Tesla has a long history of “in-sourcing” (in-housing) production. It started with the idea of following the Apple model. That meant doing the design itself, finding factories which could make it, having no production itself — only design, marketing, and sales. That didn’t work — nobody did know how to make an electric car.
Tesla then switched to making the battery electric drivetrain itself, buying a complete car without a drivetrain from Lotus. This became the Tesla Roadster.
With the Model S, the company in-housed more, building the powertrain and body of the car and doing the assembly itself. The Model X had seats that were a bit special and hard to make, so Tesla in-housed that too. There are many more examples.
There are mostly three reasons for in-housing production, and it seems Tesla routinely faced a combination of all three.
1. Suppliers don’t know how to do it, can’t reach the required quality.
2. Suppliers are too expensive.
3. Suppliers can’t reach the required volumes.
In the end, the price:quality equation is better from in-housing much Tesla production. Tesla gets a better product for a lower price.
Why is this important when looking at the UBS teardown? It’s important because of the way UBS was guessing prices for parts made in-house, especially the drivetrain parts.
“We note that this cost estimate is based on our engineering partners’ expertise and channel checks at the time the teardown was performed.”
So, what would the channel charge for “military-grade electronics?” They would probably charge scary military-grade prices, including those famous military-grade profits. With this method of estimating prices, it is inevitable that the more Tesla produces in-house to save costs, the bigger the losses are that UBS will detect using those sky-high external prices.
After this intro, let’s look at some of the UBS discoveries.
“While Tesla’s powertrain was better than peers in terms of cost per kWh and performance, their lead was not as large as we would have expected. The overall cost at $178/kWh is only ~6% better than their <$190/kWh estimate for the Model S/X provided in 2016. Moreover, the cell cost at $148/kWh is well above Tesla’s guidance of below $100/kWh ending 2018.“
It is clear what is happening here. UBS uses the prices of the suppliers that could not deliver. What is strange is that UBS put these numbers down in their report. When you find numbers that are so different from what they should be, all kinds of alarm bells should go of, with red lights flashing as well. The two most likely reasons for this discrepancy are: that Tesla executives are lying through their teeth about their costs — to the public and in their SEC filings — or the estimates are way off.
In the first case, alert the SEC. The SEC does not like fraud. In the second case, correct your estimates. Also, in that second case, in the unlikely situation where reality is very different from what was expected, you have a lot of explaining to do in order to show why your numbers are correct.
Also, talking about powertrain when you mean battery pack is sloppy — the cost of the battery is likely 2/3 of the cost of the powertrain. The report has many more of these kinds of inaccuracies
The Model S/X battery was produced at the Fremont factory at less than $190/kWh in 2016. The expectation was that volume production in Gigafactory 1 could cut 30% from that price. After the start of production in Nevada, that reduction estimate was upped to 35% ($123.50/kWh). That is $55/kWh lower than the UBS production estimate. This number is probably still too high, if you believe CEO & CTO claims that Tesla expects to cross the $100/kWh barrier at the cell level sometime this year, and at the battery pack level in 1 or 2 years.
In a few months, a new production line for batteries will be installed in Nevada, around the time of the planned start of production of the Model 3 base version. It should enable better batteries at a lower cost — they keep learning how to do it better, and implementing what they learned. That UBS did not see this coming is perhaps forgivable, since predictions are hard, especially when the future is involved.
The numbers UBS uses for its profitability estimates are not yesterday’s numbers during the ramp up — they are the numbers they expect to be true for the Model 3 when normal, steady production is reached. Other teardown groups used 10,000/week for that state. UBS does not provide a number, but it seems like the most logical number to use. Next, I’ll look at the UBS numbers as considered for a steady production of 10,000/week.
What went wrong with the battery costs estimates also probably went wrong with the seats, power electronics, inverter, battery management system (BMS), thermal management system, electric motor, and a few other parts that Tesla has in-housed because external parties were too expensive or not good enough, or both.
When we look at the profit and cost breakdown of UBS, there are a few peculiar things that catch the eye.
In the second column, there is Depreciation & Amortization (D&A) combined with warranty and freight, together ~$5,725. The warranty is 2.5%, or $1,225 for a $49,000 car. But freight is a mandatory option — if you don’t include it in the revenue ($49,000 is without freight), you can’t include the costs in the costs of revenue, which is the first $1,000 (2%) correction.
That leaves ~$3,500 for D&A. But CTO Deepak Ahuja was very clear when he said that D&A at 5,000/week is below $2,000/car. To go from 5,000/week to 10,000/week, there will be some extra capex needed. While there have been no numbers mentioned, the repeated discussions stress that it is just a fraction of the capex to build the 5,000/week line. The two journalists who visited Fremont recently mention the same in a report in the Financial Times. A D&A of $1,350 per car is probably high, but I don’t like to be too optimistic. That is the figure I will use. That’s another big correction of $2,150.
The third and fourth column are opex. They have no relation to the car’s production volume. Therefore, it is wrong to assign an amount per car. Beside that, UBS has literally no idea what the R&D and SG&A are going to be in the coming years. They are mostly discretionary amounts, and budgets can be allocated depending on the intended profit margins. That is the second reason not to use opex to compute the margin.
When we compare the UBS margin with the other margins we’ve seen — from Tesla itself, Sandy Munro, German engineers at Wirtschaftswoche — it does not look so bad at all. The Germans calculated only labor and parts. All the overhead (factory, warranty, D&A, etc.) that creep into the cost of revenue has to be added. The gross margin definition of Sandy Munro appears to be close to the way Tesla computes it. Because Musk did like the German numbers a lot, I have used them to break down the Tesla cost, but the gross margin of Tesla is the targeted 25% and Tesla the company becomes profitable at 15% gross margin for the Model 3 Long Range (LR). The UBS margin of 18% before any corrections is clearly no reason for panic.
UBS thinks the margin on the Model 3 LR + PUP is 18%. Corrected for their freight mistake, that is 20%, or $9,800.
UBS calculates the margin loss going to the SR at $9,530. That is the combination of no LR battery at $5,350 margin and no PUP at $4,180 margin. That’s right, the bill of goods for the PUP is about $820 according to UBS.
Even with the inflated prices from UBS, it still does make sense to produce the base Model 3 at $35,000. A few customers ordering different wheels, colors, or EAP/FSD clearly make it profitable. And the higher volume helps to pay the overhead of D&A, R&D, and SG&A. All three of these costs are independent of production volume — at a higher volume, there is more gross profit to pay them.
But how much sense does the Model 3 make when we correct the UBS numbers?
First, Short Range (SR) battery production is about $3,000 lower than UBS thinks. Second, D&A is about $2,150 lower than UBS thinks.
Without looking at all of the other drivetrain parts that were more expensive than even UBS thought logical, we have a $5,420 margin on the SR without any options — that is a gross margin of over 15%.
When we apply the same corrections on the LR version, we get a gross margin of over $16,000, or 32.5%.
But reality is a bitch. I think Tesla is right to focus on getting the gross margin to 25%. There are always new challenges on the production line.
As a final observation, UBS did compare the Model 3 LR with a BMW in the same price class. BMW had a 2% higher margin, according to UBS. But BMW sells wholesale and Tesla sells retail. Those margins are over different prices. And we just added over 14% to the Tesla margin, giving Tesla a 12% higher margin over a larger amount. I think the Model 3 should worry BMW a great deal.