How Insane is ARK’s Valuation for Tesla? —Let's Have a Look...

When ARK Invest released their new price target for Tesla for 2025 I was pleased to learn they used a Monte Carlo simulation. So, they not only made their valuation model available by posting it on GitHub as they'd done before but now it seems ARK also adopts more advanced techniques than simple “bull” or “bear case” scenarios—as the “traditional” Wall Street analyst does.

Of course, I couldn’t resist rebuilding ARK’s valuation model for Tesla in Analytica. (It took me quite a while to dig through all the rows and columns of the Excel model, which is how it works with complex spreadsheet models... That's why I'm quite late for the party ;)

Now, here is how it looks like:

Let’s have a look at the model

Now, really!

The reward of my endeavor is having the model in Analytica makes it transparent. You might say that sharing the Excel model already makes it transparent. But that’s not entirely true. It’s true that you can analyze the calculations, scrutinize the assumptions, and even play around with your own numbers. But that doesn’t make the model transparent.

Let me explain!

To truly understand what’s going on in the model, you need to develop your own mental model. You do this unconsciously, translating what’s buried in the rows and columns of the Excel sheets into a kind of inner representation. Typically, these representations are “objects” that are somehow connected.

Like in an influence diagram.

With Analytica, the model automatically appears as an influence diagram. This is the way models are displayed, and the way the user interacts with the model.

Here is ARK’s model for Tesla’s electric vehicle business:

In this article, I don't want to discuss the model itself or the assumptions that go into it. That's for another post or two.

But as you can see, even without knowing anything about ARK‘s assumptions and the specific calculations applied in the model, you can easily see what aspects are considered and how they influence each other, and finally Tesla’s EV business EBIT.

That’s why I said, let’s have a look...

The future is uncertain, so let’s treat it like this

As I said, using Monte Carlo simulation is a huge step forward. It allows to make the range of possible outcomes explicit and tells how likely certain results are.

Using Excel, and especially its data table function to do so, is perhaps not the best possible solution... The problem is that the simulation scenarios for inputs and every intermediate variable along the way down to the bottom line are not visible at once. You can see the numbers change when you repeatedly hit the F9 button, but that doesn’t tell you anything meaningful.

Of course, ARK shows some results in the form of percentiles which they label “bear” and “bull case.” That's quite okay. Apart from that, the only way to give the reader a flavor of the Monte Carlo simulation in the Excel model is to provide some samples for key results, as they do:

I personally like how Analytica lets you visualize the simulation scenarios for each variable in the model. The share price in 2025, the key outcome of ARK’s model, looks like this, for example:

Many people like to look at the results as a histogram or probability density chart—maybe because it looks so nice...? More useful, though, is probably (pun intended) a cumulative or an exceedance probability chart. It tells you more precisely how likely it is to get a certain result.

Sometimes it’s also useful to see the possible range of future development. This is when you use the probability bands chart.

For car sales and costs of goods sold in ARK’s model it looks like this:

And, to match ARK’s model, I also created a table that shows the individual simulation scenarios for key metrics.

But, in Analytica, you also have the possibility to choose a couple of other options to display the results, e.g. in this case in the form of percentiles.

What drives Tesla’s share price in ARK’s model?

The model ARK developed has a total of 34 inputs. For each input, they make assumptions for a bear and a bull case value as well as a minimum and maximum.

Then they use the bear and bull values to define a normal distribution with a mean right in the middle between bear and bull and the standard deviation at these cases.

The reason for using a normal distribution might be that this is the easiest way to build a Monte Carlo simulation in Excel without simulation add-ins like @RISK, CrystalBall, or Vose’s ModelRisk. ARK simply uses the RAND and NORMINV functions to do so.

There are several problems to this approach which I will discuss in a separate post. But one is that the bear and bull cases are, obviously, different from the P10 and P90 values that are typically used in risk management, and also from the top and bottom quartiles that ARK uses to define their bear and bull price targets.

What’s missing in ARK’s model, and also in the corresponding blog post, is an analysis of the impact that each individual input has on the variation of the share price.

Or, in other words, which assumptions are most critical to Tesla’s valuation in ARK’s model?

In Analytica, this is quite easy to do. There is a standard function that lets you run sensitivity analyses on all inputs at once. The resulting tornado chart tells you how the result is changing when an input is set to its P10 or P90 value, keeping the rest at their median (or P50) value.

When we run this analysis for all 34 inputs we get, within seconds, this nicely-looking tornado.

As you can see, half of the inputs seem to have almost no impact on the overall uncertainty in the share price. Zooming in to the top of the chart, we have all relevant assumptions listed according to their impact.

I'm not going to discuss the analysis here. For now, I just wanted to show the tools and approach to do this. Because this is where a fruitful discussion can start.

And, of course, you can run this analysis not only for the end result, i.e. Tesla’s 2025 share price. For the three business lines that ARK has modeled, the tornado analyses look like this.

Tesla’s Electric Vehicle Business

Tesla’s Ride-Hailing Business

Tesla Insurance

If you’re interested in the model and want to have a live presentation, just reach out to me here at LinkedIn or tr [at] syconomic [dot] com. I’d also like to hear your thoughts on the model, the Analytica model, or how you think AKR’s model could be improved. (For one, the whole energy business is missing which some folks believe might become bigger than Tesla’s auto and transportation business...)

Let me know what you think!

Happy to discuss!