I’ll ask my wife’s boyfriend if he can lend me 20 uninterrupted minutes on his Excel computer.
9 Likes
I have to admit, this ARK target report is truly a stuff of legend. Many have already commented on it, but I’ll share some highlights from Christopher Bloomstran…
Let’s go right to the insurance valuation and model update, where the analyst assumes in a BEAR case Tesla’s “Insurance Operation” is worth $23B in 2025, four short years from now. It’s clear that the author and firm have no clue how insurance works. Harsh, I know.
To begin and to be clear, Tesla has ZERO insurance underwriting operations. They are brokering auto policies in California alone for an underwriting fronting sub owned by Markel. The business written is so small to not be quantified either by Tesla or Markel in SEC reporting.
The analyst writes, “ARK believes that in the next few years Tesla could roll out its insurance offering to more states, underwriting its own insurance policies. Because its vehicles have better than average safety profiles, Tesla should be able to use real-time data to offer insurance in its vehicles, pricing it dynamically, lowering customer acquisition costs, and increasing margins. Relative to Progressive’s 13% EBIT margin in 2019, ARK estimates that Tesla could achieve margins close to 40%. If it were to sell 40% of vehicles with its own insurance offering by 2025, Tesla’s insurance revenues could approach $23 billion annually in our bear case.”
The ARK “bear case” assumes insurance revenues of $23 billion and a 40% EBIT margin in 2025 with 5 million vehicles sold, up 10x from 500k in 2020. Hold underwriting aside for a minute and think about brokerage.
Let’s give ARK credit and assume Tesla brokers the insurance on 40% of all Teslas sold over the next four years, call total sales 10 million units. At a ballpark industry average $1,600 in annual premium, total insurance policy revenue totals $16B on all Teslas sold.
Brokerages and agencies will earn on average 15% of total auto premiums as commissions, so revenues at the 2025 run rate yields $2.4B in annual commissions to TSLA. How in the world do you get to $23 billion in ARK assumed insurance revenues? Must be underwriting, correct?
Suppose instead of brokering policies for an underwriting front Tesla does get into the underwriting game. Here’s where the analyst & ARK know little or nothing of insurance. Let’s presume Tesla is admitted to underwrite auto in every state in which they sell a vehicle.
Suppose for grins Tesla writes the full forecast 40% of all policies on every Tesla sold over the next four years, say a generous 10 million vehicles (5 million in 2025). Tesla would be underwriting insurance on 4 million vehicles, its 40% share of the total. Starting NOW.
At $23B in 2025 “insurance revenue,” that’s $5,750 in annual premium/vehicle, more than $3,000/year above what the industry charges per vehicle now and in this case more than 10% of the total vehicle price (before software “upgrades”). Avg is closer to 5% with a wide range.
Let’s just say Tesla underwrites on 100% of total vehicles sold over the next four years (I’d take the under on 10 million for a whole bunch of money). At the analyst’s assumed $23B in revenues, that’s $2,300 per policy, still way higher than $PGR’s or the industry average.
Now, to underwrite insurance requires capital, yet the ARK model presumes a share count of 1B shares outstanding, well BELOW today’s fully diluted count. Tesla already has 960m shares out and 1.155 out on a fully diluted basis.
You can count on all fully diluted shares outstanding if the stock moves up to ARK’s $3,000 price target/share, a $3.47 trillion market cap, 112 times trailing 12-month revenues. Beyond asking how in the world will the company have enough capital to finance building cars at a 5-10 million annual rate with the capacity they now have and have under construction (Fremont, Shanghai, Germany and Austin) without selling new shares, where will the capital come from to underwrite $23 billion in annual premiums and expand auto production?
I wonder if ARK knows auto insurance is underwritten on an admitted basis in each state? Regulatory approval in each state will be required, as will capital - lots of it. I wonder if ARK knows admitted auto insurers can write $3 in premiums for each $1 of statutory capital?
Underwriting $23 billion in premium requires almost $8 billion in statutory surplus (equity). Progressive has about $14 billion. What will it cost Tesla to start an insurance underwriting operation on a de novo basis in each state in which it sells cars?
*It’s going to chew up a substantial portion of the $8 billion needed in surplus to underwrite $23B in business.
Here’s where the ARK report demonstrates its lack of understanding of insurance. How easy to say that Tesla will operate at a 40% EBIT margin versus $PGR’s 13%.
I’m not sure even a college junior with zero investing experience would make this Herculean leap. Tesla will, “achieve better than average margins on insurance thanks to the highly detailed driving data it collects from customer vehicles.” Wow.
First, the reason insurers don’t want to underwrite on Tesla vehicles is they are extremely expensive to repair. Their technological “advantage” against ICE vehicles means they have more technology on the vehicle - sensors, computers.
When a moron is tooling around thinking they are on FSD and rams into a semi or other vehicle, repairs are expensive, both to cars and to people. Next, the notion that Teslas are “safer” and will cost less to insure demonstrates a lack of understanding of insurance pricing.
*Beyond the notion that Teslas are safer & thus cheaper to insure, losses (claims for repairs/medical) & operating expenses (underwriting, claims paying, management, investing) combine to yield little industry profitability. It’s a regulated market with LOTS of competition. *
If an underwriter like Tesla (remember we are pretending they are one) is consistently too profitable, insurance commissioners will drive pricing down. If collisions trend toward zero (frequencies), pricing will reflect fewer claims and insurance costs will decline.
Insurance will always be required — the car accelerates into your living room on its own or the sensors miss the sinkhole in the road that just appeared and swallows the car and driver. If it made sense for incumbent insurers to write on Teslas, believe me they would do so.
Continue with profitability. I wonder if the analyst realizes that more auto insurance profitability is derived from investment income than from underwriting profit? A 40% EBIT margin on $23B in premium revenue yields a $9.2B profit before interest and tax.
PGR (Note from Julianek: Progressive Auto Insurance, the industry leader) has total investment assets of about $47B against just under $40B in premiums earned. Given Tesla will have no surplus capital and no reserving history, they will be compelled to invest most of their invested assets in fixed-income (of which they have none today).
I don’t see them earning more than $1B on a bond portfolio. Maybe the Tesla “insurance operation” will be allowed by regulators to invest solely in Bitcoin? If investments will be in fixed, the majority, maybe $8B of their EBIT profit of $9.2B, will come from underwriting.
The analyst ought to ring a few state insurance commissioners & ask, “Hey, ARK here, Tesla is going to be underwriting a ton of policies in your state over the next 4 years - we’re thinking they’ll write at a 35% pre-tax underwriting margin - you have any problems with that?”
Forget about the implausibility of a $3,000 price. Forget that the presumed $3 trillion forecast market cap doesn’t allow for new shares sold or even the vesting and exercise of more than 40 million of current option shares (there are WAY more issued but unexercised).
Forget the capital requirement to grow into 5-10 million vehicle production in 4 years (more than 5% to 10% of total vehicle global market share versus 0.5% today). The impossible assumptions regarding Tesla’s valuation just as regards insurance blasts a hole in credibility.
Ask whether a firm or analyst that presumes $23 billion in insurance revenues in four years for a yet non-existent startup insurance underwriter, even one with prospective “vertical integration,” understands enough about insurance to make the absurd claims proffered above.
When all is said and done, and the smoke screen will disappear, ARK will be remembered for what it is: a stock pumping operation toward naive retail investors. Investors who wanted to believe the hype and will be left holding the bags. But at least ARK will have made millions in management fees on its ETFs…
10 Likes
Well, why can’t it be different at Tesla insurance, if they can charge higher premiums than their competitors?! According to ARK, it can be a usage-based, pay-per-mile insurance. That’s what they’re giving as an assumption in their open-sourced spreadsheet:
*“Premium/mile for personally owned Tesla car (not in millions): 0.18”
I did a quick sanity check by googling industry averages, and competitors’ average per-mile rates should be in the single cent digits (5-10 cents per mile).
In any case, the 23 billion estimate is still low compared to Elon Musks predictions:
“Obviously, insurance is substantial. So, insurance could very well be, I don’t know, 30%, 40% of the value of the car business, frankly”
To be honest, even if this whole rant proves ARK’s incompetence in insurance business, their prediction of $23b is a very small piece of the entire revenue pie. So even if they’re wrong here, and Tesla makes $0 from insurance, it doesn’t change much. But of course, losing your credibility in one area puts your whole reputation under scrutiny.
In case you were wondering how this 10 million cars in 2025 should be reached, Steven Mark Ryan comes to the rescue with his Excel sheet. The Gigafactories in Berlin & Austin should be churning out over 2 million cars each. Additionally, 7 more Gigafactories should already be in operation. Hmm… I’m a Tesla fanboy, but I will be shocked if Tesla comes even close to these numbers.
To be clear, after having spent some time with their Github spreadsheet, I can confirm that the rest of ARK’s forecast is the same bullsh*t as their insurance analysis.
Example with the electric vehicle business. ARK forecasts 10 million vehicle sold per year by 2025. This is 20 times what the company delivered in 2020. Auto-manufacturing is not known to be asset light, and TSLA does not escape the rule. You need roughly $1 of fixed assets and working capital for each $1 of revenue (which is in line with TSLA’s 2020 numbers: revenues of $31 billion, whereas Property, plant and equipment + working capital is roughly $33 billion. Let’s assume that each vehicle will be sold at $40’000 per unit. That represents revenues of $400 billion, and the company will surely need to make investments of the same magnitude to manufacture all these cars. Gigafactories are not free! Given that the company is barely profitable, you cannot count on reinvested profits to fund the growth.
But then we enter in the magical ARK’s world, where capital growth is free:
- No equity raise needed to build gigafactories, even in the BEAR case:
- When the volume of cars sold is multiplied by 20, the working capital does not move a bit. Of course, the inventory needed to build those cars is free!
There is absolutely zero analytical rigor in this forecast. Even a junior analyst fresh out of school could call it for what it is: a big, fuming pile of bullshit.
But hey, at least they can say they used a Monte-Carlo simulation!
10 Likes
As I’m sure we agree, full self-driving and if and when they’ll be achieving is a somewhat uncertain evolution in technology - so you have to make assumptions there. But it’s not just the supposed insurance business (or the Bitcoin bullshit one of their cryptoasset analysts has been spewing out on Twitter, which we’ve touch upon above), or failure to account for dilution in stock.
They are simply making their numbers - and then tailor their data inputs for their models accordingly.
And then, even when predicting industry-leading growth and margins and capex efficiency, they’re calling it an “Example bear case”.
I’m going to cut this short, since I’ve seen Julianek has posted concurrently.
Side note: They’d be the most productive car plants in the world at that by considerable margin. No car plant in the world does this today.
The astonishing thing is how this then goes out and becomes picked up by Reuters, Yahoo Finance and Bloomberg - and ultimately on social media (Youtube, Twitter, Reddit). How little they’re being called out on it. And how much people seem to take it (maybe not at face value but at least) as a frame of reference for their own opinion, target price and investment decisions.
It is kind of strengthening my impression that the markets aren’t always rational and efficient - and on the other hand my doubts about how efficient plain index investing is.
PS: In other words: That overperformance by picking stocks is possible.
4 Likes
TSLA had a +20% day on no news a few weeks ago. And no big macro news. That is >$100bn move in a day!
The SNOWs, SHOPs, LMNDs, etc. valued as high as 200x sales at one point.
The PLUGs and QS valued at $20bn with no sales in their pipelines till 2024.
The NKLAs, WDIs, etc. valued at many billions after fraud is unveiled and their business model is videoing a car rolling down a hill.
Bitcoin miners valued at $4bn; an impossible amount for them to all return to shareholders. Of course investors have no idea how blockchain works, let alone bitcoin mining.
I can go on and on. Markets are not efficient in the short term (EMH only applies long term).
The problem is; the risk reward is not there to bet against companies when next biggest fool investing is rampant, interest rates are constantly being pressured down, and an increasing amount of systematic risk is placed on the stock market.
Even if your thesis are correct it is not easy to reap rewards due to asymmetric returns, margin calls, expensive IV on junkm etc.
Index investing avoids this and allows you to profit from EMH long term.
3 Likes
How did you unprotect that workbook? You generated a password?
I didn’t get into the details. Are you sure about what you’re saying? Is this cost of building gigafactories not covered by something else? Like “total long term debt”? We can see that it goes from $10b to $45b.
I checked the formula and it’s: (gross property, plant & equipment) * (percent of factory build debt funded). So indeed they finance factory build through debt.
I also noticed that when you reload the file, the values change, probably due to the use of random functions. That’s where the Monte Carlo kicks in. In my sheet the PP&E is $87b, much of it funded by debt.
On the website they state that they predict capital expenses of $6’000 per car, with an average selling price of $36’000.
If I understand capex correctly, it is the cost that it takes to build another car. $6’000 / $36’000 puts it far lower than your stated $1/$1. Were talking about miracle Elon here, how can you be sure he can’t make it happen?
One more thing I found in that sheet: apart from a $25’000 “Model A” they also predict a $15’000 “neighbourhood car”, something I have not seen or heard anywhere yet.
@Julianek one more thing: don’t you think it’s at least respectable that they publish a pretty detailed excel sheet with their valuation model? At least there are some concrete numbers that you might not agree with, but numbers and formulas are concrete things that really put you at risk, if you don’t know what you’re talking about.
If at least someone in ARK was aware that they’re full of shit, they would for sure not allow to publish any kind of bogus Excel file. So I think at least they themselves believe in what they put in there.
OK, one more thing: I’m curious: how a valuation of anything happens in financial investing analysis? I didn’t study the Excel to find how ARK does it. What I get is that they predict the future costs and revenues. But how does that translate to stock price, and not today, in the future!?
Let’s say there is a special coin. Whoever owns this coin will every year receive $100, into infinity. The payout is 100% certain. How much is this coin worth? How much will it be worth in 2025, the same or some different amount, depending on the predicted interest rate movement? I’m sure there would be 10 analysts and each would give a different price and reasoning, right?
I downloaded the project from github on my mac. When I opened the sheet with the Numbers App, it did not require any password. If for any reason the version I have is not the correct one, please contact me so I can update my point of view. But I have the same figures as you for gross PP&E, so I am quite confident that the document version I have is what was published by ARK.
No it isn’t, what you mention is the cost of good sold (COGS). I realize that not a lot of us on the forum is familiar with accounting, so I will try to explain it in layman terms as much as possible.
Let’s start simple with the Income statement (you can find them on Yahoo finance).
Last year (TTM = trailing twelve months), Tesla had revenue of $31.5 billion (= sales of goods and service). For simplicity, I will assume assume that all of the revenues come from selling cars (which is not too far from the truth, as it is currently their main business by far. If one wants to be picky, we could remove the regulatory credits which won’t scale with volume).
The cost of good sold (i.e the aggregate cost of making each car, without taking into account administrative expenses and other fixed costs) was $24.9 billion.
Once all expenses (including debt interest and tax) had been paid, the company had profits of $690 million.
That means that TSLA makes profits of 0.69/31.5 = 2.19% on each dollar of revenue (In other terms, its net margin is 2.19%.
We also know that TSLA delivered 500’000 cars in 2020. So I will assume 500’000 cars sold = $31.5 billion revenues.
However, those sales were not generated by pure chance. The company had to invest in hard assets like factories, plants and machines to manufacture the cars before selling them. That’s why we also need to look at the balance sheet to figure out what are the capital needs of the business to generate those sales.
Let’s first look at the “non current assets”, where most of the factories value resides. The “Property, Plant and Equipment” (PPE) line is what we are looking for. We learn that their gross value (i.e what the company paid in order to build/acquire those factories) was $29.4 billion, but they have now depreciated to $23.3 billion.
So the first relationship we can make is that “in order to generate sales of 500’000 vehicles per year (and generate $31.5 billion of revenue per year), you need factory capacity which costs $29 billion”. Which is roughly $1 revenue for $1 PPE, in line with the rest of the industry.
What we call capex (=capital expenditure) is a jargon term to describe the investment needed in additional plants, factories and equipment to generate additional sales.
So, now we have ARK telling us that within 5 years, TSLA will ramp up production to 10’000’000 units per year. That’s 20 times current production. Maybe the existing factories will have spare capacity, and maybe there will be some synergies, but overall I cannot see how the company will not need at least ten times more factories. That would mean an investment in PPE of close to $300 billion.
How to fund that? There are four ways to fund capex:
- Available cash: TSLA has $19 billion of available cash, so it won’t be enough. (In addition, it is not even sure that it could be used. For instance, the Shanghai gigafactory contract with China’s government stipulates that any profit made in China has to remain in China).
- By reinvesting profits: at current rates it would take TSLA 300 years to fund this growth with profits, given its meager margin profile.
- By taking on debt. TSLA bonds currently pay a coupon of 5.3%. Apply this rate to $300 billion and the annual coupon payment would be $15.9 billion (more than 16 times its annual profits). Any lender willing to provide the money would never see it back.
- So the only way to fund PP&E is through equity raise, which ARK did not even consider.
But if only it would stop here! In order to grow sales, you need more than additional plants. You need something accountants call working capital. That is all the temporary cash you lay down to buy raw materials, inventory and so on, as well as all the credit you provided to your customers (buy now, pay later).
-
In 2020, Tesla had to invest in $4 billion of raw material and inventory in order to make its cars. This usually grows with the volume of cars manufactured.
-
Account receivables (i.e credit given to customers, i.e cash not received yet and thus still tied to the cars) has grown from 1.3 billion to 1.9 billion, a 46% increase. (which by the way is very curious, given that TSLA’s business model is to make the customer pay in advance before the car is even built. This leads to embarassing questions about the veracity of those sales).
-
There are other elements coming into account, but the bottom line is that in addition of PP&E, the company needs to invest $12 billion of working capital to sell 500’000 vehicles. Most of it (like raw materials) varies proportionally with the volume manufactured. So if you need to produce ten times more vehicles, you will likely need roughly 10 times more working capital, and so on. This puts us with working capital needs roughly between $100 and $200 billion. Something the spreadsheet conveniently forget to mention. It assumes working capital will stay at $12 billion, which is bonkers.
He’d better find something quickly because each year where he does not find a solution is one year where ARK’s forecast is getting more extravagant. Given the time it takes to deliver FSD since its initial promise, I think we can bet safely that this capital requirement issue won’t be solved in five years.
It’s called an annuity. If payments are made into infinity, the value is simple: Value = annual payment / discounting rate, where the discounting rate is your opportunity cost (i.e the investment return of your second best alternative).
Either they don’t control what they publish, or they don’t understand what they are talking about. I don’t know which one is worse. Or they are perfectly aware of what they are doing because they desperately need TSLA to go up to avoid redemptions on their less liquid holdings.
7 Likes
I’m saying this, because the excel file I downloaded is protected, which blocks any changes to it. And only years 2020 and 2025 are visible, other colums are hidden. So I thought you somehow unlocked it to unhide the columns, because in your screenshot I see all years.
Will read the rest of your post in the morning 
So the first relationship we can make is that “in order to generate sales of 500’000 vehicles per year (and generate $31.5 billion of revenue per year), you need factory capacity which costs $29 billion”.
I think this assumption is not correct. If what google spits out is correct, the china factory has a max capacity of 500’000 per year and costs 2 billion. Tesla: Chinese factory could cost billions less than everyone thought
Tesla Plans China Plant With 500,000 Vehicle Capacity | IndustryWeek
The Fremont plant should be able to produce 600’000 vehicles.
ARK estimates 60 billion PPE for 6 million cars annually. Tesla Factory Construction and Lowering Vehicle Costs | NextBigFuture.com.
1 Like
Could it be that you only need to save a copy to make it unblocked ?
My PC at work does that with all excel files downloaded.
I’d be surprised if the ARK analysts are somehow profoundly better than elsewhere. They don’t have massive experience.
I think from linkedin, none of them had a very remarkable career before ark. If ark is somehow this brilliant fund that beats everyone, I’d expect at least some of them to have been at some prominent VC funds, or top-tier bank/financial institutions.
(not all of them, but on average you’d think a few of them would have been remarkable elsewhere)
It puts some credence to them having this one lucky allocation that put them on the map 
2 Likes
You only pay 100$ in advance to reserve the car. The rest you pay shortly before delivery.
Doesn’t help. The whole workbook is protected so I really wonder how Julianek can see data for 2021-2024. 
And thank you for that. Still, I can’t say I understand capex efficiency. What does the capex of $6’000 per car mean? The way I interpret the wikipedia definition, these are the running costs of repairing the roof, repairing the machines, that the company needs to bear in a given year, in order to maintain production. So the costs of expanding production are not included here… or are they?
If you assume average price of $40’000, that gives $10 revenue per $1 capital, 10x more than stated as industry standard by @Julianek. Is it really fair to pit 500k cars against the $29b PPE, when we know that the Fremont factory was not built from scratch in the most efficient way, Giga Shanghai is not yet fully ramped up, and a lot of capital has already been invested Giga Berlin & Texas, which have yet to produce their first car.
I found in the workbook, that they use EV/EBITDA of 13 as input. So Enterprise Value = Earnings before taxes * 13. It seems weird to me to just use that, but apparently this is how it’s done. You just take a value that is standard for a given industry.
By the way, this discussion shows the beauty of index investing. You can have no clue and still be guaranteed the return of an average investor, which I’m sure has a much higher knowledge than me.
1 Like
Actually I think you got in wrong. I am not an accounting expert like Julianek, but since I was in the plant building business (wastewater treatment, so also long lifespan etc,) I have some idea of what he is talking about.
Basically when 2 offers from a plant where compared from the price point only, you look at needed CAPEX (Capital expenditure) and OPEX (operational expenditure).
CAPEX is the cost of building the plant, including the cost of getting that capital. Basically think of it as the value of the buldings, machines, and everything you need to put it together (manhours in planning, actually building it etc.), and how much it costs you to get necessary credits etc (interest of bonds). Now this plant has a certain lifespan, where you need to replace the machines , roof etc. (not repair them, that would be different). If you estimate the amount of vehicles you can get out of the factory for that lifespan, you get your CAPEX/vehicle.
OPEX would be how much you need to run the plant (Raw materials, Running repairs of machines (spare parts also have a lifetime of running hours), workforce. Again, you take all this costs, divide it buy the number of cars and voilà.
In general, CAPEX+OPEX than represents your total manufacturing costs of the car. You will note that this does not include development of the car yet, which in general (at least where I was working) is covered by the Overhead and is included in the margin. The Overhead would be R&D,admin people, your CEO, Shareholder meetings, office buildings, communications, bookkeeping etc. Basically all the side stuff you need to make your company running. This is actually least correlated with the size of your factory, you can build 1 or 1’000’000 car, still only one CEO for example. So it is in the Overhead where the most of the Scale up factor lies. Not only but most.
@Julianek : please correct me if I am wrong somewhere.
3 Likes
I think repairs really count as capex. That’s what wikipedia says. Capex are expenses that go for non-consumable parts of the industry. I guess this distinction is not so sharp and you need to draw the line. I think if you buy something that instantly gets consumed to build the next product (like a sheet of paper or metal), that’s opex. If you repair a cog in your machine that might work for the next 10 years, that’s capex. But basically everything is subject to entropy, so this division seems pretty artificial to me.
@Julianek so it looks to me like the actual cost of building and maintaining gigafactories is much lower than the operating cost. We’re talking about $2b to build a gigafactory. In fact their current annual capex are between $2b and $4b. And regarding opex, do you really need to raise capital to cover these? Are these not covered by sales?
Ya that is what I meant with running repairs basically the spare parts you know which will run out after some running hours. The technical term for that would be wearing parts.
I guess you make the difference when the repair is actually increasing the initially thought lifespan (then it would be CAPEX). If it is just to bring the machine to the expected lifespan, it will be OPEX. But as you said, it is either in one or in the other, so not much change in the total cost.
1 Like