Analysis of My Tesla Financial Model on GitHub
Yesterday, I got some good feedback on my post about releasing a Tesla Financial Model on GitHub*.
It occurred to me an analysis of my Tesla financial model would be useful. I can’t expect everyone to download it. I was able to work and create a simple macro that copied and pasted the Monte Carlo values 5,000 times. It’s not the prettiest approach, and it takes a minute to run, but it works.
Line-by-line assumptions
Let’s go through line by line.
EV Revenue growth rate — This is how much electric vehicle revenue can grow year over year. The limits I set were -20% and 80%. Over many runs, the average should be 30% growth year over year.
EGS Revenue growth rate — This is how much Tesla’s Energy Generation and Storage revenue can grow year over year. The limits I set were 0% and 100%. Over many runs, the average should be 50% growth year over year. EGS revenue is expected to grow faster than EV revenue, since it is starting from a smaller base.
Wright’s Law decreases ASPs by — Every single time cumulative production doubles, Wright’s Law will step in and reduce revenue between 10% and 25%. (In reality, it is more of a gradual change.) In this run, the randomly generated value was 14.8%. The average over many runs should be 18%. Note that 20–25% was reported by researchers over time for Li-ion batteries. The value independently applies for EV and EGS growth, but EV and EGS may double cumulative production in different years.
Free Cash Flow growth rate — Free Cash Flow is cash flow from operations minus any capital expenditures. It is “what cash is remaining” after subtracting costs to invest in the business. Future Free Cash Flow is discounted back to the present to produce an estimate of a stock’s value. Over time, -4.8% and 5% will offset each other, or one is the inverse growth of the other (1.05 * 0.952 is close to 1).
Interest rate for NPV calculations — This number determines how much to discount future cash flow back to the present. I picked 2% to 8%, which is the same as ARK’s model, and I thought that made sense. In this simulation run, the independent, random variable gave an interest rate of 7.9%.
EV Revenue (millions) — Tesla’s full year EV revenue for 2020.
EV Produced — Tesla’s full year number of EVs produced for 2020. Both are from SEC filings and quarterly press releases.
EGS Revenue (millions) — Tesla’s full year Energy Generation and Storage revenue for 2020.
Energy Generation (GWh) — Tesla’s full year EGS in GWh for 2020.
Cumulative EGS Produced (GWh) — I fudged this number. It’s greater than 3. I haven’t gone through the filings to update this number yet for cumulative production. I am using the Energy Generation numbers from 2020. This is a large simplification of utility Megapacks, residential Powerwalls, and residential solar and solar roofs.
Free Cash Flow (millions) — This is how much Free Cash Flow Tesla had in 2020. It works out to be close to 9.5% of Tesla’s revenue in 2020. This is from its 2020 Q4 shareholder deck.
How does my model work?
Tiny! Zoom in 200 to 250% for a better view.
My model works simply. Every year, a revenue growth number is generated between the revenue bounds we set on the model assumption page. This is a random number that is weighted between the lower and upper bounds. This growth factor is applied to our starting revenue for each area. The same percentage growth is applied to EV and EGS starting production numbers. This yearly production is added to the prior year’s cumulative growth. When the cumulative growth from our starting values is greater than multiples of 2, my model will discount revenue by Wright’s Law on the model assumption tab. CleanTechnica reader “Actually Thoughtful” pointed out that Tesla may be able to reduce revenue less than how much costs drop due to Wright’s Law, and keep the extra revenue for itself. This is a good point, and I’ll have to figure out how to model that difference. It’s not present in the above model. Wright’s Law and revenue generation are independent for now. I assume that revenue goes down by the Wright’s Law cost difference every time cumulative production doubles.
The total revenue for both areas is generated. We randomly generate a free cash flow–to–revenue factor, which is added to our starting free cash flow–to–revenue factor of 9.5%. My thought process here is we don’t know the path of Tesla’s free cash flow over time. It depends on how much they invest in the business and how successful they are in converting revenue to free cash flow. It can be wildly positive or even negative. Again, to keep things simple, both areas are independent for now. This generates a yearly free cash flow value, which is discounted by our interest rate of 7.9%.
A perpetuity is a stream of constant cash flow that continues forever. To get the present value of a perpetuity, you take Tesla’s free cash value from 2030 and divide by the interest rate. Then you divide by (1.079)^10 to bring that value back to present day 2021. In our example above, Tesla’s net present value (NPV) over the next 10 years is $75 billion and the post-2030 value is $96 billion, giving us a total estimated value of $171 billion. Net present value is how you take a stream of cash going in and out, and determine its value in the present day.
How good is the model?
I may be overstating things, but the model seems decent. I know it’s wrong, but it’s still useful.
The model averages for individual variables is very close to their predicted average, over 5000 runs. I bet if I changed the free cash flow growth rate to multiply together, we would get it closer to 0%. I’ll make that improvement next time I re-run the numbers.
Thank God, Vijay, a chart! Hard to read, and my eyes are watering, but we’ll take it.
This is the distribution of the Tesla model’s discounted Net Present Value. And here are some high-level statistics:
| Average | $395,670.67 |
| Standard Deviation | $511,487.20 |
| Maximum NPV (thousands) | $5,976,195.02 |
| Minimum NPV (thousands) | ($1,064,735.57) |
| Count below $0 | 483 |
| Count above $1T | 444 |
I don’t know why, but I think a gamma model would be a good fit, perhaps because it’s easier to calculate. 🙂 A lognormal distribution looks like it would work, too.
As a reference, as of the close of the US market on 4/12/2021, Tesla’s value according to Google Finance was $673.8 billion.
What does the model tell us?
I found it interesting to look at what model values created the maximum and minimum NPVs.
The maximum was run #582. In this example, Wright’s Law was 12%, interest rates were 2%, Tesla produced 22.7 million vehicles in 2030 and 242 GWh in EGS. The free cash flow–to–revenue ratio was 22%, giving us a value close to $6 trillion.
The minimum was run #1886. Here, Wright’s Law was 17%, interest rates were 2%, Tesla produced 20.7 million vehicles in 2030 and 146 GWh in EGS, but the free cash flow–to–revenue ratio was -6.5%, giving us a value of close to negative $1 trillion.
Run #663 was close to the average. Wright’s Law was 18%, interest rates were 6%, Tesla produced 11.2 million vehicles in 2030 and 139 GWh in EGS, and the free cash flow–to–revenue ratio was 11.8%, slightly above where we ended 2020. This gave us a value of $395.6 billion.
Run #503, with $673 billion, is close to Tesla’s current value. Wright’s Law was 20%, interest rates were 2%, Tesla produced 3 million vehicles in 2030 and 236 GWh in EGS, and the free cash flow–to–revenue ratio was 16.8%, well above where we ended 2020.
What is clear is that Tesla has multiple paths to justify its current value.
What were the best scenarios? They were future paths where Wright’s Law was less than 18% cost reduction, there were low interest rates below 5%, and there were high free cash flow rates as a percentage of revenue — above 9.5%. Stocks are supposed to be valued on future free cash flow, discounted to the present. EVs as a percentage of revenue were similar to our current ratio. This makes intuitive sense. Looking at the average of the above scenarios, I get $1.03 trillion in value.
The worst scenarios were a future of high interest rates, fast production cost declines of Wright’s Law of more than 18%, and low to negative free cash flow. This makes sense too. Tesla will find it tough to scale revenue when production doubles quickly, and revenue plus costs drop by 20% or more each double. Higher interest costs reduce the value of future free cash flow. If you couple that with lower or negative free cash flow rates, this is problematic for future value. When I looked at scenarios where Wright’s Law was above 18% cost reduction, interest rates were above 5%, and there were low free cash flow rates as a percentage of revenue (below 9.5%), I got $73.6 billion in value. (Don’t kill me! That’s what the model says, not what I believe.)
The difference in current value and average model value has to be made up of higher EV and EGS growth rates than modeled, higher than expected FCF rates, and lower than expected Wright’s Law and interest rates. The first seems a lock, based on Q1 2021 production numbers, and the other three factors are unknown and will take time to be known. The big wildcard would be expected autonomy and outside revenue, which is what I plan to model next. This is currently a large difference from current value, which gave me pause. Is much of the growth from autonomy and high EV growth baked in already? FSD 9.0 Beta renewed my optimism that perhaps autonomy is closer than we thought. Once we get the Q1 financials, I’ll weight the numbers based on my model average and the actual numbers, run the model again, and then do so once more with autonomy baked in.
You can see the numbers from my model on GitHub here, under Tesla model.xlsm.
Let me know your thoughts below. I can take it. I appreciate your honesty of how much of a noob I am in data modeling and with GitHub.
Note: Author purchased shares on 4/12/2021, nothing here is financial advice, more entertainment than anything. Please due your own due diligence with a proper financial advisor before making any investments, etc., etc.
*GitHub, not Git. I’m learning.
Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.
Sign up for our daily newsletter for 15 new cleantech stories a day. Or sign up for our weekly one if daily is too frequent.
CleanTechnica uses affiliate links. See our policy here.
CleanTechnica's Comment Policy
