Tesla's conference call will disappoint the market, Musk lowers the targets for robots and Robotaxi, stating that "Tesla is at a critical moment"

Tesla's future is no longer just about electric vehicles. During the third-quarter earnings call on October 22, CEO Elon Musk spent little time discussing the automotive market, instead shifting the narrative focus entirely to artificial intelligence, autonomous taxis (Robotaxi), and humanoid robots Optimus, claiming the company is at a "critical turning point" in bringing AI into the real world.

However, Musk also lowered the near-term advancement targets for Optimus and Robotaxi, indicating that even the most aggressive technological visions must face real-world challenges when implemented.

He expects that by the end of 2025, Tesla will remove safety drivers from Robotaxis in certain areas of Austin and expand the service to 8 to 10 cities, significantly narrowing from the previous goal of covering 50% of the U.S. population. Meanwhile, the Cybercab, designed for fully autonomous driving without a steering wheel and pedals, is scheduled to begin production in the second quarter of 2026.

For Optimus, Musk provided a more cautious timeline, with the production line startup pushed back to before the end of 2026. He candidly expressed that he feels uneasy about leading the company to build a "robot army" without "strong influence," directly pointing to the upcoming shareholder vote regarding his compensation and control.

Despite Tesla announcing record third-quarter revenue and free cash flow, the market's attention is clearly drawn to the company's AI narrative. The emphasis on long-term, capital-intensive AI projects is reshaping Tesla's investment story—a grand narrative of transforming from a hardware manufacturer to an AI giant is unfolding.

Key Takeaways from the Call:

• Strategic Shift: CEO Elon Musk clearly stated that Tesla is at a "critical turning point" in bringing AI into the real world, with the company's future strategic core being "real-world AI," not just electric vehicles.

• Optimus Robot:

  • Latest Version: The Optimus V3 prototype is expected to debut in the first quarter of 2026, with Musk stating it "will look like a person in a robot suit."

  • Production Plan: The production line for Optimus is planned to start before the end of 2026, with a target of producing one million units per year. The original plan was to achieve mass production this year, but the timeline has been delayed.

  • Core Challenges: Musk emphasized that manufacturing dexterous hands and establishing a completely new supply chain are significant challenges.

  • Personal Control: Musk candidly stated that he is pushing for a new compensation plan to ensure he retains strong influence after "building this large robot army" to avoid being ousted.

• Autonomous Driving and Robotaxi:

  • Unsupervised FSD: Musk stated that he has a "clear understanding" and "100% confidence" in achieving "unsupervised full self-driving" that surpasses human safety levels, which is why he decided to expand vehicle production capacity as soon as possible.

  • Robotaxi Expansion: By the end of 2025, it is expected that safety drivers for Robotaxi will be removed in most areas of Austin, and plans are in place to expand Robotaxi services to 8 to 10 metropolitan areas.

  • FSD V14: The latest version of the FSD software has been pushed to all U.S. owners, and future versions will incorporate "reasoning" capabilities, allowing vehicles to think like "biological" entities.

• Financials and Automotive Business:

  • Record Performance: Third-quarter revenue, delivery volume, energy business gross profit, and free cash flow (approximately $4 billion) all reached historical highs. Automotive business revenue grew by 29% quarter-over-quarter.

  • Cybercab: The Cybercab, designed for fully autonomous driving (without a steering wheel and pedals), will begin production in the second quarter of 2026, which will be the largest single source of capacity expansion.

  • AI Chips: Musk detailed the next-generation AI chip AI5, claiming it is 40 times more powerful than AI4 in certain metrics, with energy efficiency and cost-effectiveness far exceeding competitors.

• Energy Business: Demand for energy storage products Megapack and Powerwall remains strong. The next-generation Megapack 4 will integrate substation functions, greatly simplifying the deployment process.

• Shareholder Voting: At the end of the meeting, company executives strongly urged shareholders to vote in support of Musk's compensation plan and the re-election of board members at the meeting on November 6, to ensure the company's future stability and development.

"We are at a critical turning point": Tesla's new story—AI and Robotics

At the beginning of the conference call, Musk hardly mentioned traditional automotive sales but instead positioned Tesla directly as a "leader in real-world artificial intelligence." He claimed that the company is at a "critical inflection point," with its updated mission being to achieve "sustainable prosperity."

Musk believes that a world without poverty can be created through full self-driving and Optimus robots, where everyone has access to the best medical services.

He stated, "I think people haven't fully realized the extent to which this will take off; to be honest, it will be like a shockwave." This confidence stems from his clear expectations for achieving "unsupervised full self-driving," which is also the premise for his decision to "expand vehicle production capacity as soon as possible."

This shift indicates that Tesla's future growth engine will no longer rely solely on the number of vehicles sold but will depend on the proliferation of its AI software and services.

Robot Timeline: Optimus mass production delayed, Cybercab to start production in Q2 next year

For future products, which are of high interest to investors, Musk provided a clear but mixed timeline The Cybercab, designed specifically for autonomous driving, is highly anticipated, with Musk confirming that production will begin in the second quarter of 2026. This vehicle "has no steering wheel or pedals" and has undergone engineering optimization to achieve the lowest unit operating cost per mile.

However, the timeline for what he calls "possibly the grandest product ever"—the Optimus humanoid robot—has been delayed. Musk expects mass production of Optimus to begin by the end of 2026, while the V3 version of Optimus will be ready for "debut" in the first quarter of 2026. He admitted that manufacturing Optimus is an "extremely difficult task," especially since Tesla must achieve deep vertical integration in the absence of an existing supply chain.

Notably, Tesla originally planned to produce 5,000 Optimus robots this year and 50,000 in 2026, but reports indicate that Musk has abandoned this year's target, primarily due to issues with the robot's hand design.

Robotaxi: Unsupervised driving by year-end, expansion plans narrowed

The autonomous driving business is a core pillar supporting Musk's AI vision. The earnings call revealed that Tesla's Robotaxi service is progressing steadily. Company executive Ashok Elluswamy stated that vehicles without a driver in Austin have traveled over 250,000 miles. Musk expects that by the end of 2025, Robotaxis in most areas of Austin will no longer require a safety driver.

At the same time, Tesla plans to expand its Robotaxi service to 8 to 10 metropolitan areas, including Nevada, Florida, and Arizona, by the end of the year, depending on regulatory approvals. Musk emphasized that the deployment in new markets will take a "paranoid" cautious approach to ensure safety. It is this confidence in the prospects of FSD technology that has alleviated his previous "hesitation" about expanding car production, planning to increase annual capacity to 3 million units within the next 24 months.

However, the expansion plans for Robotaxi clearly contradict Musk's statement during the last earnings call, where he stated that the company's goal is to cover half of the U.S. population by 2025.

"Robot Army" and Voting Rights: Musk expresses control concerns

During the call, Musk directly linked his concerns about control of the company to the robot project for the first time. He candidly posed a question: "My fundamental concern about how much voting control I have at Tesla is whether I could be ousted in the future if I build this massive robot army?"

He added, "If I don't have at least strong influence, I feel uneasy about building that robot army." This statement has been widely interpreted as a defense of his upcoming shareholder vote on the compensation plan. If approved, this plan would significantly increase his voting power. Musk also targeted proxy advisory firms ISS and Glass Lewis, calling them "corporate terrorists," arguing that their governance recommendations could harm shareholders

Technical Core: New Developments in AI5 Chip and Energy Storage Business

To drive its AI ambitions, Tesla is making a big bet on chip design. Elon Musk detailed the AI5 chip that he personally participated in designing. He revealed that Tesla will simultaneously commission Samsung and TSMC to produce this chip at their factories in the United States, aiming for an "oversupply of AI chips."

Musk stated that by eliminating non-essential modules such as traditional GPUs and image signal processors, the AI5 chip will be "40 times better" than the AI4 in certain metrics, and may achieve improvements of 2-3 times in energy efficiency and 10 times in cost-effectiveness. In terms of the energy storage business, the company released the Megapack 3 and plans to launch the Megapack 4, which will integrate more substation functions to simplify deployment processes and meet the surge in power demand brought by AI and data centers.

Financial Performance: Record Revenue, AI R&D and Tariffs Weigh on Profits, Bitcoin Also Affects

From the financial data, Tesla performed strongly in the third quarter. Chief Financial Officer Vaibhav Taneja stated that the company set new records in several areas, including delivery volume, total revenue, energy business gross profit, and free cash flow. Automotive revenue increased by 29% year-on-year, with free cash flow reaching a record of approximately $4 billion.

However, profitability is under pressure. The automotive business gross margin, excluding carbon credit sales, slightly increased from 15% to 15.4%, partly due to lower material costs and economies of scale. Vaibhav Taneja mentioned in the earnings call that the total impact of tariffs in the third quarter exceeded $400 million, and Tesla incurred legal fees related to certain lawsuits, as well as "incremental costs incurred to prepare for the shareholder meeting." He stated that tariffs and market competition are the two major challenges Tesla currently faces.

He also mentioned that in terms of other income, the company's other income decreased quarter-on-quarter, mainly due to the market value re-evaluation adjustment of Bitcoin (BTC) holdings, with third-quarter earnings being small at $80 million, compared to $284 million in the second quarter, with the remaining fluctuations attributed to foreign exchange movements in the current quarter.

Full Translation of Tesla's Q3 2025 Earnings Call Minutes (Translated by AI tools):

Event Date: October 22, 2025 Company Name: Tesla Event Description: Q3 2025 Earnings Call Source: Tesla

Presentation Segment

Travis Axelrod, Head of Investor Relations: Good afternoon, everyone, and welcome to Tesla's Q3 2025 Q&A live stream. I am Travis Axelrod, the Head of Investor Relations, and joining me today are Elon Musk, Vaibhav Taneja, and several other executives. Our third-quarter results were released around 3 PM Central Time, in the updated document linked to this live stream.

During this call, we will discuss our business outlook and make forward-looking statements. These comments are based on our forecasts and expectations as of today. Due to various risks and uncertainties, actual events or results may differ materially, including those uncertainties mentioned in our recent filings with the U.S. Securities and Exchange Commission (SEC) We urge shareholders to read our final proxy statement, which contains important information regarding the matters to be voted on at our 2025 annual meeting. During today's Q&A session, please limit your questions to one question and one follow-up. Please use the raise hand button to join the question queue. Before we enter the Q&A session, Elon has some opening remarks. Elon?

Elon Musk, Co-founder, Tech King of Tesla, CEO, and Director: Thank you. As we bring artificial intelligence (AI) into the real world, we are at a critical turning point for Tesla and its future strategy. I think it is important to emphasize that Tesla is indeed the leader in real-world AI. In the realm of real-world AI, no one can do what we can do, and we have a good insight into AI overall. I believe Tesla has the highest intelligence density of all AI, whether in the car or elsewhere, and it will only get better. We are just beginning to scale up full self-driving and Robotaxi, fundamentally changing the nature of transportation.

I think people have not fully realized the extent to which this will take off—honestly, it will be like a shockwave. Because these cars are already on the road. There are millions of cars out there that can become fully self-driving vehicles with a software update, and we are still producing millions of them each year. In fact, with our current clear understanding of achieving unsupervised full self-driving, I should say that I am confident in scaling up Tesla's production. So, our intention is to scale up our future production as soon as possible. I was cautious about this before having a clear understanding of achieving unsupervised full self-driving. But now, I feel we have a clear understanding, and it makes sense to scale up production as soon as possible.

We are also making a huge impact on the energy sector through battery storage. With Powerwall, especially Megapack, we are significantly increasing the grid's ability to generate more energy. Let me talk about this a bit; if you look at the total energy capacity in the United States, for example, the U.S. has about 1 trillion watts of continuous available power, but the average usage over a 24-hour cycle is only 0.5 trillion watts due to the significant difference in usage between day and night. If you use batteries to buffer energy, you can actually double the energy output of the U.S. just with batteries, without having to build any new power plants. And building power plants is very difficult. They take a long time, have many approval processes, and the industry is not accustomed to acting quickly. So, we see the potential of Tesla's battery packs to greatly increase the annual energy output of any given grid (whether in the U.S. or elsewhere). We are also on the verge of a truly significant breakthrough with the Optimus project, which I believe has the potential to become the greatest product of all time. This is a difficult project, and it is worth noting that this is not a given.

I do not know if Ford, General Motors, or any of our American car companies have any robotics projects; they may think of Tesla as a car company, primarily producing cars and battery packs. But it is not an obvious thing to make Optimus But we do have the elements of real-world AI, excellent electrical and mechanical engineering capabilities, and the ability to scale production, and I believe no one else has all these elements at the same time. So, with the launch of the autonomous driving V14 version, you can see people's reactions online; they are very surprised. In fact, anyone in the United States can access the V14 version as long as they select "I want advanced software" in the car.

So, if you are listening now and want to give it a try, just go into settings and say "I want advanced software," and you will get the V14 version. Yes. In terms of Megapack, we released Megablock, Megapack 3. We also have exciting plans for Megapack 4. Megapack 4 will integrate a lot of equipment that is usually found in substations and can directly output about 35 kilovolts of voltage. This will greatly enhance our ability to deploy Megapack, as it does not rely on building a 335-kilovolt substation for Megapack 4. So, this will be an engineering priority for Megapack. We look forward to showcasing Optimus V3 in the first quarter (of next year), and I believe it will be ready for demonstration. And I think that will be amazing. If you— it doesn't even look like a robot. It will look like a person in a robot suit, which is somewhat like how we started with Optimus. But it will look so real that you would need to poke it, I think, to believe it is actually a robot.

It is clear that the real-world intelligence we are developing for cars can largely be transferred to Optimus. So, this is a very good starting point. In summary, we are excited about Tesla's renewed mission, which is "sustainable abundance." So, beyond sustainable energy, we say "sustainable abundance" is our mission, and we believe that through Optimus and autonomous driving, you can actually create a world without poverty, where everyone has access to the best medical services, like Optimus being an incredible surgeon. Imagine if everyone could access an incredible surgeon. So, I think, of course, we need to ensure that Optimus is safe in every way, but I do believe we are moving towards a world of sustainable abundance, and I am excited to work with the Tesla team to achieve this goal.

Travis Axelrod, Head of Investor Relations: Great. Thank you very much, Elon. Vaibhav also has some opening remarks.

Vaibhav Taneja, Chief Financial Officer: Thank you, Travis. The third quarter has been a special quarter on multiple levels. We not only set new records in deliveries and deployments but also set new records across a range of financial metrics, from total revenue, energy business gross profit, energy business profit margin to free cash flow. This is a result of our customers' continued confidence in our products and the relentless efforts of the Tesla team. The strong growth in deliveries is attributed to strong performance across all regions The Greater China region and the Asia-Pacific region grew by 33% and 29% quarter-over-quarter, respectively. North America grew by 28%, while Europe, the Middle East, and Africa grew by 25%.

The delivery speed is the result of the ongoing excitement surrounding the new Model Y. We previously mentioned that 2025 would be the "Year of Y," and since then, we have delivered on that promise by launching the new Model Y in the first quarter, followed by the Model Y Long Range and Performance versions, and recently the Standard version Y in North America and Europe, the Middle East, and Africa. We are now operating Robotaxi in two markets, Austin and most cities in the Bay Area. Since the initial launch, we have expanded our coverage in Austin three times and are continuing to expand further as planned.

Unlike our competitors, our Robotaxi fleet is integrated into the markets we operate in, as they do not have additional sensor suites or peripherals that make them stand out. This is an underestimated aspect of our current vehicle products, all of which are designed for autonomous driving. We believe that as people experience supervised FSD on a large scale, the demand for our vehicles will significantly increase, as Elon has stated.

In terms of FSD adoption, we continue to see good progress. However, please note that the total number of paid FSD customers remains small, at about 12% of our current fleet. We are working with regulators in regions such as China and Europe, the Middle East, and Africa to obtain approvals so that we can launch FSD in those areas. Now, let's talk about the financials. Automotive revenue grew by 29% quarter-over-quarter, consistent with the increase in delivery volume. Although regulatory credit revenue decreased quarter-over-quarter, we signed new contracts and continued to fulfill previously signed contracts. Our automotive gross margin (excluding credits) increased slightly from 15% to 15.4%, attributed to improvements in material costs and better fixed cost absorption due to increased sales volume. The energy storage business continues to perform exceptionally well, with deployment, gross profit, and profit margins all reaching record highs. As mentioned earlier, this business is more affected by tariffs (measured as a percentage of cost of goods sold), as all battery cells currently procured come from China, while we are still working to find other alternatives.

However, as the Shanghai Gigafactory ramps up production capacity, this is helping us avoid tariffs, as we are using this factory to supply demand in non-U.S. markets. As Elon stated, the only way we can achieve electrification at grid scale is through energy storage. Another thing to keep in mind is that, given the intensifying competition and tariffs, we are facing headwinds in this business. The total tariff impact on both businesses in the third quarter exceeded $400 million, essentially evenly distributed between the two. The service and other businesses showed significant improvement quarter-over-quarter. This is mainly the result of improvements in our insurance and service center businesses. Please note that while the scale is small, our Robotaxi costs are included in the service and other businesses—as well as our other businesses, such as paid Supercharging, used car sales, parts, and merchandise sales. Our operating expenses increased quarter-over-quarter. The largest increases include restructuring and other expenses related to certain actions taken to reduce costs and improve efficiency through the integration of our AI chip design work

In addition, we have incurred legal fees related to certain legal cases and incremental costs associated with preparing for the shareholders' meeting. These costs are recorded in selling, general, and administrative expenses (SG&A). Additionally, our employee-related expenses are increasing, particularly in the R&D area, as we have recently granted various performance-based equity awards to employees involved in AI programs, so these types of expenses will continue to rise in the future. In terms of other income, our other income decreased quarter-over-quarter, primarily due to the market value revaluation adjustment of Bitcoin (BTC) holdings, with third-quarter earnings being small at $80 million, compared to $284 million in the second quarter, and the remaining fluctuations attributed to foreign exchange volatility this quarter. This quarter's free cash flow was approximately $4 billion, marking another record. At the end of this quarter, our total cash and investments exceeded $41 billion. In terms of capital expenditures, while we expect to spend around $9 billion this year, we anticipate a significant increase in this figure by 2026 as we prepare for the next phase of growth for the company, not only for our existing business but also for our bets around AI programs (including Project Dojo). In summary, please note that bringing AI into the real world is challenging, but we have never shied away from doing difficult things. We are incredibly excited about the future and are laying the groundwork. The benefits will be realized in the coming years. Finally, I want to thank the Tesla team, our customers, our investors, and supporters for their continued trust in us.

Q&A Session

Travis Axelrod, Head of Investor Relations: Thank you very much, Vaibhav. Now we will move into the investor question segment. The first question from say.com is, what are the latest Robotaxi metrics, including fleet size, total mileage, completed trips, intervention rates, and when will safety drivers be removed? What are the remaining barriers preventing unsupervised FSD deployment to customer vehicles?

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director: I'll start, and then Ashok can elaborate. But we expect that by the end of this year, there will be no safety drivers in most areas of Austin. So, in a few months, we expect there will be no safety drivers at all in at least parts of Austin. We are obviously very cautious in our deployment. So, our goal is to be paranoid about deployment because it is clear that even a single accident can become global front-page news. So, we better take a cautious approach. But we do expect that in a few months, there will be no safety drivers in the cars in Austin. I think that might be the most important data point. Then we do expect to operate Robotaxi in about 8 to 10 metropolitan areas by the end of the year. This depends on various regulatory approvals. And—but you can actually, I think most of our regulatory applications are online, and you can see them because they are public information. But we expect to operate in Nevada, Florida, and Arizona by the end of the year. Ashok?

Ashok Elluswamy, Executive Officer: Yes. We continue to operate our fleet in Austin with no one in the driver's seat, and we have driven over 250,000 miles. Then in the Bay Area, due to regulatory reasons, we still have someone in the driver's seat, and we have driven over 1 million miles. So—we continue to see the fleet—Robotaxi fleet operating very well. Customers are very satisfied, and there are no notable issues. On the customer side, as of yesterday, customers have driven a total of 6 billion miles using supervised FSD. So, this is an important milestone. Overall, safety remains very good. As Elon mentioned, we are on track to completely remove personnel from the vehicle, starting from Austin.

Travis Axelrod, Head of Investor Relations: Great. The next question is, how is the demand and backlog for Megapack, Powerwall, solar, or energy storage systems? With the current AI boom, does Tesla plan to supply power to other hyperscale data center providers?

Michael Snyder: Thank you. We expect demand for Megapack and Powerwall to remain very strong into next year. We have received very strong positive feedback from customers on our megablock product, which we will begin shipping from Houston next year. We are seeing significant growth in demand for AI and data center applications, as hyperscale data centers and utility companies have recognized the versatility of the megapack product in enhancing reliability and alleviating grid constraints, as Elon mentioned. Due to policy changes, we are also seeing a surge in demand for residential solar in the U.S., and we expect this to continue into the first half of 2026, when we will launch new solar leasing products. We have also started producing our Tesla residential solar panels at the Buffalo factory and will begin shipping to customers starting in the first quarter. The panel features industry-leading aesthetics and shape performance, showcasing our ongoing commitment to American manufacturing.

Travis Axelrod, Head of Investor Relations: Great. Thank you, Mike. Unfortunately, the next question is about future products. This is not the right venue to discuss this. So we have to skip it. The next question is, considering application control software, what challenges does Optimus currently face in terms of engineering hardware, training general mobile models, training task-specific models, training voice models, implementing manufacturing, and establishing supply chains?

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director:

Yes, I mean, bringing Optimus to market is an extremely difficult task, and it needs to be clear that it is not as simple as taking a walk in the park. To some extent, I mean, technically speaking, Optimus can now take a walk in the park We indeed have the Optimus robot moving around in our engineering headquarters office in Palo Alto, California, basically 24 hours a day, 7 days a week. So, any visiting guests, you can actually stop an Optimus robot and have it take you somewhere; it will really take you to that meeting room or location in the building.

So I don't want to underestimate the difficulty of Optimus; it's an extremely challenging task, especially to create a hand that is as dexterous and capable as a human hand. It's incredible; the human hand is an amazing thing. The more you study the human hand, the more you realize its incredible nature and why you need five—four fingers and a thumb, and why the fingers have specific degrees of freedom. You know why different muscles have different strengths, and why fingers have different lengths; it turns out all of this has a reason. So, making that hand and forearm is an extremely difficult engineering challenge. I would say, from an electromechanical perspective, the forearm and hand are more difficult than the rest of the robot.

But to have a useful general-purpose robot, you really need that incredible hand, you need real-world AI, and you need to be able to scale production to make it relevant, because if there are only a few hundred robots, it doesn't make sense. So, you need to be able to manufacture Optimus robots at a scale comparable to or even higher than vehicles. So, if you want to produce around a million each year, trying to make a million Optimus robots annually, that manufacturing challenge is enormous, considering that the supply chain doesn't exist at all. For cars, you have an existing supply chain. For computers, you have an existing supply chain. For humanoid robots, there is no supply chain. So, to manufacture it, Tesla actually has to do very deep vertical integration, manufacturing at a very deep level of the supply chain, making parts themselves because there is no supply chain.

So, that's why I would think, if I put myself in the position of a startup trying to make humanoid robots, I would think I don't know how to do it without a lot of manufacturing technology. So, that's why I believe Tesla is in a unique position, considering manufacturing technology, scalability, real-world AI, and a truly dexterous hand. These are three things that are usually missing when you read about other robots; they lack these three things. So, I believe we can achieve all three through tremendous effort. That's our plan.

So, regarding how much voting control I have at Tesla, my fundamental concern is whether, if I continue to build this massive army of robots, I might be removed at some point in the future. That is my biggest concern. If I—this is really the only issue I'm trying to solve. It's called compensation, but it's not that I'm going to spend that money. It's just—if we build this army of robots, do I at least have a strong influence over that army of robots, not control, but strong influence? Ultimately, that's how it is. If I don't have at least a strong influence, I feel uneasy about building a robot army.

Travis Axelrod, Head of Investor Relations:

Great. Thank you. We have discussed the expansion of Robotaxi. Unfortunately, the next question is again about future products. So we have to skip it. However, the next question is, can you provide an update on the $16.5 billion Samsung chip deal in Taylor? Given the importance of semiconductors for autonomous driving and Tesla's AI-driven future, what gives you confidence that Samsung can meet Tesla's timeline for AI-6 and achieve better yields and costs compared to TSMC?

Elon Musk, Co-founder, King of Tesla Technology, CEO, and Director:

Okay. So, I will give a fairly long answer to this question because I need to deconstruct the question and then answer the deconstructed version. First of all, I have nothing but praise for Samsung. They are an incredible company, and it is worth mentioning that Samsung is indeed manufacturing our AI full self-driving computers, and they are doing an outstanding job.

Now, regarding AI5, I need to clarify some of my previous public comments, which is that we will actually have both TSMC and Samsung initially focus on AI5. The AI5 chip designed by Tesla, I think it's an incredible design. Over the past few months, I've spent almost every weekend working with the chip design team on AI5. I don't easily praise people, but I must say that I believe Tesla's chip team is indeed designing an incredible chip. By certain metrics, the AI5 chip will be 40 times better than the AI4 chip, not 40%, but 40 times.

Because we have a detailed understanding of the entire software and hardware stack. So, we are designing hardware to address all the pain points in the software. So, I don't think, really, anyone else is doing this; I think the entire stack, all the way to calibrating with the real world, you have cars and robots in the real world, we know what the chips need to do, and equally important, we know what the chips do not need to do. To give you some examples, for AI5, we removed the traditional GPU from AI4. AI5 does not have — we just removed the traditional GPU because it is basically a GPU. We also removed the image signal processor, in fact, it's a long list of removals, and these removals are very important.

Because of these removals, we can actually fit AI5 into half a reticle, and the routing from memory to the chip — Tesla's chip accelerator, ARM CPU cores, and PCI modules like PCI X — all have good margins. So, it's a beautiful chip. I personally have poured so much life energy into this chip, and I believe this will be — this will be a winner, it's the next level. So, it makes sense to have both Samsung and TSMC focus on AI5.

Technically speaking, Samsung's wafer fab equipment is slightly more advanced than TSMC's wafer fab. Both will manufacture in the United States, one is TSMC in Arizona, and the other is Samsung in Texas. However—starting from our clear goal—is to have an excess supply of AI5 chips. Because if we prepare too many AI5 chips for cars and robots, we can always use them in data centers. We are already using AI4 for training in data centers. So, we are combining AI5 with NVIDIA's hardware. It needs to be clear that we do not intend to replace NVIDIA, but we do combine AI4 with NVIDIA's hardware, and we can always place the excess capacity of AI5 in our data centers.

Yes, NVIDIA is continuously improving. But the challenge they face is that they must meet the massive demands of a large number of customers, while Tesla only needs to meet the demands of one customer, which is Tesla. This makes the design work extremely simple, and we can eliminate a lot of complexity from the chips. I cannot emphasize how important this is. So, when you see various logic blocks in the chip, as the number of logic blocks increases, the interconnections between the logic blocks also increase. You can think of it as highways; how many highways do you need to connect the different parts of the chip? Especially if you are not sure how much data will be transmitted between each logic block, you may end up with huge highways everywhere. This becomes an almost impossibly difficult design problem, and NVIDIA does an excellent job of handling an almost impossibly difficult set of requirements. But in our case, we are pursuing extreme simplicity, and the end result is that I believe AI5 will be the best in performance per watt, possibly 2 to 3 times better, and possibly 10 times better in AI performance per dollar. So, facts speak louder than words, and clearly, we need to really manufacture this chip and achieve scalable production. But it looks like that.

Participant Question:

Great. Thank you, Elon. We have discussed unsupervised FSD. So the next question is, instead of trying to replace hardware 4 with hardware 3, why not offer an equivalent incentive to trade in for a new car?

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director

Yes. We have not completely given up on hardware 3. However, over the past year, we have offered customers the option to transfer FSD to their new cars, and sometimes we also run some promotional activities. If they have FSD, they can get better rates. So we have been taking care of these customers, but we do want to solve the autonomous driving problem first before figuring out how to take care of these customers. These customers are very important. They are early adopters. By the way, I drive a hardware 3 car for my daily commute, and I use FSD every day. So we will definitely take good care of you.

Participant:

Great. Thank you.

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director:

Once the V14 release series is fully completed, we plan to develop a "light version" of V14 for Hardware 3, expected to launch in the second quarter of next year.

Participants:

Great. Thank you, Ashok. Alright. Our last question from Say is: How much longer until we see autonomous driving tests for the semi-truck? Do you think this technology can replace trains?

Elon Musk, Co-founder, King of Tesla Technology, CEO and Director:

Yes. So I'll start with the production plans and timeline for the semi-truck. The factory is progressing as planned. We have completed the building and are now installing equipment. Our validation truck fleet is on the road. We will ramp up production on a larger scale by the end of this year, then have our first online production in the first half of next year, and enter true mass production in the second half of the year, around the second quarter. So this is progressing quite smoothly, and this is obviously the first step to getting autonomous trucks on the road.

As for trains, they are excellent for long-distance point-to-point transportation. They are super efficient, but the last mile, that is, loading and unloading, can be better served by autonomous semi-trucks over shorter distances, which would be great. So we do expect this shift to happen. As Elon said, we are really changing people's perception of transportation. So we look forward to that moment. Ashok, I know you can talk about the fully autonomous driving part.

Ashok Elluswamy, Executive:

Currently, the team is very focused on solving the autonomous driving problem for passenger cars. That said, once we get some data from [inaudible], the same technology will easily scale to trucks.

Great. Now we turn to analyst questions. The first question comes from Emmanuel at Wolfe. Emmanuel, please unmute.

Emmanuel Rosner, Analyst:

Great. Thank you very much. Hello everyone. So, you mentioned that after gaining confidence in unsupervised autonomous driving, you will expand vehicle production as soon as possible. How should we understand this in the context of your existing capacity of 3 million units? Is this the production target you hope to achieve, and what kind of timeline are we talking about? Does this require some level of demand boost or incentive? Will this essentially prioritize sales over short-term profitability considering long-term opportunities?

Elon Musk, Co-founder, King of Tesla Technology, CEO and Director:

Well, our capacity is not fully 3 million yet, but it will reach 3 million at some point. You know, ideally, we might reach an annualized capacity of 3 million within 24 months, and I think it could be less than 24 months. And remember, there is a whole huge supply chain that will move forward in sync with this. So I think we will ramp up production as quickly as our suppliers can keep up. Then we have to consider where we will build new factories beyond that

The largest single expansion in production will be the Cybercab, which will begin production in the second quarter of next year. This is a vehicle truly optimized for fully autonomous driving. It actually has no steering wheel or pedals, and is engineered to minimize the cost per mile, fully considering the operational cost per mile.

So for other vehicles, they still have a bit of a "horse-drawn carriage" feel; obviously, if you still have a steering wheel and pedals, you are designing a high-performance car that people might want to pursue extreme acceleration and tight cornering, which is different from a vehicle optimized for a comfortable ride, not expecting to exceed 85 or 90 miles per hour, and always pursuing a smooth drive.

So yes. Do I think we will sacrifice margins? I don't think so. I believe demand will be incredibly crazy. But the real killer application ultimately is, can you text in the car? If you tell someone, yes, this car is really good now, you can text on your phone the whole time in the car. Anyone who can afford this car will buy it. End of story.

So this is what everyone wants to do. In fact, not everyone wants to, they just do it. That’s why, in fact, you see the global accident rate rising is because people are texting while driving. So Autopilot actually greatly enhances safety here because if someone is looking down at their phone, they are not driving well.

So this is the real game changer, and we do see, at this point, I feel basically 100% confident, I mean not basically, but 100% confident, that we can achieve a level of safety far exceeding that of humans with unsupervised fully autonomous driving. We have released version 14.1. We have what I think is a very amazing technology roadmap. We will add reasoning capabilities to the car. The world simulator we use for reinforcement learning is incredibly impressive; when you see the Tesla reality simulator, you cannot distinguish between the videos generated by the Tesla reality simulator and actual videos; they look exactly the same.

So this allows us to have a very powerful reinforcement learning loop to further improve Tesla AI. We will increase the number of parameters by an order of magnitude, which is not yet in version 14.1. There are many other improvements to the AI, which are quite radical. So this car will feel like a living being. This is the level that AI can achieve before using AI4 computing. Then there is AI5, which, as I said, is 40 times better by some metrics; I can conservatively say it is a 10 times improvement. So for a car, this might be a bit too smart. I do wonder how much intelligence a car should have. It might... feel bored. Then, one idea I heard is that if we have all these cars that might feel bored, then when they... if they feel bored, we can actually have a huge distributed reasoning fleet, and then say, if they are not actively driving, we let them form a huge distributed reasoning fleet You know, at some point, if you have tens of millions of vehicles in your fleet, or at some point reach 100 million vehicles, and assuming by then, I don't know, they have a kilowatt of high-performance inference capability, that would be 100 gigawatts of distributed inference capability, and cooling and power conversion have been solved. So, this looks like quite a considerable asset.

Travis Axelrod, Head of Investor Relations:

Great. Thank you, Elon. The next question comes from Adam at Morgan Stanley. Adam, feel free to unmute. Adam, please go ahead. We seem to have some audio issues with Adam. So, we will get back to you later. The next question will come from Dan at Barclays.

Dan Levy, Analyst:

Hi, good evening. Thank you for the opportunity to ask a question. Elon, I know Tesla is very focused on bringing AI into the physical world through Masterplan 4. I think we have seen in the past that Tesla is willing to engage and enter new markets, new Total Addressable Markets (TAMs). So, when you consider growth prospects, how do we define which areas truly belong to Tesla's core competencies, and where do you draw the line for markets or AI applications that are outside of Tesla's core competencies?

Elon Musk, Co-founder, Tesla Tech King, CEO, and Chairman:

Actually, I'm not quite sure what you mean by "AI applications beyond Tesla's core competencies." But I mean, when we started, we didn't have any of these core competencies, yours. So, like, the initial core competency was zero, our overall core competency was zero. So, you can think of Tesla as, I don't know, a company with a dozen startups inside it. And I started each of those startups. So, we didn't do stationary battery packs before, but now we do, producing for homes, we produce for utilities at scale with Powerwall and Megapack, we created the global Supercharger network.

No one else created the global Supercharger network. In fact, the Supercharger network in North America is good, reaching almost all other manufacturers in North America have switched to our standard and use our—Tesla Supercharger network. But if it were so easy, why don't they do it themselves? The chip design team also started from scratch. The Tesla AI software team also started from scratch. I really mean to say, hey, we are going to start doing this. I posted on Twitter, now called X. And then, if you want to join us in building, come on.

In fact, I believe Ashok was the first person I interviewed to join the Tesla Autopilot team, which we now call the Tesla AI software team because it is the AI software team. So, the core competencies are created instantly. And the scalable Optimus, it is like an infinite money printing machine, it's hard to express its magnitude, like if you have something, if Optimus, I think it could reach five times the productivity of a person per year because it can operate 24/7 and doesn't even need to be charged It can be operated while being constrained. So, it has been plugged in all the time, which is why I call it that if you achieve sustainable abundance, work will be optional. AI has limited capabilities in enhancing human productivity, but there are no real limits for physically embodied AI. That's why I refer to it as an "infinite money printer."

Ashok Elluswamy, Executive:

I mean, I want to add that people forget that the first version of our Autopilot was ten years ago. So, Elon started this a long time ago.

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director:

Yes. We have (inaudible) to prove it.

Ashok Elluswamy, Executive:

Exactly. And then regarding Optimus, even though people think, oh, this is something new, I still remember over four years ago, when Elon and I were at a financial meeting, Elon said, hey, our car is a robot on wheels. That's where we started developing. In fact, most of the engineering team working on Optimus comes from the vehicle side. That's why when we talk about manufacturing progress, we have this capability because the engineers who worked on the drive units back then are now working on actuators. So, that's what we can do—if any company can scale this, it will be us.

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director:

But we have also added a lot of new engineers to the team. So, a lot of the credit for the Optimus engineering actually goes to the new engineers, many of whom just graduated from college.

Ashok Elluswamy, Executive:

Yes.

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director:

So, the Optimus engineering team is a very talented engineering team. I want to say, wow. In fact, yes. And the review for Optimus is currently—there's an engineering review and a manufacturing review happening simultaneously, with an iterative loop between engineering design and manufacturing because then we see—we design something, and then we say, oh my god, this is really hard to manufacture. We need to change that design to make it easier to manufacture. So, we have fundamentally improved the design of Optimus while adding functionality, but making it truly manufacturable. I want to say, Optimus 2 is almost impossible to manufacture, frankly speaking.

But my two five-volt points, we have evolved from a person in a robot suit to what people see as Optimus 2.5, which is doing kung fu, just like Optimus did kung fu at the premiere of "Tron," right in front of a large audience, doing kung fu with Jared Leto, with no one controlling it; it was doing kung fu with Jared Leto at the premiere of "Tron." You can see videos online, and interestingly, many people think it's just a person when they walk past it.

Even with Optimus 2.5, you can see that its waist is only three inches wide, and it is clearly not human. However, its movements are so human-like that many people do not realize - many people do not realize they are seeing a robot. So, what I want to say is that Optimus 3 will have a huge improvement based on this and will be mass-produced. But as I said, this is a very difficult thing. The engineering and manufacturing review of Optimus, along with the meetings with Optimus on Friday nights, sometimes go until midnight, and then my meeting on Saturday is in the afternoon with the AI5 chip design team. So, these two things are crucial for the future of the company.

Travis Axelrod, Head of Investor Relations:

Dan, do you have a follow-up?

Dan Levy, Analyst:

Yes. As a related question, perhaps you could talk about Tesla's AI efforts and to what extent xAI is complementary, or are they just different forms of AI? Maybe you could help the audience distinguish between them. Thank you.

Elon Musk, Co-founder, Tesla Tech King, CEO, and Director:

Yes, they are different forms of AI. So, xAI is - Grok is a huge model, and you definitely can't fit Grok into a car. It is a massive model. For Grok, it tries to solve the general artificial intelligence problem through a lot of AI training computation and reasoning computation. For example, Grok 5 can only effectively run on GV 300, and that’s the power of Grok 5. Tesla's model may only be less than 10% of its size, possibly closer to 5%. So yes, they do approach the problem from very different angles. XAI is competing with Google's Gemini and OpenAI's ChatGPT. So some parts are complementary. For example, voice interaction in the car is cool. For Optimus, voice recognition and voice generation are Grok. So that helps in this regard. But they are addressing the problem from two opposite ends of the spectrum.

Travis Axelrod, Head of Investor Relations:

So we continue, the next question is from Walt of [Ph]. Walt, please unmute.

Can you hear me now?

Questioner Walt:

Yes.

Questioner Walt:

Great. Thank you. Back to the Austin question, if you can remove the safety driver by the end of the year, is the limitation in the Bay Area just a regulatory issue, or is it a learning process market by market? Similarly, among the 8 to 10 markets you mentioned that you want to expand into, is the decision to have a safety driver in the passenger seat or a safety driver there? Is this a gradual process of market expansion, or does it really just depend on the regulations of individual markets?

Elon Musk, Co-founder, King of Tesla Technology, CEO and Director:

Well, I think even if regulators didn't force us to do this, we would still take it as a cautious way to enter new markets. Just to ensure that we remain paranoid about safety. I think it makes sense to have a safety driver or safety occupant in the car when we first enter a new market, just to confirm that we haven't missed anything. Because if just one in ten thousand trips goes wrong, you have a problem. So it's like, whether a city has any particularities, like a very difficult intersection, or I don't know, some unexpected challenge in the city, targeting that one in ten thousand situation. I think we might be able to roll it out directly in those cities, but we just don't want to take the risk.

What we're talking about here is possibly having a safety driver in a new metropolitan area for three months to confirm that it's okay, and then we remove the safety driver, that's it.

Questioner Walt:

Okay. And about FSD14, it feels different from 13, and I think it feels a bit different from what it was in Austin as well. Is this basically a different path you are taking in development, in terms of the software itself, compared to the version you pushed to early adopters? When you push these new versions, are you looking for a significant improvement in intervention rates, or is the issue basically resolved, and it's more about adding features like parking and driving modes, or just overall comfort?

Unidentified Speaker:

No, when we release a major new software architecture for Autopilot, the top priority is safety. So it starts with safety, obviously safety first, and then we address comfort issues, which is why I don't recommend people use the initial version, and that's why I say most people should wait for version 14.2 before downloading version 14. Because by version 14.2, we will have addressed many comfort issues. The priority is very clear: safety first, then comfort issues. That's why most people might feel that it will be safe but a bit bumpy. We just need time to polish the rough edges while ensuring safety and addressing comfort issues for a major change in the Autopilot new architecture.

But it really is... I mean, I know the roadmap for Tesla's real-world AI, and the details are very specific. Obviously, Ashok is leading this project, I mean, I've spent a lot of time with the team studying in extremely detailed ways what we are doing to improve real-world AI. Like I said, this car will feel like it's a living being, which is achieved in AI4, even before AI5.

Elon Musk, Co-founder, King of Tesla Technology, CEO and Director:

Yes, this roadmap is super exciting, just like we are very much looking forward to releasing everything we are working on. In terms of the versions we deliver to customers versus the version for Robotaxi, they are mostly the same Clearly, customers have some features, such as the ability to choose where they want the car to park in a specific spot or lane, which is not particularly relevant for Robotaxi, but there are only a few small modifications like that. But most of the algorithms and architecture are the same between these two platforms.

Yes. But as I mentioned earlier, we will add reasoning capabilities in... I don't know, probably in version 14.3% or 14.4%.

Participant:

Yes. Or definitely within a year.

Elon Musk, Co-founder, King of Tesla Technology, CEO and Director:

Yes. So with reasoning capabilities, it will really think about... which parking spot to choose. It will say, this is the entrance, but in reality, if the parking lot is quite full, the probability of having an available spot at the entrance is very low. But in reality, it will simply drop you off at the store entrance and then go find a parking spot itself. It will become very smart at finding parking spots. It will discover... it will be better than humans at finding available spots. It has 360-degree vision, it will... yes, as I said, it will use reasoning to solve problems.

Participant

And putting all of this into a computer with A4 is the real challenge, which is the direction the team is working towards, because obviously, you can do reasoning on a server, which requires whatever, but in the car, you need to make real-time decisions, so putting it into the computer in the car is the challenge.

Elon Musk, Co-founder, King of Tesla Technology, CEO and Director:

Yes, that's why I say I have a good understanding of AI, from the giant model level of Grok to Tesla, I can confidently say that Tesla AI has the highest "intelligence density." When you examine the intelligence per terabyte, I believe Tesla AI may be an order of magnitude better than anyone else. It has no choice because that AI must fit into the AI4 computer. However, having that kind of AI intelligence density discipline, when you turn to something an order of magnitude more capable, like AI5, will yield tremendous returns. Now you have the same intelligence density, but the computer's capability is ten times stronger.

Operator:

Okay, the next question will come from Colin at Oppenheimer. Colin, please unmute when you're ready.

Questioner Colin:

Thank you very much, guys. I appreciate the challenges you mentioned regarding the dexterity of humanoid robot hands, as well as the complexity of the supply chain and the vertical integration you are pursuing. I just wanted to coordinate the timeline for starting production next year with the current state of the supply chain. It sounds like there is still quite a bit of work to do in terms of dexterity before you can really freeze the hardware design and start scaling production.

Elon Musk, Co-founder, King of Tesla Technology, CEO and Director:

Well, hardware design will not actually freeze even after production begins. There will be continuous iterations because you find that many things are very difficult to manufacture, and you only discover this late in the game. So we will make rolling changes to the design of Optimus, even after production starts. But I do believe that the new hand is an incredible engineering feat, we will actually have a production intent prototype ready in the first quarter (next year), possibly in February or March, to showcase, and then we will build an Optimus production line with an annual output of one million units, hoping to start production by the end of next year.

But that production ramp-up will take some time to reach the speed of one million units per year, as it will progress at the pace of the slowest, clumsiest, and unluckiest of the ten thousand unique parts. But it will reach one million units. Then, ultimately, we will do Optimus 4, which will be ten million units, and Optimus 5 may be fifty million to one hundred million units, I mean, this is really crazy. Yes.

Travis Axelrod, Head of Investor Relations:

Okay. Unfortunately, that's all the time we have for our Q&A session today. Before we wrap up, Vaibhav has some closing remarks.

Vaibhav Taneja, Chief Financial Officer:

Thank you, Travis. I want to take a moment to talk about an extremely important vote that will take place on November 6th. This meeting will shape the future of Tesla, and we ask you, as our shareholders, to support Elon’s leadership through two compensation proposals and the re-election of Ira, Kathleen, and Joe to the board. Please note that this is a team sport, and at Tesla, the board is an indispensable part of this winning team. Shareholders are at the center of everything we do at Tesla, and our special committee has developed a compensation plan, as Elon said, we don’t even want to call it a compensation plan.

Elon Musk, Co-founder, King of Tesla Technology, CEO, and Director:

Yes, the key is that I just want new voting control to give strong influence, but not so much that I can’t be fired if I go crazy. I think that number is around just over 20%. As a publicly traded company, there is no— we have explored all possible ways to have something like super voting stock, but after you go public, there really is no way to have super voting stock, but for example, Google, Meta, and many other companies have this. But they had it before going public. So it counts as, I guess, being exempted by the "grandfather clause." Tesla doesn’t have this. So, as I said, I just feel uncomfortable building a robot army here and then being ousted because of some (inaudible) suggestions from ISS and Glass Lewis, these clueless guys. I mean, those guys are corporate terrorists. The issue is, yes, let me explain the core issue here is that so many index funds, passive funds, all vote according to the recommendations of Glass Lewis and ISS Now, they have made many bad suggestions in the past, which would be extremely destructive to the company's future if those suggestions were adopted. But if you have passive funds, you essentially push the voting responsibility onto Glass Lewis and ISS. Therefore, if too many shares of listed companies are controlled by index funds, it could lead to extremely disastrous consequences for the listed companies. It is essentially controlled by Glass Lewis and ISS. This is a fundamental issue of corporate governance because they do not vote in a way that is truly beneficial to shareholders, and that is the big problem. I mean, ultimately, ISS and Glass Lewis are corporate terrorism.

Vaibhav Taneja, Chief Financial Officer:

Yes, I want to say that the special committee has done an excellent job in building this plan for the benefit of shareholders. Nothing will go through until shareholders receive substantial returns. So, in the end, I want to say that I urge you not only to vote for this plan but also to vote in favor of all three directors because of their exceptional knowledge and experience. Moreover, we at Tesla indeed work day in and day out with these directors. I mean, there is not a day that I haven't talked to one of the directors, or my colleagues haven't talked to them, and the directors here are not just reading from a PPT script; they work with us day in and day out. So again, I just urge you as shareholders to vote according to the board's recommendations. Thank you all.

Travis Axelrod, Head of Investor Relations:

Great. Thank you, Vaibhav. We appreciate everyone’s questions today. We look forward to talking with you next quarter. Thank you very much, goodbye