Next-Generation Model 2 To Launch Early? (Tesla 1Q24 Earnings Call Summary)
Below is the summary of Tesla's first quarter 2024 earnings conference call. For the financial report analysis, please refer to " The Contribution of FSD, Who Still Says Tesla is "Paper Mache"? 》
I. Review of Core Financial Information:
II. Detailed Content of the Earnings Conference Call
2.1. Key Points from Executive Statements:
- Business Highlights:
① Energy Business: In energy storage deployment, Megapack sales reached a historic high, driving record profits in the energy business. It is expected that this growth trend will continue and may exceed the growth rate of the automotive business.
- Technological Progress:
① FSD: In the FSD Beta V12, we have made very good progress. Soon, we will see 3 million vehicles starting to use the FSD Beta, which is purely AI-based.
We see continued improvement in our reasoning capabilities. All these vehicles have very strong reasoning power.
In North America, we have further actions for Hardware Version 3, and soon Hardware Version 3 will be expanded to approximately 1.8 million vehicles. So far, about half of the owners are using it, and this ratio is increasing every week. The total mileage of FSD Beta 12 is now 300 billion miles.
We have adjusted prices in the United States, with the subscription price adjusted to $99, and the selling price also adjusted accordingly, both lower than before.
② Robotaxi: We also announced earlier that we will showcase our Robotaxi autonomous taxi in August, which is a very important move. The Robotaxi can form a very good network.
③ Model 2: Tesla is accelerating the launch of new models, planning to advance to early 2025 or even possibly later this year. These new models will utilize the technology of the new and current platforms and will be produced on existing production lines.
④ Significant progress has been made in artificial intelligence training capabilities, no longer limited by training capabilities. Approximately 35,000 H100 computing units have been installed and activated, and it is expected to reach 85,000 units by the end of the year.
- Financial Performance:
For the automotive business, we did see a decrease in revenue compared to the previous quarter, with seasonal factors being very apparent. Additionally, we also saw a slight decrease in automotive profit margin, from 18.9% to 18.5%.
Apart from the impact of the cybertruck, in terms of pricing actions, we actually need to consider more the reduction in unit costs and the offsetting of automotive component revenue recognition that the U.S. did not have certain functions before. That is, we have achieved a reduction in unit costs, and now we need to further confirm some of the revenue from certain componentsThere is a mutual offset between this and the impact of the cybertruck.
In addition, the Model 3 has been further improved at the Fremont factory, with upgrades to the Model 3 at the Fremont factory. It was also mentioned that there was an arson attack in Berlin causing production interruptions. This may lead to a slight increase in costs in the short term, but to a large extent, such cost increases are offset by our cost reduction efforts. If we do not consider the impact of the cybertruck and the upgrade of the Model 3 at the Fremont factory, our car profit margin and revenue have improved. We have also introduced the Autopark function. Earlier, I mentioned that the car profit margin and revenue have improved, without considering the positive impact of the introduction of Autopark. Furthermore, we see that at the Austin and Berlin factories, the cost of each Model Y is very close to the level we have achieved in Fremont. This is a very good sign that we are able to reduce costs without sacrificing quality. This is attributed to the improvement in business efficiency and the strengthening of execution by the entire team.
We also see that other complete vehicle manufacturers are gradually reducing their investments in electric vehicles, and their reliance on credit scores is actually very strong. This means that our current strategy is in stark contrast to theirs. Overall, our sources of income are stable, and we see the entire automotive industry moving away from electrification, which is not the future we want. We are more willing to contribute to the future of the industry. We have taken many measures, such as reducing the subscription and purchase prices of FSD. In addition, in the United States, we have introduced very attractive leasing options for the Model 3, and you can check the relevant prices in our financial report. We expect to continue to achieve better results on this path.
In terms of finance, we see that the free cash flow in the first quarter of 2024 is -25 billion US dollars.
In terms of AI, we have more investments, and we expect future investments to increase further. The number of our employees has actually decreased, and the amount saved may exceed 1 billion US dollars. We are very concerned about capital efficiency and hope to further improve the capital expenditure efficiency of different business segments, including the energy storage business.
2.2, Q&A Analyst Q&A
Q: How is the situation with the 4680 battery? What is the current production capacity?
A: The production capacity of the 4680 battery increased by about 18%-20% from the fourth quarter of last year. Cybertruck production exceeds 1k per week, equivalent to about 7GWh per year. We expect battery production to continue to outpace Cybertruck's capacity in the second quarter as we expand the third production line of the first phase of four production lines and maintain several weeks of battery inventory. COGS continues to decrease as line yield improves and production increases. Our expected goal is to lower costs to even lower levels by the end of the yearQ: What is the current status of Optimus? Are they currently executing any factory tasks? When do you expect to start mass production?
A: We are able to perform some factory tasks in the laboratory. We expect to start limited production of Optimus before the end of this year and carry out some useful tasks. It may be possible to sell to the public by the end of next year.
I believe the value of Optimus will exceed the sum of all other robots, because if you have an intelligent humanoid robot that can navigate and perform tasks in reality, there is no limit to the economies of scale. Tesla's AI reasoning efficiency far surpasses that of other companies, which will bring returns in many ways.
Q: What is Tesla's current assessment of the regulatory approval path for unsupervised full self-driving (FSD) in the United States? How should we consider appropriate safety standards compared to human drivers?
A: Several states have already adopted laws for autonomous vehicles, laying the foundation for the operation of unmanned driving vehicles. These states and the data we provide will pave the way for broad regulatory approval in the United States, and also in other countries. Other autonomous vehicle companies have made regulatory breakthroughs, which also help us. They have been operating in San Francisco for some time and have obtained approval from the city of Los Angeles, and these approvals are happening rapidly. I think it's hard to ignore as long as we have large-scale data and the data clearly shows that the accident rate of autonomous vehicles is only half that of human-driven vehicles. You don't even need to think about it, just like taking an elevator, it will take you to your destination without considering how the elevator works.
Tesla will manage the entire fleet, similar to the operating model of Airbnb and Uber. Some vehicles are owned and operated by Tesla, while others are owned by end users. Users can add or remove their vehicles from the fleet at any time and set usage permissions, such as allowing friends and family to use them or restricting use to five-star users or anyone. As the fleet grows, we will accumulate a large amount of data and continue to train and optimize. Therefore, as our fleet grows from 7 million vehicles globally to 9 million vehicles, and eventually to 10 million vehicles, there will be a continuous feedback loop.
The success of Google Search is built on its strong search and feedback loop mechanism, where users continuously provide data through searches and clicks, allowing Google to continuously optimize search results. Similarly, our autonomous driving technology is also advancing on a similar path.
We firmly believe that we have strong capabilities in reasoning and computing power. Currently, our vehicles are equipped with Hardware 3.0, and we are actively advancing the design work for Hardware 4.0, with Hardware 5.0 almost ready, expected to be deployed on vehicles by the end of next year.
We are well aware that cars cause resource waste when not in use. Therefore, we expect cars to be utilized more fully over time. This is similar to the distributed computing achieved by the AWS cloud platform, where users can submit requests to the cloud platform from anywhereThe network formed by our fleet is very powerful.
We need a large amount of computing power to train AI models. Considering the large scale of our fleet, the number of vehicles may reach 100 million, with each vehicle having very strong inference computing capabilities on average. This means that we have approximately 100 GW of distributed inference computing power globally. Integrating 100 GW of inference computing power together is very difficult, even in autonomous computing. However, this distributed method allows for the stacking of these computing capabilities. The utilization rate of future vehicles will increase significantly, not just 10 hours per week, but 50 hours per week, allowing vehicles to deliver greater value, whether in travel or inference computing, thus enabling more efficient use of resources.
Q: How do you view the process and safety issues of autonomous driving?
A: We have multiple levels to verify safety. We train hundreds of neural networks every week to improve safety by replaying key events from millions of trips. Additionally, we use simulation systems for closed-loop testing. Once verified, we hand it over to quality assurance drivers, collect feedback, and then provide it to external customers. We also have real-time monitoring dashboards to track key events, continuously improve the quality and safety of versions. Failure data is fed back into training to automatically enhance model performance without human intervention.
We have some understanding of the situation in the next three to four months because we have advanced models, and the capabilities of vehicles will see significant improvements, but there will also be some minor issues to address. We can observe model scaling losses, where increasing model size may lead to performance improvements, along with other scaling methods including increasing training data, training computation, and improving architecture. By combining these scaling methods, we can roughly infer future performance. These experiments take time, but based on past trends, we can predict progress in the future.
Q: Can you reveal the official timeline for the $25,000 model?
A: We are accelerating the launch of a low-cost model, utilizing existing and open capacity to achieve higher capital efficiency. This is our mission, to deliver the most affordable model to customers as soon as possible. We will discuss this further on August 8 before spending high capital on other things, utilizing all our capacity with marginal capital expenditure.
Tesla's main goal is to achieve autonomous driving and provide services to the vast fleet. When unsupervised fully autonomous driving is achieved, it will be one of the biggest events in asset value growth in history.
Q: How is the production progress of the Cybertruck?
A: Several weeks ago, Cybertruck production reached 1,000 vehicles per week, achieved within four to five months since the end of last year's SOP. Despite our costs decreasing, mass production still faces many challenges due to various new technologies and supplier constraints. We will continue to focus on cost-effectiveness and quality to drive production growth this yearQ: Which traditional car manufacturers have contacted Tesla to discuss the issue of FSD authorization opening?
A: We are in talks with a major large-scale car manufacturer on this matter.
Q: What is the timeline for the scale of Semi?
A: We are finalizing the engineering design of Semi, achieving ultra-high efficiency and large-scale production based on the experience we have gained from pilot fleets. At the same time, we have started production at the Reno, Nevada factory and plan to launch the first batch of vehicles by the end of 2025, with external customers starting to purchase in 2026.
Q: Can we transfer FSD permanently for free until achieving full Level 5 autonomous driving (currently limited-time transfer policy)?
A: No.
Q: How is the production growth in Lathrop? What capacity do you expect to reach by the end of the year?
A: The production plan in Lathrop is proceeding as expected. We have activated the second general assembly line, increasing capacity from the current 20GWh to 40GWh by the end of the year. Due to the long project sales cycle, we can understand the order situation 12 to 24 months ahead of the delivery date, allowing us to plan production in advance to meet the growth in orders. This enables us to gradually increase factory capacity according to business and order growth needs.
Q: Elon, you are currently leading many important companies. Could you share your current interests and whether you will reduce your involvement in Tesla in the next three years?
A: Tesla occupies most of my working time, and I work almost every day. There are very few Sunday afternoons when I take a break. I ensure the prosperity of Tesla, and it will continue to be so in the future.
Q: The growth rate expectation for 2024 is significantly lower than 2023. What is the likelihood of the growth rate exceeding zero? Does this statement imply that sales may decline?
A: No, I believe our sales this year will be higher than last year.
Q: Regarding future products, if you successfully achieve the next generation of cheaper and better vehicles, such as priced at $25,000, with a range of 300 miles, and more aggressive manufacturing technology, how long do you think your top Chinese competitors will take to replicate such a vehicle?
A: Actually, we don't know what competitors are doing, but we know we are doing very well, and we can stand unbeaten in the competition. There are indeed some complete vehicle manufacturers in China doing very well. We hope to do our business well from our perspective, rather than just looking at what others are doing. Look, some competitors in China have seen a decline in sales, compare their decline with ours. Our decline is actually less than theirs, so from this perspective, it proves that we are doing better than them.
We should not simply view Tesla as a car company or a complete vehicle manufacturer, but as an AI company, a robotics company. Essentially, if we only look at our automotive business, it is wrong to treat our business frameworkTesla places great emphasis on autonomous driving, but our business is also very extensive. Think about what we just mentioned - if a car is only used for 10 hours per week and 1.5 hours per day, it is a waste. If we can significantly increase the utilization rate, the change we bring to transportation will be revolutionary. Take a look at what improvements the FSD Beta version can bring, it gets better every day, so give it a try. The latest 12th version is gradually improving from the previous versions, and future versions will be even better. Yes, what we care most about is what we are doing and how we are improving. Let me reiterate, Tesla should not be seen solely as an automaker. We have various business segments including AI, robotaxi, energy storage, etc., which are all integral parts of our business. We will also further invest in technology.
Q: Elon, you have mentioned that you want to have 25% of the company's voting control. Do you have any mechanisms in place to ensure that level of voting control? Any additional comments on this topic?
A: I believe that no matter what happens, Tesla will solve the problem of autonomous driving. Even if the progress may be slower, at least the issue of autonomous driving will be resolved.
On the Optimus side, there are other aspects of future product research and development. I believe Tesla will make significant progress and ultimately succeed. Even if vehicles eventually disappear, we will have more products, more future-oriented products, a whole set of products for the future. We are really optimistic, even a bit naive, on the Optimus side. But we think this optimism, even naivety, is necessary for us to better understand our future business. Personally, I am more focused on Tesla's current and future business. As for the 25% voting control, I believe there are corresponding approval processes for this, so I won't elaborate further on this.
By the way, if the company generates a significant amount of positive cash flow, we can consider stock buybacks.
Q: Can you quantify the specific impact of OpEx reduction and discuss more about the qualitative impact of these layoffs on the business? What specific activities will be abandoned due to the layoffs?
A: We have made comprehensive personnel adjustments to align with future growth stages. It's like pruning a tree, we are pruning. Through this process, we will become stronger and more capable of facing future challenges. From 2019 until now, we have experienced long-term prosperity and have not given up on anything important. Instead, we are reorganizing the company to adapt to the next growth stage. Just as humans start from a single cell and gradually develop into different individuals, companies also need to reorganize as they evolve through growth stages to ensure growth.
Q: Can you provide more details on the progress of the FDF license you mentioned today? What product milestones does Tesla need to achieve in order to successfully reach a licensing agreement?A: I think we only need to further develop a better FSD according to our ideas and roadmap. Currently, we are working on version 12.3. It is obvious that our solution for enabling automatic driving only requires relatively low-cost computers and standard cameras, without the need for lidar, radar, or ultrasonic sensors.
Let them use the same cameras and our licensed software. It may be somewhat difficult to directly promote this because their vehicles do not have the same hardware as ours, so there are still some issues to overcome and resolve regarding FSD licensing.
Launching a smartphone, possibly using iOS or Android, many people in the early days did not understand why a smartphone should be designed this way. It seemed like a brick with no great functionality, but what are smartphones like now? It seems that no one goes back to using non-smartphones, and the design of smartphones is not like a brick, no one makes it a flip phone. To promote a smartphone design like this, including hardware and software, it is necessary to ensure that more companies have similar hardware forms. In the process of authorizing FSD to other automakers, we need to consider whether we can gain some inspiration from the example of smartphones just mentioned.
In the future, non-smart cars will be difficult to sell. Therefore, the use of FSD technology will become more widespread. We have had discussions with some automakers and noticed that they spend a lot of time on the product lifecycle, and their pace is very slow. As their development pace takes several years, we have significant differences with them, so we may reach some FSD licensing agreements with them to adapt to their development pace, allowing them to truly benefit from our software capabilities in product integration and integration. These agreements will take effect in two to three years to match their progress.
We do not need major design changes. Like the example of smartphones mentioned earlier, implementing FSD only requires maintaining a basic similar form, which requires installing cameras and relying on computer vision technology. Next, regarding Tesla's future pricing strategy, the reduction in car prices is driving demand growth. From a regulatory perspective, regulatory credits and offset policies will continue to be provided, and Tesla also offers leasing discounts.
Q: I would like to understand Tesla's future pricing plans. Previously, the company said that lowering prices can increase demand. Now that car prices are already very affordable, especially considering that some vehicles can enjoy IRA tax deductions and leasing discounts provided by Tesla. Do you think it makes sense to further reduce prices for existing products? Can the company further reduce prices while maintaining positive annual free cash flow?
A: I believe we can maintain a fairly large free cash flow, and our cost reductions can offset price reductions. Our goal is not to achieve profits by lowering prices, but to pass on the benefits to customers. Selling excellent products at discounted prices will bring excellent sales. We must ensure that the product continues to improve cost-effectiveness, and the price is acceptable to people. Therefore, we need to continuously improve affordability, make prices more attractive, constantly enhance product value, and improve production costs. The new Model 3 is an excellent car, and we have added a lot of value to it, while you can lease it for as low as $299 per monthQ: Can you help us understand the schedule for the launch of FSD in other regions? Could you also explain your recent comments on China?
A: Considering there are many untapped markets in various regions globally, we need to evaluate which markets to enter. Once the decision to enter a new market is made, what measures will we take? We need to analyze the market and whether the end-to-end neural network-based autonomous driving technology can bring greater value and be smoothly deployed in these new markets. Our existing technology and capabilities perform well in almost all markets, and in most cases, no modifications are needed when expanding to new markets. However, the bigger challenge lies in regulatory requirements. Whenever we release a supervised version of FSD in any market, we need to obtain the corresponding regulatory approvals, and this applies to the Chinese market as well. For example, the basic principles of driving are universal globally, and regardless of the country, drivers must adhere to local traffic rules and driving requirements. Therefore, our technology is applicable to different markets, but attention must be paid to regulatory differences in each country, such as local driving qualifications, license exchanges, etc. Hence, when promoting FSD in China, we must also comply with local regulatory requirements.
Q: I would like to know if you think the supply constraints mentioned in the release have affected the results, could you help us quantify this? Is this one of the reasons you are confident in unit growth in 2024?
A: Seasonal factors are significant, and the macroeconomic conditions remain subdued. In addition, our factory in Berlin was subjected to an arson attack, and upgrades are being made to the Model 3 at the Fremont factory. These can be seen as short-term constraints. However, we remain confident in the performance for the second quarter and will address these challenges one by one. We will continue to comprehensively analyze market conditions and the demand for different products, such as whether there is excess demand for Model 3 or Model Y, and whether we can quickly meet these demands. Furthermore, we are committed to streamlining operational processes and avoiding unnecessary complexity. Over the past week, we have significantly simplified the sales process. Our goal is to make the Tesla purchasing process as simple and fast as possible, ideally completing the purchase within one minute.
Q: Considering Tesla's pursuit of becoming a leader in real-world artificial intelligence, could you provide a detailed explanation of the specific applications of distributed reasoning technology in the deployment and use of vehicles? Additionally, what new possibilities can be unlocked by vehicles through distributed reasoning capabilities?
A: This question involves multiple aspects, and I will focus on providing an answer. As we have mentioned before, our cars are not just transportation vehicles but machines with wheels. It is expected that the utilization efficiency of these vehicles will significantly increase in the future, with the weekly usage time potentially increasing from the current approximately 10 hours to 50 hours. This means that about one-third of the time the vehicles will be in operation during a week, while the rest of the time will be used for charging, cleaning, and maintenance. In the future, cars will be able to handle various workloads. In terms of large language models, we can utilize the computing power of the fleet to do more work, unlocking more possibilitiesThis requires a very large neural network scale, and we need to further expand. At the same time, we need a large amount of computation to handle the workload. We believe that these vehicles can play a greater value in a distributed system. Considering people's investment in these high-priced vehicles, we hope to maximize the use of their computing power to ensure that the time for vehicle operation, charging, cleaning, and maintenance adds up to fill all possible time, thereby improving efficiency.
Furthermore, we hope to enhance the value of AI and computing power through a distributed design similar to Amazon Web Services, achieving better commercialization. If we can connect millions or even more vehicles into a complete network, their distributed computing power will be very strong. By properly planning the idle time of vehicles to engage in meaningful activities, if the scale of vehicles reaches 100 million, we can achieve almost unlimited possibilities. By then, the computing power of vehicles may reach hardware version 6.0 or even 7.0, estimated to reach a computing power level of 1000GW, or even in the future through hardware upgrades, achieve higher levels of computing power.
In any case, cars need to become smarter, which means they need powerful onboard computing power. Considering that cars cannot be driving at every moment, we plan to fully utilize these computing resources during the idle time of vehicles to maximize the utility of smart vehicles. In addition, we have explored the deployment potential of artificial intelligence and related workloads in a distributed system.
Compared to smartphones, our vehicles have extremely high computing power, not in the same order of magnitude as smartphones. The battery capacity of vehicles may be 60 kWh, which can provide a large amount of power support when the vehicle is not running, helping to perform more calculations, which is an important foundation. In terms of distributed power access and cooling, we can explore the application of battery packs in new technologies and business models using current and future technologies.
Q: Regarding the large-scale production of 4680 batteries, can you talk about how close you are to the target output, and when you may start accelerating the expansion of production capacity for this technology?
A: We are very pleased with the progress of the 4680 battery cells. By the end of this year, we expect to further increase the production capacity of 4680 battery cells. The production situation of 4680 battery cells is expected to exceed supply capacity, as sometimes suppliers are unable to keep up with supply, indicating a significant improvement in our production speed. Currently, we are particularly focused on the production capacity increase related to Cybertruck. We have extensive cooperation with suppliers, and they have also made significant investments. By the end of the year, we expect suppliers to face challenges in meeting our demand for 4680 battery cell production, although we hope they can keep up. Supply continuity is crucial for us, so we will do more calculations internally and inform suppliers of possible situations.
One important reason for Tesla's internal sales of 4680 batteries is to counteract possible situations with suppliers. Because every large car manufacturer is ordering a large number of batteries, causing the price of lithium-ion batteries per kilowatt-hour to soar. With internal battery production, we have a countermeasure against demand shocks. This is our way of thinking, and we are doing the battery project to address the crazy increase in supplier costs per kilowatt-hourQ: At last year's Investor Day, you mentioned that the per unit cost of the next generation vehicles will be reduced by 50% compared to the current Model 3 and Model Y. I believe this implies a cost of around $20,000. About one-third of this comes from improvements in the manufacturing process. But I'm wondering if you see other factors, such as reduced powertrain costs or material cost savings, being widely applied in some of the new products you are currently discussing?
A: For us, a car is not just a standalone entity, it consists of multiple subsystems such as drive, powertrain, automation units, battery, heating system, seats, interior components, and DR controllers. We plan to further analyze the costs of these components to determine which costs can be transferred. Our goal is to optimize the costs of these components and apply these optimizations to the entire product line. Based on years of engineering experience, we hope to fully utilize existing, already identified components to produce future cars.
Q: I'm curious, given the rationalization of your current automotive manufacturing plans, is there an opportunity to convert and sell 4680 batteries to other car manufacturers to create additional revenue streams? What are your thoughts on this?
A: In the current market situation, many complete vehicle manufacturers are experiencing a significant decline in order volume. From our perspective, many suppliers currently have excess capacity issues, and we hope that their capacity can be absorbed in the future. Elon has previously mentioned the supply chain issues of 4680 cells, which can actually help us solve many battery supply chain issues. The demand for 4680 cells will be very high in the future, and we will reach more deals with suppliers, consuming more of their capacity. We hope to be more involved in the upstream of the supply chain and communicate and collaborate more with suppliers.
In terms of overall pricing, we will also consider what the upstream supply is like, and if there are any fluctuations, whether we need the necessary buffers. Batteries will have corresponding cycles, with boom and bust periods. Whether production exceeds supply or supply exceeds production may change back and forth, but overall we will see a positive trend.
There are corresponding government acts, such as the inflation reduction act in the United States, which brings additional complexity to us. We will combine such acts to look at the supply of batteries and where our entire production activities should go.
Risk Disclosure and Statement of this article: Dolphin Investment Research Disclaimer and General Disclosure
