-- Audio Transcript (English) from https://youtu.be/suBQAv1lm4g

Below is a transcript I've created from the simplified version of the captions file with timestamps in the MM:SS format and text combined into full sentences where possible. The structure maintains the flow of the conversation, grouping related entries to form coherent sentences while preserving speaker attribution.

00:02 Emmanuel Huna: I'm in my Tesla app and I've selected my car, our Model Y. A very cool thing is that I can go here into Google Maps or Apple Maps and search for a place, then share it with my Tesla app. When I share it, it will send it automatically to the Tesla car and there it is in the navigation. Now I can just put my seat belt on, press a button, and Tesla full self-driving will take us there, awesome!

00:50 Elon Musk: Let me walk you through why I'm so excited about the future of Tesla. First of all, autonomy. The team and I are laser-focused on bringing Robotaxi to Austin in June. It will first be solved for the Model Y in Austin and then... Actually, we should parse out the term robotic taxi or Robotaxi and what's the Cybercab. We've got a product called the Cybercab, and then any Tesla, which could be a Model S, 3, X, or Y that is autonomous, is a robotic taxi. The vast majority of the Tesla fleet we've made is capable of being a Robotaxi. Once we have made the system work where you can have paid rides fully autonomously with no one in the car in one city, that is a very scalable thing for us to go broadly within whatever jurisdiction allows us to operate. We're solving for a general solution to autonomy, not a city-specific solution. Once we make it work in a few cities, we can basically make it work in all cities in that legal jurisdiction. If we make it work in a few cities in America, we can make it work anywhere in America. Once we can make it work in a few cities in China, we can make it work anywhere in China, likewise in Europe, limited only by regulatory approvals. This is the advantage of having a generalized solution using artificial intelligence and the AI chip designed specifically for this purpose, as opposed to very expensive sensors and high-precision maps of a particular neighborhood where that neighborhood may change, and then that car stops working. A general solution instead of a specific solution.

02:47 In regard to Optimus, we're making good progress. We expect to have thousands of Optimus robots working in Tesla factories by the end of this year (2025). We expect to scale Optimus faster than any product in history to get to millions of units per year as soon as possible. I feel confident in getting to a million units per year in less than five years, maybe four years. By 2030, I feel confident in predicting one million Optimus units per year, and it might be 2029.

03:31 With respect to energy, our energy business is doing very well. The Megapack enables utility companies to output far more total energy than would otherwise be the case. When you think of the energy capability of a grid, it's much more than its total energy output per year. If power plants could operate at peak power all 24 hours, as opposed to being at half power or sometimes a quarter power at night, you could double the energy output of existing power plants. To do that, you need to buffer the energy so you can charge up something like a battery pack at night and then discharge into the grid during the day. This is a massive unlock on total energy output of any given grid in the course of the year. Utility companies are beginning to realize this and are buying our Megapacks in scale. At this point, a 1 GWh class battery is quite a common thing. We expect the stationary energy storage business to scale ultimately to terawatts per year, very big numbers.

04:56 Q1. First quarters of a year are usually pretty tricky. It's often the worst quarter of the year because people don't want to go buy a car during a blizzard. We picked Q1 as a good quarter to do that cutover to the new version of the Model Y. We changed production of the world's best-selling car, remembering that the Model Y is the best-selling car of any kind on earth with a 1.1 million unit renewal output of a single model. We did this changeover at the same time in factories all across the world. Congratulations to the Tesla team on an amazing job pulling off what was a very difficult transition. That was very impressive work.

05:41 In conclusion, while there are many near-term headwinds for us in the auto industry, the future for Tesla is brighter than ever. The mission of the company is delivering sustainable abundance with our affordable AI-powered robots. I like this phrase: sustainable abundance for all. If you say, what's the ideal future you can imagine? That's what you'd want—abundance for all in a way that's sustainable, good for the environment. This is the happy future, the closest thing to heaven we can get on earth. Thank you again to the Tesla team for all their efforts during this challenging time, and I look forward to continuing to lead the team to great success in the future.

06:45 Thank you very much. Fantastic. Now we will move on to investor questions. We will start with questions from say.com. The first question is: what are the highest risk items on the critical path to Robotaxi launch and scaling?

07:06 Is that Ashok? We've got Ashok online. Let's disambiguate the Cybercab from Robotaxi once again. The Teslas that will be fully autonomous in June in Austin are Model Ys. That's currently on track to be able to do paid rides fully autonomously in Austin in June and then in many other cities in the US by the end of this year (2025). It's pretty difficult to predict the exact ramp week by week, month by month, except that it will ramp up very quickly. It's going to be like an S-curve where it's difficult to predict the intermediate slope, but you know where the S-curve is going to end up, which is the vast majority of the Tesla fleet being autonomous. That's why I feel confident in predicting large-scale autonomy around the middle of next year, certainly the second half of next year. I think there will be millions of Teslas operating autonomously, fully autonomously, in the second half of next year.

08:59 It does seem increasingly likely that there will be a localized parameter set, especially for places with very snowy weather, like the Northeast. It's kind of like a human—if you're a very good driver in California, are you going to be as good a driver in a blizzard in Manhattan? You're not going to be as good, so there is some value in a localized set of parameters for different regions and localities. But we can put that in the nice-to-have category, not the required category. The car is very much like a human—its digital neural nets and cameras versus humans' biological neural nets and eyes. The same strengths and weaknesses will be present.

10:47 Ashok: Speaking to the location-specific models, we still have a generalized approach, and you can see that in our deployment of FSD supervised in China, where with very minimal China-specific data, the models generalize quite well to completely different driving styles. That shows the AI-based solution we have is the right one. If we had gone down the previous rule-based solutions or hardcore HD map-based solutions, it would have taken many years to get China to work. You can see those in the videos that people post online themselves. The generalized solution we are pursuing is the right one that's going to scale well. You can think of these location-specific parameters like a mixture of experts. If you're familiar with AI models like Grok and others, they use this mixture of experts to specialize parameters to specific tasks while still being general. This makes the model use a limited amount of compute to solve for the reliability of tasks it has to solve.

12:02 In terms of addressing the question, what are the critical things we need to get right? One thing I would like to note is validation. Self-driving is a long-tail problem where there can be a lot of edge cases that only happen very rarely. Currently, we are driving around in Austin using our QA fleet, but it's super rare to get critical interventions for a Robotaxi operation, so you can go many days without getting any single intervention. You can't easily know whether you are improving or regressing in your capacity and need to build out sophisticated simulations, including neural network-based video generation. That's all happening in the background to make sure we deliver a safe product and are able to measure our safety even when we're driving around the block.

12:50 Elon Musk: In very basic terms, if we're seeing an accident every 10,000 miles, you have to drive 10,000 miles on average before you get into an accident or an intervention. People must be very worked up by the sheer number of Teslas doing circuits in Austin right now. It's weird looking, pretty bizarre. There's always a convoy of Teslas going all over Austin in circles. I can't emphasize enough, to figure out long-tail things, if it's one in 10,000, 20,000, or 30,000 miles, and the average person drives 10,000 miles in a year, try to compress that test cycle into a few months. That means you need a lot of cars doing a lot of driving to compress what would normally take someone a year into a month.

14:08 Ashok Elluswamy: If you haven't looked at those videos coming out of China, people are putting it to the real test—dark roads...

14:17 Elon: Those videos are amazing. Frankly, I think the Chinese consumer might be the most discerning consumer. Customers in China are awesome. They have a lot of fun with the cars. I saw one guy take a Tesla autonomously on a narrow dirt road across a mountain. That's a very brave person. He was driving along a road with no barriers, where if he makes a mistake, he's going to plunge to his doom. But it worked.

14:58 Thank you. If the question was on Cybercab itself, we're in B-sample validation now. We have our first big builds coming at the end of this quarter, within Q2. In the coming months, we'll start the large-scale installation of all equipment in Giga Texas, still on schedule for production next year.

15:25 Ashok: I just want to clarify because people don't understand—there's no new building being built for Cybercab. It's happening in the same factory, upstairs and along blinds, while we're still building Model Ys and Cybertrucks every day.

15:46 Elon: It's worth noting that the Tesla Gigafactory in Austin is three times the size of the Pentagon, including the ground zero garden. The Pentagon used to look big, but not anymore.

16:08 Thank you very much. The next question is: when will FSD unsupervised be available for personal use on personally owned cars? Before the end of this year, within the US. We do want to test... At Tesla, we're absolutely hardcore about safety. We go to great lengths to make the safest car in the world with the lowest accidents per mile and fewer lives lost. We want autonomy to be definitively safer than manual driving—not just as safe, but meaningfully safer. We want to be cautious with the rollout and not jump in at the deep end. With that said, I think we should be able to have it work in several cities later this year for personal use. The acid test is: can you go to sleep in your car and wake up at your destination? I'm confident that will be available in many cities in the US by the end of this year.

17:51 Thank you very much. That's unfortunately all the time we have for today. We appreciate all your questions and look forward to talking to you next quarter. Thank you very much and goodbye.