Skip to content 
In this blog, we discuss the latest on Tesla’s strategy through the lens of its recent investor day presentation and its 2022 earnings.
Tesla’s first investor day
Last week, Tesla conducted its investor day presentation. If you just look at the mainstream media coverage, which is plentiful, all commented how the presentation was full of vision but lacked specific details about the business. And supposedly, this prompted a market sell-off in after-hours trading, causing the share price to fall by 7%.
After watching the 4-hour stream, I’m not sure I agree with the market. While it’s true that the company painted a grand vision, Tesla also showed tremendous technical progress that contributes directly to the bottom line. In this article, we will delve deeper into these achievements.
A message of hope
A common refrain in the solar energy community is how abundant it is compared to human needs. Whenever we wrote academic papers as graduate students, we always mentioned that there’s enough energy in the sunlight that hits the earth, in ~1.5 hours, to power human activities for one year. The amount of energy is so abundant that it would take only 0.2% of the earth’s land area to achieve a sustainable energy economy. So, the lack of adoption of solar energy is not due to lack of abundance but rather due to lack of infrastructure, cost competitiveness (although this is changing rapidly), and energy storage.
It is with this same message that Tesla opened the presentation.
“What we are trying to convey is a message of hope and optimism. Optimism that is based on actual physics and real calculation. It’s not wishful thinking. Earth can and will move to a sustainable energy economy and will do so in your lifetime”. — Elon Musk
How will Tesla help move us to a sustainable energy economy and replace fossil fuels? By tackling various points of the energy value chain (see below).
While it may sound like an impossible task, the company believes that it is achievable, with a reasonable capex, within our lifetime.
When we think about the electrification of the economy, we tend to think about the scarcity of minerals (for electric motors and batteries) and the mining needed. But in reality, Tesla argues that a sustainable economy involves less mineral extraction. This is because minerals in batteries are recyclable, creating a closed loop. In contrast, fossil fuel economy is linear: it involves drilling and processing said fuels and then burning them. Furthermore, Tesla’s history suggests mineral reserves always expand to meet demand (see below). Whenever prices go up high enough, supply expands. This is another argument against the mineral scarcity debate.
Relentless innovation
The remainder of the presentation details Tesla’s efforts in rethinking the automobile manufacturing processes from first principles – to reduce costs and increase throughput.
Take the Cybertruck, for example, which is finally coming this year. It has a stainless steel exterior. Stainless steel cannot be stamped into shape. As a result, the company had to design the truck using linear bending processes, which is not part of the current auto-manufacturing methodology. This is why the Cyber truck looks the way it is. Unlike other car companies, Tesla’s design team is vertically integrated with the manufacturing team. In other words, they develop the cars hand-in-hand.
Another example of rethinking established manufacturing practices is the redesign of car assembly. The image above shows the current way of assembling a vehicle. Henry Ford invented the assembly line more than 110 years ago, an innovation that enabled a new Model T to be produced every 90 minutes. The practice is still being used today, with major improvements in process and the addition of automation. However, a century later, optimization around the assembly line is almost at its end. Today, Toyota can produce a car with an internal combustion engine at a rate of 3 cars per second (1,800x faster than the Model T) while producing cars that are arguably 1,000x more complex.
Here’s a GIF of how Tesla’s cars are made using the traditional assembly line process. Notice how assembly follows the following order: assemble the metal frame and body panels, paint, take the doors out, then insert the interior, battery, etc., and then re-attach the doors again.
With the new paradigm shift to EV, design, and manufacturing can be rethought from the ground up. In the past two years, Tesla spearheaded two major innovations with the Model Y:
- Tesla introduced the largest high-pressure aluminum stamping machine created to stamp chassis parts, deleting hundreds of separate parts and making the assembly line faster (the company were inspired by how toy cars are made)
- Tesla made the battery part of the floor, which allows it to contribute to structural integrity. This saves weight and increases the speed of assembly
These, along with other changes, allowed Tesla to create an improved assembly process for its next-generation platform, which looks something like this.
In the GIF above, the future process enables the assembly of modules in parallel. With this process, Tesla expects to reduce operator density by 44% and increase space-time efficiency by 30%. This means that the future Gigafactories can be 40% smaller and 50% cheaper while maintaining similar or higher throughput.
The innovation does not occur just in the manufacturing process but occurs all throughout the cars as well. Despite looking the same (since its introduction in 2017), Model 3 cars have become cheaper to produce and more efficient.
Much of the power savings and weight reduction efforts in Tesla’s cars come from the power management unit. EVs have higher semiconductor requirements to manage the charging process, the conversion of DC to AC current, the powertrain, and many others. This means that compared to a traditional gas car, EVs require a more complex wiring system, specialized controllers, and more expensive silicon carbide chips (SiC). All these add to more weight, costs, and inefficiencies.
Tesla’s next-generation platform will significantly reduce these requirements by using a new drive unit, improved wirings, and a power controller system. Shares of SiC chip companies fell with this announcement.
The result of its relentless focus on efficiency is that Tesla’s cars are more efficient than its competitors. On a per-mile basis, Model Y is about 25% more efficient than Volkswagen’s ID.4.
Self-driving (FSD) update
Part of the vision of sustainability is to increase the utilization rates of cars. Most cars are parked during the day. If self-driving cars are a reality, then conceivably, utilization rates can be increased by orders or magnitude. Rather than have your car parked while you’re at work, it can continue to be used as part of a self-driving fleet.
There are three main parts of Tesla’s self-driving effort (Full Self Driving or FSD, which is a bit misleading, in my opinion).
The first is architecture. The FSD architecture is designed to create a general vision and planning system. It uses an eight-camera system that inputs to a video-based neural network, outputting a single 3D output.
The above is an example of the neural network in action navigating a complex left-hand turn while being surrounded by moving objects. The car has to be able to plan and execute within a 50 ms (0.005 s) compute time.
The second is data. The Tesla fleet not only captures visual data from all cars in the fleet but also auto labels the data, enabling them to reconstruct a 3D environment. All of these require a specialized supercomputer system. This is the third part, the supercomputer network, which manages and trains the AI. It comprises 14,000 GPUs (likely NVIDIA’s), split between auto-labeling video images and AI training. All this video data requires more than 30 petabytes of video cache (or 3 million GB), which is slated to grow to 200 petabytes in the near future.
AI + Data + Compute = a safer car. According to Tesla, vehicles with FSDs are getting into an accident 6x lower than non-FSD cars (one accident per 3.2 million miles driven for FSD vs. 0.5 million for the US average).
While this might be true, perception is different from reality. In reality, an accident that occurs on a car that has an autonomous driving capability degrades users’ trust in the entire fleet. If Citibank in New York City is robbed, you wouldn’t think all Citi branches are unsafe. But this is not true for Tesla cars. One Tesla car getting into an accident diminishes the trust that users have for the company. This is the rational thought process for human beings. After all, a single flaw in one car can be a manifestation of a systemwide flaw. This is why Tesla crashes always get so much media attention, even if they often are caused by driver error.
For some investors, the FSD and the robo-taxi fleet are part of the Tesla investment thesis. However, the full deployment of FSD has been delayed many times. In the short and medium term, it’s probably not wise to heavily index this as an investment thesis.
Exponential growth of physical manufacturing is hard
The relentless focus on efficiency is necessary because Tesla is trying to grow exponentially. Its long-term goal is to make 20 million cars per year. To put that into perspective, Tesla’s goal is double what Toyota makes today and 15x that of 2022’s total car production.
To do this, Tesla drives continuous process improvements in its Gigafactories. In both the new Shanghai facility and the more mature Freemont, CA, factory, the company can increase output while reducing labor input (see below).
Currently, Tesla has four factories, with a total capacity of about 2 million cars. It also announced a new facility to be built in Mexico.
In the recent past, Tesla has shown exponential growth in manufacturing output. It’s taking less and less time for the company to produce 1 million cars:
- 12 years to build the 1st million cars
- 18 months to build the 2nd million
- 11 months to get to the 3rd million
- 7 months to build the 4th million car, a milestone it achieved this month
All the improvements we outlined above helped Tesla to produce a new car every 45 seconds.
Battery Megapack
When analyzing Tesla’s business, I often wonder why it dabbles in so many different solar/electricity-related verticals. It’s as if building cars is not hard enough. In reality, Tesla is trying to bootstrap an entire electric energy ecosystem. To do that, it often has to integrate vertically.
This means it has to build its own lithium refining plant (to address supply issues), create its own software stack to simulate magnetic fields (to make more efficient motors), and develop its own battery technology and power electronics to manage the system. And because car manufacturing at the grand scale is a well-orchestrated continuous process (to make a car every 45 seconds), the input to production cannot be interrupted.
To ensure this, often, Tesla has to build excess capacity, especially for the battery and power electronics. But building excess capacity for the sake of it is capital inefficient. To better utilize the excess capacity, Tesla sells other things that, even though requiring the same input, can be produced and sold at a different cadence than the car production.
The Megapack is one such product (it also fits the mission). It is a grid-level battery installation to store electricity on a large scale. These are being used to replace gas-peaking power plants (power plants that are run only when demand is high and the base load is not sufficient). Megapacks are often paired with wind and solar energy. Currently, Tesla’s Megapacks are industry-leading. They have the highest energy density per unit area, at 300 MWh per acre. This is 2x higher than the energy density of a traditional gas-powered peaker plant.
These are not just dumb batteries in a box. Tesla has built software suites that made managing and operating these Megapacks easy.
During the stream, Elon mentioned an insatiable demand for Tesla’s megapack. He believes growth is constrained by production capacity. So far, the company is reporting a 65% CAGR of capacity deployed since 2016.
Cumulative charts can be deceiving (and VCs hate them…). To get a better sense of growth, let’s look at the quarterly deployment trend. Overall, the growth is very lumpy. Deployment seems to have slowed down in 2021 and peaked in the second half of 2022. Notice how the solar segment is flat (blue bar below).
Tesla’s next generation platform
The company did not give many details on its next generation EV platform other than it will be more efficient and cheaper to build. Its target is a 50% reduction in costs for the next-gen vehicle.
The demand for Tesla
Another concern that investors have about Tesla’s prospects is demand. As more and more car companies introduce their own EVs, will demand for Tesla decrease?
In the presentation, Elon and management commented that they had not seen any weakness in demand. Rather, Tesla was observing sales slowness due to affordability. Higher interest rates make financing more expensive, while at the same time, inflation is pushing costs up. In other words, demand for Tesla is constrained by affordability, not desire.
To address this, in late Q4 2022, the company cut prices (some of which have been reversed). The price reduction was then followed by a surge in demand. It seems that management is correct. There appears to be a high price elasticity for Tesla.
To see this trend in Tesla’s data, we can look at the total cars produced and compare the numbers to the total delivered. Delivery numbers represent when customers accept the vehicle. It is the point where revenue is recognized. When we look at the trend of the two (see Figure 2 below), we see a divergence between the two. Starting in the second half of 2022, Tesla produced more cars than it could sell. It could be that, as the global economy weakened, consumers started to cancel their orders. This might have prompted Tesla to cut prices in January 2023.
The grand vision
The long term objective for Tesla is to produce 20 million cars annually (a 15x increase from today’s capacity) and achieve 1 TWh annual storage production. This requires an additional $122 billion – $147 billion in additional investments, and it hopes to fund this by operating cash flow.
Here’s the non-GAAP annual trend of operating expenses as a percentage of revenue.
Over time, Teslas has achieved economies of scale that has allowed them to gain more leverage on operating expenses. It can produce more cars that are better and cheaper. If the company can maintain operating expenses at about 6-7% of revenue, it should be able to fund the required investment with its operating cash flow and achieve a 20 million car production rate by the end of the decade. It is a big if, however. The hope is that demand will continue to be strong, and the company can continue to make cheaper cars to address the affordability issue.
Tesla is often compared to Apple, because both companies sell high end hardware with tight software integration. But in reality, while Apple designs much of its system, everything is produced and assembled by external vendors. This is not true for Tesla. The company has tight integration between design and manufacturing. This integration is needed to push forward the frontier of efficiency for the new EV industry. This requires extreme discipline and high technical know-how. To me, a more apt comparison would be Tesla versus the old Intel.
Thanks for making it this far! For a in-depth discussion on Tesla’s strategy, please read a piece we previously published: Tesla Strategy Analysis.
