Tesla, Inc. (TSLA.US) Q3 conference call: looking forward to showcasing Optimus V3 in the first quarter of next year.

Recently, Tesla, Inc. (TSLA.US) held its 2025 fiscal third quarter earnings conference call. Tesla, Inc. believes that as artificial intelligence is brought into the real world, it is at a crucial turning point. Musk emphasized that Tesla, Inc. is the absolute leader in the real-world AI field, unmatched by anyone else. Tesla, Inc.'s cars have the highest "intelligence density" in all AI, and this advantage continues to strengthen. Key business developments include: the full-scale launch of fully autonomous driving (FSD) and Robotaxi, which will fundamentally change the form of transportation; the energy storage business (Powerwall, Megapack) through grid energy storage buffering to significantly improve energy output efficiency; and the Optimus humanoid Siasun Robot & Automation, which is expected to become the largest product in history, with human-like motion and interaction capabilities. Tesla, Inc. is at the starting point of a large-scale expansion of fully autonomous driving (FSD) and Robotaxi, which will fundamentally change the nature of transportation, with its momentum akin to a "shockwave." As Tesla, Inc. has a clear understanding of achieving unsupervised fully autonomous driving, Musk is confident in expanding Tesla, Inc.'s capacity to achieve the goal of "expanding future production as quickly as reasonably possible." Furthermore, Tesla, Inc. is making a significant impact in the energy sector through battery storage, especially the Megapack. The use of batteries for energy buffering can effectively double the energy output of the grid without the need to build new power plants. The company has announced the Megapack 4 plan, which will integrate the functions of substations, directly outputting a voltage of about 35 kilovolts, greatly enhancing deployment capabilities and speed. Optimus is on the verge of a major breakthrough, with the potential to become "the greatest product ever." The company has all the core elements needed to achieve it: real-world AI, excellent mechatronics engineering capabilities, and scalable production capabilities. Optimus V3 is expected to be showcased in the first quarter of next year, with a realism level that has never been seen before, appearing as "a person wearing a Siasun Robot & Automation suit." Q&A Q: What are the latest metrics for the Robotaxi? Fleet size, cumulative mileage, completed ride counts, intervention rate, and when will the safety driver be removed? What are the remaining obstacles to deploying unsupervised FSD to customer vehicles? A: In terms of Robotaxi business deployment, we are steadily progressing with the removal of safety drivers. The company expects to achieve fully driverless operations in most operational areas in Austin by the end of this year. We plan to achieve this goal in at least some areas of Austin in the coming months. Due to safety concerns, we are proceeding cautiously with deployment as any accident could become global headline news, hence we have chosen a more prudent approach. We plan to launch Robotaxi services in about 8-10 metropolitan areas by the end of the year. The specific expansion progress will depend on regulatory approvals in various regions, and we have submitted applications in Nevada, Florida, and Arizona, which are publicly accessible. In terms of actual operating data, our driverless vehicle fleet in Austin (without a safety driver in the driver's seat) has accumulated over 250,000 miles. In the Bay Area, due to regulatory requirements, we still have safety drivers in the driver's seat, but the mileage has exceeded over 1 million miles. It is worth emphasizing that the total mileage of customers using the supervised version of FSD has recently reached 6 billion miles, which is a significant milestone. We plan to gradually remove in-car safety drivers as planned, starting first in Austin. Q: In the current AI boom, what is the demand and backlog status for Mega Pack, Powerwall, CECEP Solar Energy, or energy storage systems? Does Tesla, Inc. plan to supply power to other large-scale data centers? A: The market demand for Megapack and Powerwall remains strong, with orders already booked into next year. Our Mega Block product has received positive feedback from customers and will start shipping from the Houston factory next year. With the advancement of AI technology, the application demand for energy storage products in the data center field has significantly increased. Large-scale data centers and utility companies are increasingly recognizing the value of Megapack in improving grid reliability and alleviating power supply pressures. In the U.S. market, policy changes are driving a surge in residential CECEP Solar Energy demand. We have introduced a new CECEP Solar Energy leasing product, and we expect this trend to continue until the first half of 2026. In the first half of 2026, a new leasing product will be launched, and residential CECEP Solar Energy panels produced at the Buffalo factory will be delivered in the first quarter of 2026, showcasing industry-leading aesthetic design and performance, underscoring Tesla, Inc.'s commitment to U.S. manufacturing. Q: What are the existing challenges in bringing Optimus to the market? A: Bringing Optimus to the market is an extremely challenging task. It is important to note that this is by no means an easy feat. Currently, Optimus is already capable of running 24/7 autonomously at our Palo Alto engineering headquarters, providing guidance services to visitors. However, I do not underestimate the difficulty involved. The biggest engineering challenge lies in the development of the hand and forearm. Creating a robotic hand as dextrous as a human hand is extremely difficult. The human hand is a sophisticated biological construct, and through in-depth research, we have found that the configuration of four fingers and a thumb, the different degrees of freedom of each finger, the varying muscle strength, and finger lengths have all evolved precisely. Therefore, the hand and forearm of Optimus (with its actuators primarily concentrated in the forearm area) present a massive mechatronics engineering challenge, with the complexity surpassing all other components of the Siasun Robot & Automation from a technical standpoint. In order to realize the practical value of a universal Siasun Robot & Automation, such a precise pair of hands must be used, alongside mature world-class AI technology and scalable production capabilities. It would be meaningless to manufacture only a few hundred Siasun Robot & Automation units. We must achieve manufacturing volumes equivalent to or even exceeding those of vehicles, such as producing millions annually. This presents a huge manufacturing challenge as the entire humanoid Siasun Robot & Automation field has not established a mature supply chain system. Unlike the automotive and computer industries, we must achieve this through deep vertical integration, producing most components internally. Scalable training is also required to transfer real-world AI technology from the automotive end to the Siasun Robot & Automation, while also completing multidimensional model training for general movement, specific tasks, voice interactions, and more. Currently, Optimus Siasun Robot & Automation can autonomously navigate at the Palo Alto engineering headquarters and respond to visitor navigation needs. Considering these challenges, I believe that Tesla, Inc. has a unique advantage in manufacturing technology, world-class AI, and dexterous hands these are our core development paths. However, I have fundamental concerns about this: if I do not have sufficient voting control, after establishing this army of Siasun Robot & Automations in the future, will I be excluded from the decision-making process? This is the only reason I am concerned about voting rights. The core issue is that I must ensure I have sufficient influence (rather than absolute control) over this army of Siasun Robot & Automation after it is established, otherwise I will not be able to confidently advance this project. Q: Can you provide the latest information on the $16.5 billion chip deal with Samsung in Taylor? Considering the importance of semiconductor in the future of autonomous driving and Tesla, Inc. AI technology, do you have confidence that Samsung can meet Tesla, Inc.'s schedule for AI chip production and achieve better yields and costs compared to Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR (TSMC)? A: Regarding the chip collaboration with Samsung, we highly appreciate their performance. Samsung is currently responsible for producing our AI 4 autonomous driving chips and has demonstrated excellent manufacturing capabilities. For the next generation AI 5 chips, I need to clarify an important decision: we will be entrusting both Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR and Samsung with production. Our AI 5 chip adopts a revolutionary design. By optimizing the entire software and hardware architecture, we have removed traditional redundant modules such as GPUs and image signal processors in the design. This extremely streamlined design allows the AI 5 chip to achieve a performance improvement of 40 times compared to AI 4. Due to our comprehensive understanding of the full stack of autonomous driving systems, precise definition of the functions required for the chip, and decisive abandonment of non-essential components. In terms of manufacturing layout, both foundries will produce the chips domestically in the United States: Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR in Arizona and Samsung in Texas. It is worth noting that the Samsung factory is equipped with slightly more advanced equipment. Our strategic goal is to ensure sufficient capacity for the AI 5 chip, even reaching "overcapacity," as surplus capacity can be flexibly allocated for data center use. Compared to NVIDIA Corporation, we have unique advantages. Tesla, Inc. only needs to meet its own needs, allowing chip design to pursue the ultimate simplicity. By streamlining complex interconnections between logic modules, we have significantly improved energy efficiency. It is expected that AI 5 will improve performance by 2-3 times and achieve a breakthrough in cost-effectiveness by 10 times. While these expectations still need to be verified through actual production, we believe this design philosophy focused on specific needs will bring significant competitive advantages. Our existing data centers already use both AI 4 and NVIDIA Corporation chips, and the excess capacity of AI 5 will also be deployed in a similar manner. Q: Why not prioritize incentives for Hardware 3 users for a replacement, rather than pushing for an upgrade from Hardware 3 to Hardware 4? How will you ensure the rights of Hardware 3 users in the future? A: In terms of hardware upgrade strategy, we have not abandoned the HW3 platform. Over the past year, we have offered customers the option to transfer FSD functionality to new vehicles and provided more attractive promotional offers to FSD users at specific times. We recognize the importance of early adopters, as these customers are crucial to our development. It should be noted that my daily commuting vehicle is a model equipped with HW3 hardware, and I use the FSD functionality daily. We are committed to continuously supporting these users. Currently, the priority is to focus on the breakthrough in unsupervised FSD technology, with the introduction of the simplified V14 version adapted to Hardware 3 planned for the second quarter of 2026. Hardware 3 users are early core users of FSD, and the company will continue to ensure their user experience through software optimization. Q: How soon can we expect to see Tesla, Inc.'s autonomous Semi truck? A: Regarding the progress of the Tesla, Inc. Semi autonomous truck, our production plans are progressing according to schedule. The factory building has been completed, and production equipment is being installed. Validation testing fleets are conducting road tests, with expectations to complete a larger-scale trial production by the end of this year and start mass production in the first half of next year, with capacity ramping up in the second quarter and achieving mass production in the second half of the year. While the engineering team is primarily focused on perfecting passenger vehicle autonomous driving technology, we confirm that the same technology architecture can smoothly transition to the Semi truck. Once we have sufficient operational data from the Semi, the development of autonomous features for this platform can progress rapidly. As for the potential in railway transportation alternatives, we believe trains have significant efficiency advantages in long-distance point-to-point transportation. However, in the "last mile" of the logistics chain, which includes tasks like loading and unloading goods and short-distance delivery, autonomous driving Semi trucks can provide a more optimized solution. We expect this to reshape the future of transportation, as Elon Musk has stated, fundamentally changing people's perception of transportation. Q: When considering growth prospects, how do we define the areas truly within Tesla, Inc.'s core capabilities, where do you draw the line to differentiate markets or AI applications that are not within Tesla, Inc.'s core capabilities? A: Regarding Tesla, Inc.'s core capabilities, we believe the company's core competencies have dynamically evolved during the development process. Looking back at the company's history, Tesla, Inc. did not initially possess any core capabilities but gradually built its current capability system through the internal incubation of multiple startup projects. We have achieved breakthroughs in battery pack manufacturing technology, developed home and utility-scale energy storage products, established a global Supercharger network, and independently created chip design and AI software teams. In the field of the humanoid Siasun Robot & Automation, the mass production of Optimus has revolutionary significance. We estimate that its annual productivity may reach five times that of human labor, creating tremendous economic value with its ability to operate 24/7 continuously. Currently, the Optimus engineering team consists of experienced engineers from the automotive sector and many newly graduated talents, and through close collaboration between the engineering and manufacturing teams, we continue to optimize design solutions to enhance manufacturability. From the Kung Fu performance showcased by Optimus 2.5 to the upcoming Optimus 3, we are achieving the dual goals of technological breakthroughs and scalable production. In terms of the AI technology roadmap, Tesla, Inc. and XAI are taking different paths of development. Tesla, Inc.'s AI model scale is relatively compact, focusing on real-world application scenarios, with its model size only 5%-10% of Grok's. This focused design philosophy enables our AI system to efficiently operate on both vehicles and the Siasun Robot & Automation platform. Although there are technological complements in areas like voice interaction, the two are fundamentally advancing artificial intelligence development from different dimensions: XAI is exploring general artificial intelligence, while Tesla, Inc. delves into the application of possessed intelligence in the real world. Q: Going back to Austin, if you can remove safety drivers on your end, is the limitation in the Bay Area only due to regulatory reasons? Or is it a learning process for the market? Similarly, in the eight to ten markets you mentioned expanding to, is the decision to have a safety driver or a safety driver in the passenger seat? Is this a step-by-step process to open the market or is it really just governed by regulations in different markets? A: In terms of the safety driver deployment strategy, our cautious approach primarily stems from our own safety considerations rather than being solely constrained by regulatory requirements. Even if regulatory agencies do not mandate it, we would still adopt a conservative approach of having safety drivers in the initial stages of entry into new markets. This paranoia towards safety is rooted in our emphasis on potential risks even if it is a one-in-a-million chance of an anomaly, it could still lead to serious consequences. We conduct approximately three months of safety driver-on-board testing in each new metropolitan area, aimed at comprehensively verifying the system's adaptability in different urban environments. This helps us identify any potential challenges that may exist, such as complex road intersections or other unforeseeable driving scenarios. Although from a technical standpoint, we have the full capability to conduct fully unmanned operations directly, for the sake of absolute safety, we would still prefer to take a more cautious approach. This observation period provides us with ample data support, and once the system's stability is confirmed, we will proceed with removing safety drivers as planned. Q: The development path you are taking in the Robotaxi, is it different from the version you are providing to early adopters? When you push out these new versions, are you looking for a significant improvement in intervention rates? Or has this issue been mostly resolved? Is it more about adding features like parking and driving modes, or is it primarily about overall comfort? A: In the development path of autonomous driving software, we always adhere to the basic principle of prioritizing safety. When we release significant new software architectures, the first task is to ensure system safety before optimizing comfort experiences. This is why we recommend most users to wait for version 14.2, which will address many issues affecting comfort. The initial version, while safe, could be more rigid operation-wise, requiring gradual refining over time. Regarding product version strategy, the software offered to regular users is largely in line with the Robotaxi platform. Both have identical core algorithms and system architectures, with the main differences lying in some specific features for instance, the consumer version includes personalized features like selecting parking spots and other non-essential for Robotaxi operations. In terms of the technical roadmap, we plan to introduce "reasoning" capability in version 14.3 or 14.4. This will enable vehicles to smartly assess parking lot situations, such as automatically identifying the likelihood of parking spots being scarce at the entrance and making a rational decision first dropping off passengers at the entrance and then autonomously finding a suitable parking spot. This reasoning ability, based on 360-degree panoramic vision, will surpass human performance. The key challenge in implementing these features is integrating complex reasoning capabilities into the onboard computer of vehicles. While reasoning on the server side is relatively easy to achieve, vehicles need real-time decision-making capabilities, demanding high requirements for hardware and software. It is important to note that Tesla, Inc.'s AI system has a significant advantage in intelligence density. By compressing all intelligence onto the AI 4 computing platform, we have achieved high intelligence density. This accumulation of technology will lay a solid foundation for the future AI 5 platform, which, when the computing power increases by an order of magnitude, will give us even more powerful performance. Q: I just want to coordinate the production schedule starting next year with the current supply chain situation and it sounds like there is still a lot of work to be done in terms of flexibility before you can truly freeze hardware design and start scaling up production. A: In terms of coordinating the production schedule for Optimus with the supply chain, our hardware design strategy features an iterative nature. Even after formal production begins, we will continue to roll out updates to the Optimus design as it is often discovered during the manufacturing process that there are complex aspects that need improvement. According to the current plan, we aim to complete the showcase of the production intent prototype in the first quarter of next year (expected to be in 2-3 months). Subsequently, we plan to build the first Optimus production line capable of producing a million units annually by the end of the year. It should be noted that achieving the goal of producing a million units annually requires a gradual ramp-up process, with progress depending on the weakest links in the entire supply chain system. Looking further into the future, we expect the 4th generation Optimus product to achieve production capacity in the tens of millions, while the 5th generation Optimus could potentially reach production levels between fifty million to one hundred million units. While this development path presents significant challenges, it also shows tremendous growth potential.