Tesla communication meeting, Tao Lin revealed for the first time: latest details on FSD entering China, Optimus mass production, AI training center, etc

On February 6th, Tesla Vice President Tao Lin elaborated on the company's strategic planning and business layout for 2026 at an exchange meeting held in Beijing, covering several key topics including the company's capital expenditures and investment priorities, the introduction of autonomous driving in China, energy storage expansion, humanoid robot mass production, and market positioning in China. For Tesla, 2025 is a challenging turning point year. According to Tesla's Q4 2025 financial report, the company's revenue and sales experienced declines and fluctuations, and it faced some business adjustments. Tesla CEO Elon Musk stated during the earnings call that production of the Model S and Model X would be halted in the second quarter of 2026. In response to this, Tao Lin said, "The sales fluctuations for Tesla in 2025 are mainly due to capacity impacts from product transitions, rather than a decline in competitiveness. The core path for the company over the next 3-5 years is clearly directed towards cutting-edge fields such as AI, autonomous driving, and robotics." "2026 is significant for Tesla," Tao Lin said, "We are undergoing a transformative shift focused on AI and energy." In the subsequent discussion, Tao Lin further introduced Tesla's latest progress and plans in several core business directions. The core points are as follows: Tesla's FSD does not have a clear rollout timeline in China; Musk's earlier mention of "February" primarily refers to markets like Europe, with China being "subsequently followed." Tesla's driver assistance data does not need to be sent abroad; computing centers and training are conducted locally in China, strictly adhering to data compliance requirements. Tesla has established its own AI training center in China to deploy localized training capabilities, preparing for larger-scale applications in the future. Tesla expects Robotaxi to be implemented in China within five years, but will not rush to pursue the number of urban coverage and order scale. The main bottleneck in enhancing the capabilities of the Optimus humanoid robot is not in the training data, but in whether the hardware possesses the foundational capabilities for training, especially in key structures such as dexterous hands and joint coordination. Maintain calm regarding "parameter competition" such as 800V and 5C fast charging. Overall experience is superior to a single metric; currently, a 15-minute charge providing 300 kilometers meets the needs of the vast majority of scenarios. Below are selected Q&A highlights from Tao Lin, adjusted without changing the original meaning: Tesla's Investment Focus for 2026 Question: Recently, did Musk's team come to China to inspect the supply chain related to Tesla? Tao Lin: That report should not be about Tesla; it may be related to SpaceX or other projects. Regarding the supply chain, we adopt a global procurement strategy. As long as a local supply chain can meet the needs for cost, stability, delivery speed, and overall optimal conditions, we will choose to procure from there. Question: Tesla's capital expenditure in 2026 is expected to exceed $20 billion. How will this funding be allocated? How will you balance short-term returns with long-term planning? Tao Lin: Regarding the specific allocation of this $20 billion, it has not yet been detailed down to individual projects, but the overall investment direction is already very clear, mainly focused on six areas: First, the advancement of Cybertruck mass production. The core production line construction at the U.S. factory has been basically completed, with most of the initial investment completed by 2025. Entering 2026, we will continue to invest to ensure the smooth realization of large-scale mass production of Cybertruck. Second, the construction of AI computing power centers, which is the most critical investment direction. The training center located in Texas, USA, has accumulated investments exceeding $10 billion, and significant additional investments will be made in 2026. This center will support all AI-related applications of Tesla, including autonomous driving and robot model training, and will serve as the computational foundation for a unified "world model," catering to global markets including China. Third, the transformation and upgrading of robotic factories. We have initiated upgrades to the Model S/X production lines, and in 2026, we will advance larger-scale transformation projects, aiming to have mass production capabilities for Optimus robots by the end of the year. Fourth, the expansion of energy storage business. We will continue to increase investments in energy storage manufacturing to further enhance overall capacity and delivery capabilities to meet the rapidly growing global energy demand. Fifth, the upgrade of the global manufacturing system. The focus here is to simultaneously enhance hardware automation levels and software capabilities, making the entire manufacturing system more efficient, intelligent, and capable of stronger scalable replication. Sixth, the continuous construction and opening of the charging network. We will continue to expand the coverage of the charging network and gradually open it to more car manufacturers. In the future, the charging network will not only serve as a refueling facility but is also expected to become an important component of a new energy network. Overall, this investment reflects Tesla's firm determination to transform towards "AI + energy infrastructure." In the AI era, before the large-scale application truly explodes, foundational infrastructures such as computing power, manufacturing, and energy must be laid out in advance, which is key to determining long-term competitiveness. Question: How much of this $20 billion capital expenditure will be invested in the Chinese market? Tao Lin: This investment indeed includes a portion aimed at the Chinese market, but the specific proportion has not yet been disclosed publicly, and relevant planning is still ongoing. Overall, we will continue to increase our investment in China, and the direction will remain consistent with our global strategy, still focusing on the two core areas of energy and AI, including the continuous construction and upgrading of software and hardware capabilities. Currently, we have deployed localized training capabilities in China to support autonomous driving assistance and various AI applications in the Chinese market. Although the current scale of AI users in China is still in a relatively early stage, the relevant training system has already been laid out in advance and put into operation, preparing for larger-scale applications in the future. Progress on FSD landing in China Question: Is there a plan for Model 3 to enter the Chinese market? Tao Lin: So far, there is no such plan, and our focus is not on this for the time being. Question: How do you view the competitive environment of the Chinese automotive industry? Tao Lin: China is one of the most vibrant and innovative automotive markets in the world. Both consumers and the industrial ecosystem demonstrate very strong creativity. There are many new force car companies in China willing to invest in this extremely challenging industry. Many founders already have stable successes but still choose to enter high-risk areas, which is very commendable. The value of Tesla, on one hand, proves that this path is feasible; on the other hand, it accelerates the entire industry's progress by continuously promoting technological advancements. As long as we continue to develop in China, we will definitely increase our investment and set important goals to be achieved in China. Question: What is the status of FSD's implementation in China? Elon Musk mentioned a timeline that it might be launched in February? Tao Lin: We will actively participate in the promotion of automated driving assistance work in China. In the United States, the development of FSD is very rapid, with a large user base and mileage serving as verification of technological reliability. The official website updates data daily, and the cumulative mileage has exceeded 7.5 billion miles, approximately 12.1 billion kilometers. Regarding specific timelines, we cannot provide a clear answer at the moment. The "February" mentioned by Musk mainly refers to potential progress in Europe, with China "following up subsequently," but this does not mean that China will also launch in February. Currently, all work is actively progressing, but there is no publicly available timetable. Question: The domestic L3 (automated driving assistance) is advancing quickly, with many car companies promoting regional pilots in cooperation with local governments. Will this accelerate Tesla's actions? Tao Lin: Tesla will actively participate in the development of the driving assistance field in China. Although it has not been officially launched yet, our training and adaptation for the Chinese market have been ongoing. Once it can be showcased, it will be in the best condition. In fact, all driving assistance systems on the market are still in the early stages, and it is not yet possible to clearly distinguish the advantages and disadvantages of different systems. It may take another year or two for the strengths and weaknesses to become clearer. Question: Recently, the government issued related restrictions on data export. Will this affect FSD? Tao Lin: This is unrelated to driving assistance. Our driving assistance does not require data to be exported because our computing power centers and training are conducted locally. Question: Will FSD be opened to other car companies in the future? Tao Lin: Tesla has always adhered to an open philosophy. Just as the charging network has been opened to other brands, we are also willing to open FSD capabilities to more car companies in the future. Of course, this requires certain collaborative development, and currently, our focus is mainly on improving our own system, so we have not formally initiated this, but it is a clear long-term direction. We believe that in the future, every car should have automated driving capabilities. However, not every car company has the ability to develop a complete system from scratch, which requires huge funding, a large team, and massive data Tesla's "world model" can be more efficiently adapted to different vehicles, helping the entire industry enter the autonomous driving era faster. The significance of FSD entering China may, to some extent, be similar to the role of the Shanghai Gigafactory in promoting the new energy vehicle industry back in the day. Question: In 2026, Tesla seems to be shifting from car sales to AI and robotics. What is the core path for the next 3-5 years? Tao Lin: The phase fluctuations in global sales in 2025 are mainly due to capacity impacts from product transitions, such as the switch to the new version of Model Y. Tesla has always operated on a "no inventory" model, where capacity essentially determines sales. The peak in sales in 2023 was due to stable products and full production line operations. Therefore, changes in sales do not entirely represent a decline in product competitiveness; rather, they reflect supply changes due to adjustments in production rhythm. From a longer-term perspective, our focus is continuously extending towards AI, autonomous driving, and robotics, which is the core development path for the company in the coming years. About the Competition of Robotaxi Question: What is the timeline and business model for the rollout of Robotaxi in China? Tao Lin: Currently, there is no clear timeline that can be announced. Once there are confirmed developments, we will communicate them externally as soon as possible. As for the operational model, whether it will be independently operated or co-built with partners has not yet been finalized. Question: From an overall progress perspective, how much longer will it take for Robotaxi to truly land in China? Tao Lin: From a technical perspective, Robotaxi is already operating in the U.S. and is unsupervised. So technically, Robotaxi is capable of providing services to the public in the U.S. However, in China, this is not just a technical issue; it also involves legal regulations, insurance, and various other factors. Personally, I expect it to be achievable within five years. Currently, in the U.S., it is also operating in limited areas, such as certain regions in Austin and Silicon Valley, where users can call for these unsupervised Robotaxis. This will definitely be a gradual process. Question: Why is Waymo already operating in six cities with 400,000 orders per week, while Tesla's progress seems slower? Tao Lin: We could actually expand to many cities, but we are taking a more cautious approach. We carefully test in each location, continuously adjust, and ensure absolute safety before scaling up. It's like a child learning to walk; whether they walk at 10 months or 11 months is not important. Right now, it doesn't matter who has more cities or more orders. Once the technology is truly mature, the rollout speed will be very fast. Just like a child who has just learned to walk must ensure they can walk steadily in each new place; once they have mastered it completely, they can walk anywhere freely. As long as all scenarios are validated without issues, the promotion will be quick Question: How do you view the competition with companies like Baidu in the Robotaxi field? Tao Lin: At this stage, the number of cities covered is not the most critical metric. These advancements may bring some first-mover advantages, but we do not focus on that. More importantly, we never simply understand this as a zero-sum competition. The entire automotive and mobility market is still far from saturated, and there is a vast space. To give an analogy: just like people can learn to drive at different driving schools, as long as they can safely get on the road in the end, there is essentially no issue of needing to have "only one" provider. Similarly, as long as more vehicles can achieve safe driving, we do not believe that all cars in the future must come from Tesla. From a global perspective, currently, no single company's vehicle scale exceeds a few million, while the total number of vehicles in the world is an extremely large figure. At this stage, the industry needs to jointly promote technological maturity and safety implementation, rather than prematurely entering fierce market distribution. Therefore, it is still too early to discuss who has more or less, who wins or loses. Question: Recently, He Xiaopeng stated that he believes current autonomous driving technology has jumped directly from L2 advanced assistance to the "quasi-L4 安心 stage." How do you view He Xiaopeng's perspective of "skipping L3 and directly moving to L4"? Tao Lin: Regarding the differentiation of autonomous driving levels, L2 has a clear definition, but the definitions of L3 and L4 are actually quite vague. We differentiate more from the dimensions of "human supervision" and "no human supervision." When there is human supervision, the responsibility always lies with the driver; when there is no human supervision, the responsibility lies more with the car company. Therefore, the specific names of the levels are not very important; the key is the division of responsibility and technological capability. Technical Challenges and Mass Production Planning of Optimus Robots Question: Tesla plans to transform the Model S/X factory for robot production and has set a goal of shipping 1 million units by 2027. How will this goal be achieved? Tao Lin: At the current stage, the most critical aspect of humanoid robots is not the scale of mass production, but whether the product itself is truly mature. If we compare it to human growth, the entire industry is still in the "embryonic stage," and there is still a significant distance to true large-scale application. The core technological breakthroughs at this stage mainly focus on two aspects: First, it is the dexterity of the hands. Over 70% of human work relies on hand functions, and the ability for fine hand operations is the core of humanoid robots. Some robots on the market can perform simple actions, but they are still far from high-precision operations like threading a needle. Second, it is the sensitivity and coordination ability of the joints throughout the body. Only when all joints approach the flexibility of humans can robots truly enter the stage of large-scale mass production. Our third-generation Optimus will achieve significant improvements compared to the second generation. Once the functional design matures, mass production itself will not become the biggest challenge. From the perspective of global potential demand, 1 million units is not a particularly large number; the real key is whether robots can effectively replace humans in completing work What we are building is a general-purpose humanoid robot that does not need to distinguish between industrial or household scenarios. The same robot can switch between different scenarios through training, and essentially the hardware system is unified. Question: How to solve the problem of insufficient training data in the training process of humanoid robots? Tao Lin: At this stage, the core bottleneck is not entirely about data or training itself, but more about whether the hardware capabilities have reached a "trainable" basic level. Take the dexterous hand as an example. If the physical structure cannot achieve joint flexibility and tactile feedback close to that of a human hand, then no matter how much training is done, it is difficult to truly enhance capabilities. In the future, we will certainly enter a large-scale data training phase, including various sources such as virtual data and artificial data. But at the current stage, the most critical aspect is still the breakthrough in hardware-level details. Only when the hardware capabilities are in place can the value of data and training be truly released. Question: How will Optimus cooperate with the Chinese robotics industry chain? What is the current level of localization? Tao Lin: Currently, Optimus mainly relies on its self-developed system, with core components sourced globally. As for the specific localization rate, it can only be accurately assessed once the product enters a stable mass production phase; it is still too early to discuss this issue now. Our principle is very clear: we will prioritize regions that have a comprehensive advantage in cost, stability, and supply capability. If the Chinese supply chain performs best in these aspects, we will certainly prioritize adopting Chinese solutions. From a more macro perspective, the entire robotics industry chain is still in the "initial construction" stage, somewhat similar to the early automotive industry. At that time, there was also no mature and complete supply system, and it needed to be gradually integrated and improved from a more fundamental level. Once humanoid robots truly achieve large-scale mass production, it will become an important milestone for the global robotics industry, indicating that a complete supply chain has begun to form and has the capability to support the continuous development of more robotic products. Tesla Builds Local AI Computing Center in China Question: Does Tesla have its own AI training capabilities in China? How is localized data training conducted? Tao Lin: For the assisted driving and AI application scenarios in China, Tesla has independently invested in and is using an AI training center in China to deploy Tesla AI's local training capabilities. Question: What is the scale of computing power at the training center in China? Tao Lin: We do not calculate based on a specific number of cards; the training center can meet our training needs in China. Since the chips used are not exactly the same as those in the U.S., there is no direct comparability. Tao Lin: Is the training center self-built or rented? Tao Lin: It is completely self-built; there is no need to rent. We also built it ourselves in the U.S. Question: What is the source of data used by the computing center? Tao Lin: We do not collect data from car owners. All data-related matters are strictly executed in accordance with Chinese laws and regulations. If the laws and regulations allow us to collect data, we will; if not, we will not collect it. But in either case, we can conduct local training Question: Under data compliance requirements, is Tesla considering using Chinese data for overseas model training or collaborative training? Tao Lin: We have not yet activated these functions because there is currently no significant necessity. Our models are already world-class and trained; we only need to make some localized adjustments in China. Moreover, many adjustments do not necessarily require collecting actual road data from China, such as road signs and turning rules, as this information is already available. Question: Is the training of data completely reliant on video clips found online? Tao Lin: We will comprehensively utilize various data sources, but we do not necessarily need to collect a large amount of actual road data. We can ensure the effectiveness of the training. About Semi Truck and Cybertruck Question: What is the plan for the construction and mass production timeline of the Semi truck? Tao Lin: Currently, the Semi factory is officially scheduled to start construction in 2026. From a product positioning perspective, the Semi is not only a brand new commercial vehicle but will also become an important flagship product for Tesla in the fields of electrification and autonomous driving. Question: Why is Tesla prioritizing the Semi over other potential products? Tao Lin: The company's resources are always limited, and we must proceed according to priorities, addressing key issues one by one. While there are many directions we could pursue, what is truly important is to start with projects that have the most strategic value and can bring long-term impact. Under this judgment, the path represented by the Semi, which combines electric freight and autonomous driving, has a very clear priority. Question: Did the Semi truck have autonomous driving capabilities from the beginning? Tao Lin: The Semi was designed from the outset with autonomous driving scenarios in mind. Compared to passenger vehicles, the freight sector is more standardized in terms of operational routes, usage scenarios, and business models, making it one of the most suitable scenarios for the early large-scale implementation of autonomous driving technology. Question: Will the Cybertruck include features for automated logistics delivery? Tao Lin: These plans are currently mainly targeted at the U.S. market, primarily to enhance the cargo capacity of the Cybertruck. We have a Cybertruck at our Shanghai factory, and we often use it for hauling goods; it works very well. Question: Is the design of the Cybertruck suitable for hauling cargo? Tao Lin: It can haul cargo; although its load capacity is not as high as that of the Semi truck, it is already a good cargo capacity for a small pickup in the U.S. Question: Will the Cybertruck enter the Chinese market? Tao Lin: It is still too early to discuss this, but we certainly hope it can enter the Chinese market. We will wait until it is officially mass-produced and has been operating in the U.S. for a while before making a decision. There is still a lot of space in the Chinese truck market. Charging Technology and Battery System Planning Question: What is your view on advanced technologies such as 5C fast charging and 800V high-voltage platforms that have been implemented by domestic car companies? What are Tesla's plans? Tao Lin: These technologies will continue to evolve, and a significant portion of our capital investment is also allocated to enhancing infrastructure capabilities, including the ongoing construction of charging systems. Taking the 800V platform as an example, we have conducted thorough internal assessments. Even if the vehicle itself supports 800V, if the charging network cannot fully match it, what ultimately determines the user experience is still the amount of electricity actually charged in a given time. Currently, with the support of the supercharging network, Tesla vehicles can replenish approximately 300 kilometers of range in just 15 minutes, which is sufficient to meet the vast majority of practical usage scenarios. Therefore, the demand for higher voltage platforms is not particularly urgent. In terms of technological reserves, we always maintain cutting-edge exploration. For example, the 48V low-voltage system used in the Cybertruck is a breakthrough design. However, whether a technology can truly be implemented depends on whether it can effectively enhance the user experience in a specific market environment. Our core principle is not to pursue leadership in a single parameter but to ensure the overall optimal performance of the entire system in real-world usage. If a vehicle has an 800V architecture but the charging network cannot support it, then users are essentially paying for a capability that is difficult to use. What truly determines the charging experience is the comprehensive result of various factors such as speed, stability, reliability, network coverage, and management capabilities, rather than a single isolated technical indicator. Risk Warning and Disclaimer The market carries risks, and investments should be made cautiously. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk