Today was Tesla’s Q4 2021 earnings call for investors. This morning Elon and the team spoke about the successes in ramping production and deliveries last year and the resulting financial rewards that come as a result.
As impressive as the numbers are, 87% growth in deliveries, $17.6B in cash/equivalents on hand, and 30.6% GAAP Automotive gross margin in Q4, things got really interesting when Elon provided a product roadmap update.
While Tesla has successfully navigated the global chip and supply chain shortages to date, it sounds like that’ll continue to be a major battle through this year and into 2023.
Cybertruck / Semi
Tesla will not introduce any new vehicles to their lineup this year, as the company continues to ramp production of existing vehicles (primarily Model Y), along with starting and scaling production at the two new Gigafactories – Berlin and Texas.
Model Y was the real success story from 2021 and will continue to be this year, with the mid-sized SUV market thirsty for an EV with a decent range (and hopefully a decent price).
Sadly this means the much-awaited Cybertruck, with more than a million pre-orders, will now arrive sometime in 2023, along with Semi, and maybe Roadster. This revised timeline is interesting, given the competition is launching vehicles in the Ute/Truck segment.
There are already Rivian R1T rolling off the production line (albeit in small numbers), Ford’s F150 Lightning is proving successful with them recently doubling production estimates and you also have the Chevy Silverado and lesser so (due to its crazy price), the GM Hummer EV.
I’m running a Twitter poll that asks if those with Tesla Cybertruck orders are prepared to wait or are considering alternatives. At the time of writing, 80% say they’re happy to wait, while 20% responded Yes, meaning they’re looking at an alternative.
4680 Battery Packs
Back in September 2020, Tesla held their Battery Day event, where they announced they were entering the battery production space, with a new form factor, the 4680. This new cell would provide higher performance and range, at a lower cost.
One of the biggest questions that we had is how this would arrive to consumers. Would Tesla add less of the cells, to achieve the same or similar range, while capturing more profits from the vehicle, or keep a similar capacity pack, delivering more range for a cheaper price to the customer.
While we don’t have any adjustments to the ordering page of the Tesla website to confirm which direction Tesla will go, Musk did confirm that cars (Model Ys) will be leaving the factory in Q1 2022, with the 4680 cells.
At the time they announced the new cells, they explained that they would combine this new cell with a structural pack and when combined with the front and rear single-piece casting, you have many benefits in both production, but also a reduction in weight, increased rigidity which should correlate to better handling.
Of course, these 4680 cells were also slated to go into the Cybertruck and Semi, so now they’ve been pushed to 2023, there’ll be a lot more available to ship in Model Y this year. There is no word on plans to add them to the Performance Model 3 at this stage.
In a surprising move, Musk announced plans to prioritise the Tesla robot, known as Optimus internally. The robot was announced during AI Day back in August last year and at the time we were told we’d see a prototype this year.
The concept here is that Tesla can leverage their skills and expertise in understanding the real world thanks to FSD, that they’ll be able to leverage that object detection, path planning etc, to build a robot to take care of manual labour.
The bot would use cameras like the car and also use the FSD HW3/4 computer to perform operations in the real world. For many of us, it’s hard to understand how well, how human-like this could be, so seeing the prototype is critical in setting realistic expectations.
Today we had our expectations of 5-10years reset, as Musk announced this would be a priority for Tesla. During a response to a question, he also revealed that they will first be deployed in Tesla factories, before selling externally. This dogfooding approach is common in technology companies and not that surprising.
I would still expect a prototype update sometime in 2022, likely Q4. There is no question that interacting in the 3D-space is an incredibly difficult challenge and the volume of objects it would need to understand to be useful is enormous. Also, the training interface for skill development is equally important and will there be a store to buy new skills, or can the bot learn by watching a human complete the task.
While many have spent hours discussing how impactful this Battery Day vehicle will be, today we were told it is not a priority. On the surface at least, it seems a great, cheaper electric vehicle would be one of the most important projects for the company, to execute on their vision, but it seems that is actually a number of years away now.
If Tesla can succeed with the TeslaBot, it may be a key component in reducing the cost of building future vehicles. The Tesla compact car which was speculated to be US$25,000 could be 10-15% smaller than the size of the Model 3, but needs to drop close to 30% of the price. Current Model 3s have gone up in price and if that continues, reaching that price gets even more difficult.
Even with the best battery updates, and the raw material cost reduction, Tesla needs to get creative to reach that price point any time soon.
One of the most interesting parts of Tesla, is their work on autonomous vehicles. Their Full Self Driving Package has been in development for years and each year, it seems Elon Musk declares this is the year we’ll reach level 4 (or sometimes 5) autonomy.
During today’s earnings call, Musk provided what was the strongest conviction about the timeline to date. He said that he would be shocked if it wasn’t that capable by the end of this year. While it’s hard to pin down how exponential the development will happen, we also are not well-positioned to make time estimates that far out. There is clearly a significant to-do list in front of Tesla’s Autopilot team to complete FSD to the point where you don’t need to pay attention.
Based on what we see from the FSD Beta testers in the US (today we learned there are now 60,000 of them), the system is absolutely advancing. We often look to Elon to more accurately predict the timeline for FSD, understanding that he gets to experience the latest software in his car, and work with the team directly on the roadmap.
There is no question an autonomous vehicle creates an amazing opportunity for increased autonomy for society and those with mobility issues, particularly if the cost of riding in a robotaxi drops considerably compared to an Uber. The challenge is, solving the technology piece, the getting regulatory approval.
Tesla’s dedicated chip architecture for their AI training was shown off at AI Day and we knew then, the path to implementation had to pass a specific benchmark. That was to be better (aka faster) than running the training on their existing GPU cluster.
Musk confirmed today, Dojo is not ready yet. At best, Dojo may be active sometime later this year and if it performs as they expect, it would enable faster iterations of the AI models that are tested and then deployed to customers cars.
Importantly we did get confirmation from Elon that Dojo is not required to reach level 4 with FSD, it would simply speed up the training of the neural nets for faster learning.