In my article yesterday, I broke down what Tesla’s Q2 production numbers looked like in a worst case and best case scenario. Now, I’m going to share what I think those numbers will actually be, and why.
A Note On That Article
I asked people to point out anything I missed, and Taylor Marks made a solid point that not every vehicle delivered in Q2 will be delivered in Q2, and suggested revising my worst case numbers down by 10,000. Indeed, there is always some difficulty getting a certain number of vehicles to the right place fast enough, and there is always a decent number of vehicles unsold or not delivered because they are in transit or just in the wrong place.
In a steady-state world, these vehicles would match in volume to vehicles in the previous quarter with which the same thing happened. This quarter is especially hard to determine, though, since increased production capacity at the end of the quarter would make it harder to deliver everything. I think Taylor’s 10,000 number is too high, but I can see an argument for revising down my worst case scenario of 66,161 for it. Thanks, Taylor!
Explaining Theoretical Capacity
Capacity is a fascinating piece of nearly any business, and there seems to be a lack of understanding about exactly what it means. In one of my previous jobs, I was an analyst focused specifically on looking at capacity, and trying to determine how we could get as close to a theoretical capacity rate (or run rate) as possible.
Theoretical capacity rate is when you figure out the absolute most of something that you can produce using a particular process. It assumes that whatever we are doing will completely meet capacity at all times. This is completely unrealistic.
I want to point out that my best-case scenario of 139,434 deliveries isn’t the theoretical capacity. Theoretical capacity is likely much higher.
To explain the difference, let’s say that you’re in line for a new roller coaster (and there isn’t a pandemic). We’ll use Superman as our example:
There are 32 seats per train. For argument’s sake, let’s say that you can load and send people on each train once a minute. This gives you a theoretical hourly capacity of 1,920 people per hour.
There are three problems, though:
- It’s extremely unlikely that every train will be sent with every seat filled. In fact, I zoomed in on this photo I took years ago and found an empty seat in the first, third, fourth, sixth, and seventh rows, meaning that there were only 27 people on this train. I have no idea if this is a normal amount of riders per train or not, but let’s assume that it is. Instead of a theoretical run rate of 1,920 people per hour, having 5 empty seats per train — which doesn’t seem like much — actually drops our capacity by 300 to 1620, or a utilization rate now of 84.3%.
- It’s extremely unlikely that you could send a train once every minute. To do so would mean you would never have a delay in someone getting in their seat, with the restraints working, ensuring their heights are properly checked, putting their shoes on, or whatever else. If you’ve been to an amusement park in the last 10 years, it seems there is always someone delaying something. Let’s say it takes just an extra 15 seconds to send a train — which based on my latest Six Flags experiences would be super generous. Instead of sending 60 trains per hour, you can now send only 48, for a utilization rate of 80%.
- Sometimes, things happen and the ride stops and everyone waits for someone in the maintenance department to come fix it. I can’t guess what percentage would cause a downtime like this, but it would skew the hour or hours of downtime significantly. Let’s assume over time this averages out to the ride breaking down and needing a repair for only 2% of the time. This gives us a utilization rate of 98% — although, this type of downtime is obviously the most disruptive at any single point in time, since it reduces capacity to 0% for the period it happens.
Here’s the problem: when we combine these different delays, it gives us the actual capacity compared to the theoretical, and they combine to be way off. Using the examples above, the ride would see 27 riders per train and 48 trains per hour, or 1296 riders per hour — a utilization rate of just 67.5%. Take away another 2% from downtime, and you would expect 1270 riders per hour on average, or a utilization rate just shy of ⅔ of the theoretical capacity.
I illustrate this not because roller coasters have anything to do with Tesla, but because problems which may seem extremely minor can affect capacity far more than most people realize, and the “real world” capacities of many systems can be massively different than their theoretical capacities. By themselves, a few empty seats and a small delay in sending the train out can seem like pretty minor issues, but it adds up quickly.
You can do things to mitigate your capacity issues, however. For example, on a roller coaster, maybe there is someone at the front of the line checking your height, or telling you where to put loose articles. Some parks use single-rider lines, so they can quickly fill the empty spots on the train in with people. This allows you to increase your “best case” for a capacity to be as close to theoretical capacity as possible.
I use roller coasters in this example because it’s a simple way to illustrate capacity challenges that was as close to possible like a factory setting but many people could relate to. A lot of “people businesses” — like buses, for instance — introduce the complexity of daily demand and you could get me to write another 2500 word article just on them, but that’s not why we’re here.
Why Are We Talking About This?
So, anyway, I bring this up for a few reasons:
First, it’s important to understand how seemingly small changes can drastically impact production. If Tesla Shanghai is operating at a 429 cars per day capacity, raising output by 1% only results in 4 additional cars per day. That number is so small that it might seem like it’s worth overlooking.
But a 1% increase sustained for a 91 day quarter, would give us an additional 364 cars. At an average sales price of $40,000, that is an increase of $14.56 million for the quarter — nearly equal to Tesla’s $16 million profit in Q1 of 2020.
Second, we don’t have enough information to really determine what a theoretical capacity would be. To find it, we’d probably need to find the best day for each individual automobile assembly line and then add them together. Best case can be impacted by a lot of different issues, and it’s something you can aim to get as close to as possible, but it’s impossible to pull off.
My best case scenario numbers are not impossible to pull off. To be clear, I think they are far higher than what reality will show, but they would be within the realm of what could actually be achieved if everything went perfectly.
My Best Case / Worst Case Numbers
From my earlier article, doing all the math I did, my best case scenario was 125,162 vehicles produced and 139,434 vehicles delivered, and my worst-case was 51,889 vehicles produced and 66,161 vehicles delivered.
Next, I’m going to start looking for evidence to see what the actual capacity has been. I expect it’s somewhere between these two numbers. If I come up with a number above the best case numbers, I need to really figure out what happened, and if I come up with a number below the worst case, I’d want to check the assumptions I made to come across my worst case.
My Evidence — US (& European) Production & Deliveries
It’s extremely hard to find much evidence right now about US operations and deliveries because Tesla doesn’t share things until a few days after the quarter ends. I did take my car in for service in late May, and I was pleasantly surprised to see multiple Model Ys rolling past me onto the lot of the Tesla store where my service was being done. I asked the service technician if I could look at them, and he said I was free to walk around them and they had a “huge amount” that were being shipped there for delivery starting the day I was there. By the time I left, about 25 Model Ys had arrived in their lot, and I could see two more carrier trucks arriving.
I bring this up mostly to dismiss it as evidence. I’m not close to California, and if you’ve watched how Tesla delivers vehicles for any length of time, you’re aware that they generally ship cars further away from California sooner than later, and save their California market until the end of quarter, where they can deliver cars practically as they come off the factory floor. Even in my worst-case scenario, 25 cars is a drop in the overall bucket, less than 1/2,000th of the 66,161 deliveries numbers I found in my worst case.
A lot of people and even articles will often find one thing like this and use it as evidence of a company’s health. Don’t fall for that.
There is a bit I was able to learn that is slightly useful, however. First, the location confirmed this was the first shipment they received since before the pandemic shut the factory down. Secondly, the manufacturing date on the vehicles in the lot was May 19, meaning it was just 8 days after the reopen that they were produced. My only takeaways were production had definitely resumed, and locations were definitely starting to receive vehicles.
My Fremont production range was therefore 34,300 to 82,600 vehicles. Using the idea that multiple things can affect capacity, and my only data being production and shipping seems to have resumed at a normal-ish rate when Tesla said that it did, I feel pretty comfortable assuming that Tesla produced at ⅔ of its standard capacity over this time, resulting in 55,067 cars produced in Fremont in Q2.
Based on the (usual) evidence of shifting timelines and cars at the end of quarter, I assume that the majority of these would have been delivered. Note that while I’d dismiss a few shifting timelines once, like my Model Y experience above, since there is evidence of this every quarter, I think we can assume it’s happening.
There was also a surplus of 14,272 cars produced in Q1 that were not delivered due to the pandemic. Let’s use the same ⅔ deliver ratio I used above, which gives me an additional 9,515 delivered vehicles. This gives us a total of 64,582 deliveries resulting from Fremont in Q2.
(And, note, I put Europe in these numbers since it gets its cars from the same factory.)
My Evidence — Chinese Production & Deliveries
Here’s where things get more interesting. We have some delivery data from China! Unfortunately, our first bit of data wasn’t too good, as Reuters reported that Tesla registrations fell 64% in April compared to March, down to 4,633 units. Shanghai-produced registrations were at 3,635. (Note: The 1,000 remaining cars were surely all in the 14,272 transport number, so I’m going to ignore those as I counted them above.)
The article I just linked to is important, though, for something else — it notes that Tesla expected to make 4,000 units a week in June. That number is higher than anything I used in even my best-case scenarios in my last article.
The other thing is I don’t know how Tesla moves capacity around in China. With how Tesla pushes all the deliveries to happen by the end of the quarter, perhaps some of the weakness in this number could be attributed to Tesla moving around inventory like it does in the US.
In May, things got better. Reuters reported that Tesla sold a record 11,095 Model 3 vehicles in May, all of them being made in Shanghai. I’d also like to bring back up that the factory in Tesla was shut down for the first 9 days in May for upgrades, making this number even more impressive than it appears on it’s face.
I estimated Tesla Shanghai’s production at 17,589 to 42,562 for the quarter. Considering deliveries from Shanghai for the first two months were 14,730 alone, it would take a historically disastrous final month to be at the minimum number.
Instead, I think it’s reasonable to assume that May was more representative of actual sales, and that there is at least some increased production, as well as some push to deliver as many of the produced vehicles as possible, like we see in the US. Let’s assume the first 9 days of May moved possible May sales down by 25%, and didn’t impact June. This would give us 13,869 June deliveries. I’m going to add another 10% based on the usual end-of-quarter rush I expect to see happen in China too, which adds 1,387 vehicles to that. This gives me a total of 29,986 deliveries in China. I’m going to back the 1,000 number out from Fremont, meaning I’m going to peg production at 28,986 for the quarter.
(Note: as I explained at the beginning of the article, they clearly won’t deliver every car they made on June 30th to customers, but I also believe that this will be offset by earlier production and will ultimately be a wash.)
I’m looking at Fremont for production of 55,067 vehicles, with previous supply leading to 64,582 deliveries. Shanghai I’m estimating production and deliveries of 28,986, not including the 1,000 vehicles that came from Fremont.
This gives me my own estimate of 84,053 vehicles produced, and deliveries of 93,568 vehicles, far above the consensus on Wall Street of 60,000 to 70,000 deliveries. In fact, since I started writing this, I saw that Credit Suisse’s Dan Levy sent a note to investors that Tesla could see a delivery number between 90,000 and 100,000 vehicles. His numbers — expecting 31,000 from China, 35,000 from Fremont, and 25,000 from inventory — confuse me a bit because I don’t get where the extra 10,000 vehicles in inventory will come from, but his production numbers — which were 76,000 to 88,000 — are pretty close to mine.
There is further evidence that I’m being super conservative on China’s production. Here’s an article from Ran Finance Studio in March (check out Google Translate if you want to read it) noting that production capacity was expected to reach 3,500 per week that month, and will reach 5,000 in July. If this is true, theoretical max capacity of 3,500 per week for the quarter would result in 45,500 vehicles produced. The run rate of 4,000 per week for June mentioned by Tesla would make that number 47,500. That’s an upside of nearly 64% above what I estimated. I think there is a lot of potential for my numbers to be low.
And here’s the thing — Tesla China has a lot of room to grow that seems to go unnoticed. The same article notes that Tesla China has been accelerating the localized part contracts, and it was estimated 40%+ of the car was already sourced in China, moving to 70% by the middle of the year and 100% by the end. They note that the goal is to continue to reduce prices and promote sales while maintaining a stable gross margin. They anticipate the price of a Chinese Model 3 could be reduced to 230,000 yuan, or around $32,500, a significant decrease that will lead to increased demand.
Another Capacity Tangent
Back to the theoretical capacity for a moment — I highlighted theoretical capacity to show how small issues can lead to significant disruptions. What I didn’t highlight is that as you add the ability for additional capacity that operates independently of the earlier capacity, the less disruptive each problem becomes. Amusement parks don’t just have a single ride, they have dozens, meaning if one ride spends a full hour down for a maintenance issue, the park capacity — or ability for a guest to go on a ride — is not significantly impacted.
Tesla Shanghai’s ramp, if it is anywhere near what Tesla has said it was going to be — and at this point, why would be doubt it? — would have an extremely positively affect on the company’s capacity. If Chinese demand hasn’t collapsed — and I see no evidence that it has — Tesla Shanghai is going to give us all one hell of a ride.
I started this quarter by noting that it seems like almost every business should just have a mulligan for Q2. Maybe Tesla doesn’t need it. If my delivery estimates are even close to correct, and I believe they will be, I would be more surprised to see a loss for the quarter than a profit. (In fact, as I was writing this, an email from Elon leaked that stated that “breaking even is looking super tight.” And to add to the reporting by CNBC, which stated it wasn’t clear if he was talking about profits or something else, I’m certain he was talking about profits.) In times of economic strife, things often accelerate toward new technology and markets, and it seems that we’re seeing that now.
Oh, and in case it wasn’t clear, this is my work. I spelled out how I did it all, but please feel free to disagree. I’m certain it won’t be exactly right, and as I also tried to highlight, small changes in assumptions can lead to large discrepancies in totals.
My actual vehicles produced guess is 84,053, and I’d guess that we will see a real number between 80,000–92,000. Similarly, I feel that my delivery number of 93,568 would translate to a range of somewhere between 89,000–101,000 deliveries.
Longtime readers of my articles will know I don’t like just using numbers by themselves, so to translate this into the way I think, it shows me that Tesla’s business is far more resilient and better positioned than I had guessed.
What do you think Tesla’s production and delivery estimates will be?
I am a Tesla [NASDAQ:TSLA] shareholder who has purchased shares within the preceding 12 months. Research I do for articles, including this article, may compel me to increase or decrease stock positions. However, I will not do so within 48 hours after any article is published in which I discuss matters that I feel may materially affect stock price. I do not believe that my voice could or should influence stock price by itself, and I strongly caution anyone against using my work as your sole data point to choose to invest or divest in any company. My articles are my opinion, which was formulated using research based on publicly available data. However, my research or conclusions may be incorrect.
Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.
Former Tesla Battery Expert Leading Lyten Into New Lithium-Sulfur Battery Era — Podcast:
I don't like paywalls. You don't like paywalls. Who likes paywalls? Here at CleanTechnica, we implemented a limited paywall for a while, but it always felt wrong — and it was always tough to decide what we should put behind there. In theory, your most exclusive and best content goes behind a paywall. But then fewer people read it! We just don't like paywalls, and so we've decided to ditch ours. Unfortunately, the media business is still a tough, cut-throat business with tiny margins. It's a never-ending Olympic challenge to stay above water or even perhaps — gasp — grow. So ...