I read the Ark model on TSLA.
One thing one has to give them credit for is they aren’t shy about their assumptions.
After 11 yrs of car-making, they made 500k cars in 2020. The ‘bear case’ for 2025 is 5mm units. Assuming capacity grows 1mm cars a year from 2021-2025 that gets
them to their bear case of ~11mm cars on the road by end 2024 and 13mm on average in 2025.
The bear case assumes 40% of all Teslas sold will be on a human ride-hailing platform. That’s 5.4 million human-driven ride-hailing Teslas on the road in 2025.
The bull case is
of course, more aggressive at 10mm units sold in 2025, suggesting avg 22mm TSLAs on the road in 2025, 13.2mm of which are used for human-driven ride-hailing, and there is a 50% chance that 13.2mm is only 8.8mm with the *other* 13.2mm used as robotaxis.
The bull model has a weighted average 90% of Teslas sold being used for robotaxis and ride-hailing in 2025.
The model on GitHub would suggest that in the Ark bull case autonomous driving, in 2025, Teslas would drive 2 trillion miles a year for ride-hailing and robotaxis.
By the end of 2025 that would be a run-rate of 2.5 trillion miles a year.
22.5mm Teslas used as taxis driving 3.3trln km a year (4trln run rate by end-2025) suggests TSLA will have near 100% of the market globally, and average taxi usage will grow dramatically from here.
The market share model seems to ascribe US market share to rest of world (and China/Europe show that’s an iffy idea). The taxi model suggests TSLA’s sales market share outside the US will be a fair bit higher than inside the US. And given ex-US taxi market models, that seems
another iffy proposition. The *average* miles driven by a taxi with paid customers in the highest per vehicle market in the world (Shenzhen) is 60k miles/yr. Tesla’s mean estimate for it’s Monte Carlo-derived TP of $3000 is 59,952 miles/yr. The bull case is 110k miles.
The first problem I see is that in cities, globally, dead-heading (i.e. return trip or trip W/I customer) is 45-55% of total taxi miles driven. The model assumes 110k miles driven earning. Which is 7-8x US taxi avg including deadhead and twice global max.
Total robotaxi miles in the github model is only 1.43trn with gross ride-hail billings of $889bn on 13mm cars is $68,400/car. The amount which would come to the car owner is about 40% of that, or $27k/yr. That is on a $25k car according to the same model.
Presumably, there will be eBay listings for queue space for Model A (the cheap car) cars and some families will get every member to buy one, with a loan. Put down $5k, get paid $27k/yr for letting TSLA use your car. Seems easy. I wonder if TSLA will underwrite the revenue.
So TSLA will have $93bn of EBITDA in 2025. That’s the bear case with no autonomous driving. With autonomous driving and the bull case, EBITDA will be $208bn, with $168bn of free cash flow.
With all confidence, I can say... It’s a model...
• • •
Missing some Tweet in this thread? You can try to
force a refresh
I would have tried that recipe but I was in want of every single ingredient listed except for a slab of rump steak, an onion, and celery so my own version this even used balsamico + honey instead of treacle. The sauce was the sous vide juices, with a butter/wine reduction (syrah)
Carson Block of @muddywatersre writing in the FT lays the blame of stonk gyrations like GameStop (sub $20 on 12 Jan, up 18-fold in 10 trading days) squarely on low rates and passive investing.
This is Hogwash, Blatherskite, Buncombe, and Taradiddle.
That fall into the close on Tesla is classic ‘big name inclusion trade unwind’ pattern - people say “let’s not leave it all for the close, just in case” and put their ‘sell ahead of the close vs sell at the close’ ratio higher than the buyers do.
Now I’ve jinxed that. 🤓
Closing cross on S&P funding trade was about 8-9% larger than the S&P Global model would have suggested (I had it a trifle larger than even that but indexers might trade the tail. If they trade the tail on both sides, they’re already screwed going into the weekend b/c TSLA AH up.
This is a take. The thread is worth reading for the take.
I expect it overstates things.
Couple of problems.
1) while vanna can drive moves, I expect it is more gamma curvature and density than vol (in real trader terms, vanna of one-week options is meaningless).
2) if this were the case, it would be the case for other stocks too, but TSLA did not have the highest realized volatility of S&P500 (or S&P+TSLA) for the 52wks to 13Nov announcement. It's #21. Sure airlines and oilcos and cruise companies were on the list. But banks are too.
3) What really drives that number - volatility - over time is volume, and volume is best understood by how much of a company's float is traded every day.
Every time someone buys/sells S&P500 and dealers/arbitrageurs have to buy/sell the shares, increasing intra-correlation
Currencies do not become reserve currencies because they are easy to trade and ownership can be on the blockchain. They are reserve currencies because the owner can sleep at night with $100bn in one place in one asset.
In the end it is about trust. And reserve currency status is about a multitude of sovereigns trusting one other with the value of their accumulated ‘net retained export earnings’ - and the key is that the others with who’ll they deal have to respect that choice of collateral.
Afghanistan, the DRC, and Russia all have very low debt to GDP. But if some country held only treasury bills of those three countries’ currencies, one would not regard that as a wise choice of assets which would be deemed acceptable collateral to a vast array of other sovereigns.