OK, now that the article title has your attention, I'd like to start by saying I commend Ark Investment Management LLC for providing a financial model regarding how the firm arrives at its Tesla (TSLA) valuations and providing it as "open source" rather than simply providing a headline price and only providing details to its paying clients. It advances the ability of individual investors to make rational investment decisions.
Seeking Alpha has further advanced the cause by providing a recent news post and summary of the article, with links to the underlying model, recently provided here.
I have identified two major flaws in Ark's "bear case" model which I'd like to focus on. The first is estimated additions to "property, plant and equipment" (P,P&E), which has a major internal inconsistency and multiple minor errors. The bear case shows annual production/sales increasing to 1.7 million units per year by 2023 with gross investment in property, plant and equipment being less than $3 billion in total over the entire five-year time frame, not much more than Tesla plans to invest this year alone.
The second item regards capital raises and outstanding share counts in 2023. The model seems to completely ignore the impact of employee stock option exercise, restricted stock vesting and Employee Stock Purchase Program issuances.
This article is focused almost entirely on these two issues and is not meant to debate whether estimates such as 1.7 million units and 25% gross margins are part of a reasonable 2023 bear case. I will simply point out that the valuation that ARK proposes is physically impossible based upon the errors in these two items in the model even if all other assumptions are correct.
Unfortunately, ARK's public model is somewhat incomplete. It only provides minimal income statement items and does not contain cash flow or balance sheet projections. The model is actually titled: "Excerpt From ARK's Tesla's Valuation Model," so we are not being provided with their full model.
It's also something of a "hybrid" model, not addressing energy generation and storage, or even services, which is an integral part of Tesla's automobile business. Therefore, ARK's price targets don't necessarily represent the company as a whole but simply the market value of the automobile business.
Property Plant & Equipment Investments
I have provided the first section of Ark's model below, which provides the P,P,&E figures:
The table shows gross P,P,&E as $14.029 bn. at Dec. 31 2018, which is the actual number reported in the 2018 10-K. It shows the projected gross amount under the bear case as being $17 bn. in 2023, (presumably year-end), an increase of just under $3 bn. in five years.
For many reasons, a total increase in P,P,&E of only $3 billion in five years is extremely unrealistic. Tesla actually has a capex budget of $2-2.9 billion in 2019 alone, (which is not quite the same as P,P,&E additions, but that's mostly a discussion for another day), but of course a small portion of that may be related to its energy division.
We also can simply look at this from the perspective of how much in capex is likely to be needed to reach capacity of 1.7 million cars in five years as opposed to its 2018 year-end capability of less than 500,000. Tesla has estimated a cost for the Shanghai Gigafactory of about $2 billion, which would possibly add another 500,000 of annual capacity. An additional 700,000 cars would be an additional 1 1/2 Gigafactories, or $3 billion. Of course, if Tesla sells 1.7 million cars that year and already has a few million on the road that it sold in prior years, it would need to have a greatly expanded sales, service and supercharger network in place, which would be billions more. Will an additional battery Gigafactory or two also be required? Will additional capex at Sparks be required for Tesla's expanded planned auto production? Even a rough estimate indicates Tesla may need to add $7-10 billion to P,P,&E, not less than $3 billion, to reach production of 1.7 million cars in 2023.
More simplistically, you can just look at Tesla's recent annual P,P,&E expenditures and multiply by 5. In 2018, this figure was $2.1 billion with at least this amount planned for 2019. Even assuming a bit of that was for non-automotive operations, this method gets us to a figure approaching $10 billion.
Where did Ark get its $3 billion figure?
Ark explains its thought process in the note next to the model's "Gross P,P,&E" line as follows: "GM produces ~8mm vehicles on 60bln PP&E: 7.5k per vehicle. Ford 6mm vehicles on 35bln: 6k per vehicle. Bear case Tesla: 10k per vehicle, bull case 6k per vehicle."
This is wrong in so many ways:
- Ark's methodology for its bear case is to multiply 1.7 million cars by $10,000/car, arriving at $17 billion total, and then subtracting the $14 billion it already has invested in gross P,P,&E. They are basically saying that although it cost $14 billion to build capacity for the first 500,000 cars (or $28,000 per car), it's only going to cost $2,500 per car for the remaining 1.2 million.
- Tesla has an integrated business model, so you would expect the P,P,&E per unit of additional manufacturing capacity to be much higher for Tesla than for Ford (NYSE:F) or GM (NYSE:GM). The company not only produces its own batteries but even various components like seats. It also is its own dealer network. For starters, Ark needs to add the P,P,&E for Ford and GM's dealer networks and then can make downward adjustments from there if it thinks Tesla's model is more efficient.
- The P,P,&E in accounting records is historical cost, not replacement cost. Some of the property Ford owns could have been purchased 100 years ago. For GM it's even worse. The company went through a bankruptcy reorganization with many assets on the balance sheet revalued as a result.
- There are constant adjustments to gross P,P,&E on the balance sheet, so the recorded P,P,&E doesn't even necessarily represent the historical cost of assets currently in use.
- Tesla made a huge accounting adjustment in Q1 resulting in gross P,P&E decreasing from $14.029 billion at Dec. 31, 2018, to $12.769 billion at March 31. Tesla's explanation for the decrease is as follows: "As of December 31, 2018, the table above included $1.69 billion of gross build-to-suit lease assets. As a result of the adoption of the new lease standard on January 1, 2019, we have de-recognized all build-to-suit lease assets and have reassessed these leases to be operating lease right-of-use assets within the consolidated balance sheet as of March 31, 2019 (see Note 2, Summary of Significant Accounting Policies). This includes construction in progress associated with certain build-to-suit lease costs incurred at our Buffalo manufacturing facility, referred to as Gigafactory 2." If Ark simply uses the same method, but starting with the March 31 figure rather than the Dec. 31, amount, the additional capex required would be $4.231 million, not $3 billion. Of course, this disclosure also indicates that Ark was using some of the Buffalo solar microfactory investment in its calculations.
- There's also the complex accounting for equipment owned by Panasonic at the Gigafactory, which is included in the Tesla P,P, & E figure as well. Per the March 31 10-Q, "As of March 31, 2019 and December 31, 2018, we had cumulatively capitalized costs of $1.41 billion and $1.24 billion, respectively, on the consolidated balance sheets in relation to the production equipment under our Panasonic arrangement." If this is subtracted from Tesla's P,P,&E to get a true figure for Tesla owned equipment, then the additional amount needed to reach $17 billion becomes $5.64 billion.
- Some of Tesla's P,P,&E is for Tesla Energy, including assets acquired from SolarCity. Subtract a few hundred million dollars for this and the additional investment required is easily over $6 billion. This total also starts to jive more closely with Tesla's estimate of the additional $2 billion required for additional 500,000 of capacity in Shanghai.
As Seeking Alpha contributor Keubiko has previously identified, both in comments on Seeking Alpha and on Twitter, Ark also made a huge error in its calculation of annual depreciation expense. Instead of using Tesla's annual depreciation expense in 2018 of $1.1 billion to calculate its baseline depreciation percentage, it used the accumulated depreciation amount of $2.6 billion. Keubiko explains this error in detail here.
The Equity Raise Misstep
Ark's model assumes that Tesla will issue 26 million shares and raise $10 billion, which requires an average price of $385/sh., resulting in total shares outstanding of 204 million in 2023 under Ark's bear case. A reformatted excerpt from ARK's model is below:
We could debate forever whether $385/share, double the current stock price, is reasonable or not, but that's not my concern. It's that Ark has completely ignored the potential shares to be issued between now and 2023 as a result of various Tesla stock compensation programs.
Tesla has three major stock compensation programs: Stock option grants, restricted stock grants and Employee Stock Purchase Program issuances. The expenses for these items are recorded before the actual stock issuance, typically as the benefits vest and the employees are able to exercise their rights, even if they don't do so until years later. Restricted stock generates no cash when the shares are issued, while the cash generated by stock options and ESPP purchases depend upon the exercise price, which can vary greatly and we don't have "granularity" on this issue.
However, we do have total amounts for all three programs combined from Tesla's financial statements. The 2018 10-K cash flow statement shows Tesla received $296 million due to "proceeds from exercises of stock options and other stock issuances" while according to the "Consolidated Statements of Redeemable Noncontrolling Interests and Equity" the number of shares issued was 3.568 million for these programs. This means Tesla received an average of $83 /sh. for issuance under them. The results were similar in Q1 2019, with 1.079 million shares issued generating $78 million in cash. ($72/sh.)
If we simply assume new employee stock issuances will average $100/sh. and the issuance continues at 1 million shares per quarter for the next 19 quarters, the share count will increase by 19 million while raising only $1.9 billion.
In addition, per p. 63 of the TSLA 2019 proxy statement, at year-end 2018, Elon Musk had exercisable options for more than 4.7 million shares at $31.17/sh., expiring in 2022. These options will undoubtedly be exercised prior to 2023 (in fact 175,000 of them were just exercised) generating total proceeds of about $148 million.
The above estimates suggest a total of 23.7 million shares, generating a bit over $2.1 billion, could be issued under these stock compensation programs. Since ARK's model is projecting that a total of $10 billion of equity needs to be raised, this would require an additional $7.9 billion. Using ARK's average price of $385/sh., this would require the issuance of another 20.5 million shares. As a result, making this adjustment to ARK's bear case model would result in about 225 million shares being outstanding in 2023 rather than 204 million.
Ark's model has some very basic errors and logical inconsistencies with respect to both estimated capital investment and stock issuance requirements for its bear case prediction of 1.7 million sales in 2023. Some of the errors are so fundamental that if the model were submitted by an accounting or financial analysis student for an introductory class, it might struggle to even receive a passing grade. It's puzzling that a high-profile investment firm, such as Ark, with a team of trained analysts, would make such errors.
Disclosure: I am/we are short TSLA. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.