the option market. ---not a big # relative to the float of 3.5B, but its a damn big # vs daily volume.
* any big buy move in near-to-expiry ATM calls (where gamma is highest) will trigger the gamma squeeze. for it to work, need to see a lot of open interest in the strikes ATM, and ideally not a lot of stock liquidity (ie holiday week).
of course, we dont know what a market maker has on their books, but we do know they sell volatility and delta hedge and given the p/c ratio biased towards calls, they're a structural buyer of tsla if the market is structurally long calls---aggravating every price move higher, but happy to see (and help) it oscillate so they can bank the vol and OTM options expire worthless.
what does the above mean?
1) there's 2 markets for tesla: one for the stock (for which fundamentals SUCK as brad makes clear), and one for the option market, for which the only fundamentals that matter are A) volatility and B) liquidity. (i see nothing short of a complete wipeout in tesla changing this)
2) the 2nd market, the option market, can and does drive tesla's share price at times (and at times like the Q1 earnings squeeze, i'd bet elons family office was behind it). this means there's an ever-present risk of a gamma squeeze for shorts. --makes position sizing critical.
3) volatility cuts both ways. it will go up and down so long as A) fundamentals suck and B) the open interest remains massive with structural bids on the call side to delta hedge when the tail (options market) wags the dog (stock market). ---this is the perfect formula for volatility: Bears (fundys) vs Bulls (net call length with liquidity & volatility to monetize)
knowing this, i'd say be careful around event risk and weeks with high open interest & low aggregate market activity (like a holiday week ala 1st week in july) for subsequent gamma squeezes. last week had just < 0.5MM OI for weekly DTEs (vs 1+MM for first week of july), and it still got squeezed higher on Fri after jackson hole---and i'd note the market makers pinned the close at 220, just as one would expect if they wanted to wipe those long that strike.