Elon Musk's annual speech at the 2024 Tesla Shareholders Meeting

On the early morning of June 14th Beijing time, Tesla held its 2024 shareholder meeting at the Gigafactory in Texas. This shareholder meeting attracted significant attention due to two important shareholder votes - re-approving Elon's 2018 compensation plan and relocating Tesla's registration to Texas.

The meeting first announced the voting results of all company and shareholder proposals, and as Elon had hinted the night before, both of these important proposals were successfully passed. Following that, Elon took the stage to give a speech. Elon appeared visibly excited and relaxed today. To show his gratitude to the shareholders, he started with a few jumping moves and a five-second awkward dance as soon as he got on stage.

Elon first ad-libbed a review of the progress of FSD autonomous driving and Optimus robots, then summarized the achievements the company has made in the past year in various aspects such as factory construction, automotive business, supercharging, 4680 battery cells, energy storage products, software development, AI hardware and software, and finally looked ahead to Tesla's path to becoming a $25-30 trillion company in market value.

01:48 FSD autonomous driving and Robotaxi

08:28 Optimus humanoid robot

12:16 Carbon emissions and factory construction

15:05 Electric vehicles

23:21 Supercharging network

25:44 4680 battery cells, lithium refining

28:07 Energy storage

30:38 Software development

31:55 AI hardware and software

37:14 Inference computing services

39:20 Bottlenecks of FSD

43:49 Optimus and the path to a $25 trillion market value

Full transcript of the annual speech:

Hello everyone! Welcome to the Tesla shareholder meeting!

First of all, I want to say, I love you all. We have the best shareholder base, it's incredible. It's incredible, comparable to any public company.

We are here to hold a grand shareholder meeting, to summarize the company's achievements and look ahead to our future. I believe we have an incredible future. We are not only opening a new chapter for Tesla, we are even starting to write a brand new history.

For those who have been closely following Tesla, you can understand. What's important is that you understand. I believe Tesla's future will absolutely shock people.

Clearly, we have made great progress in addressing sustainable energy issues, including electric vehicles, stationary energy storage, and solar energy.

1. FSD Autonomous Driving and Robotaxi

I believe, in terms of creating value for the company, autonomous driving will be shocking. As I have said, those of you here understand, but I think most people in the world still don't. When we talk to many large institutional investors, they are usually in cities like New York, where they don't drive themselves I ask, have you tried the FSD v12.3 autonomous driving system?

They said, oh, not yet.

I said, oh, well, you should give it a try, it's a good idea.

If you plot the points on the curve of autonomous driving progress, and then believe in the trend of this curve, it is advancing towards unsupervised FSD at an exponential rate, that's just how simple it is.

In fact, I want you to test it yourself, see how many miles you need to drive without having to intervene with each iteration, it's that simple.

With each version, you will find significant improvements, exponential progress. Clearly, it will reach a level of safety far beyond human driving.

I can foresee the situation for the next several versions, because basically, with each version hitting the road, we can more or less anticipate the next version and the one after that. We have a certain foresight into its development, which is truly amazing.

Without a doubt, its safety will far exceed that of humans, yes, without a doubt.

I have mentioned this before, but sometimes reiterating and connecting information is helpful, because sometimes people may wonder, how can profits be made from this huge fleet once autonomous driving is achieved?

In fact, in a way, it's like a combination of Airbnb and Uber: Tesla will own some cars, like Uber; and our customers will have their own fleet, like Airbnb.

You can add or remove your car from the fleet at any time. You can say, I want to leave for a week, just tap on the Tesla App, your car will join the fleet, and make money for you while you're away.

You can have your car in the fleet for a few hours, a few days, a few weeks. Whenever you want to take back your car, just click "come back", and it will return to you immediately.

I am confident that the owner's income will far exceed the actual monthly payment.

Obviously, Tesla will take a cut, but most of the money will belong to the car owner. This mechanism can really work, and it's about to happen very soon. Remember my words, it's just a matter of time.

I admit, sometimes I am overly optimistic. I am not completely unaware of myself, but if I am not optimistic, none of this would exist, and this factory would not exist either.

(Audience applause)

At this point, my brother is doing great, he's right here, just re-elected (as a director).

A little anecdote, when we were in school, my brother would tell me the wrong school bus time, so I wouldn't miss the bus. Then I would get mad at him, why would you tell me a time earlier than the actual arrival time of the bus?

He said, otherwise you would miss the bus.

This actually happened.

I think I have been somewhat pathologically optimistic since birth. Anyway, all of this is happening because of my somewhat pathological optimism. But I will eventually fulfill my promises, that's the key.

If I'm not mistaken, they predicted Tesla's valuation will exceed $5 trillion, considering only autonomous driving, not including Optimus I agree with their statement.

(Audience applause)

I believe that by only considering autonomous driving for cars, we can increase the company's value by 10 times from what it is now. I believe all of this will happen.

Who knows, it's only 4:20 in the afternoon now, and I just noticed.

(Elon smirked)

II Optimus Humanoid Robot

I think this prediction is quite accurate, but if we consider the Optimus robot, it becomes even more crazy.

It is a humanoid robot that can do anything you want it to do. It can be your companion, stay at home, take care of your children, teach them, be a good teacher, and work in factories.

How many super useful humanoid robots do you want to have?

Who wouldn't want a C-3PO robot from Star Wars? One C-3PO, plus R2-D2, and several other robots would be fantastic.

I believe that everyone in the world would want a robot, absolutely everyone. And obviously, robots will also be involved in production in the industrial sector. I think the ratio of humanoid robots to humans could be at least 2 to 1, and it could definitely reach 1 to 1.

In other words, there may be around 10 billion humanoid robots, maybe 20 or 30 billion.

Now, let's assume the production speed... I think the production speed could reach the scale of producing 1 billion humanoid robots per year, even if Tesla only accounts for 10% of that—probably much more—then that would be producing one million Optimus robots per year.

For reference, the automotive industry produces approximately 1 billion vehicles per year, at least in terms of quantity.

I believe that at a very high production scale, the cost of producing a robot could be around $10,000, cheaper than a car. If sold for $20,000, in the case of large-scale production, Tesla could make about $1 trillion in profit each year.

If the P/E ratio is 20 or 25 times, it means that just Optimus alone could bring a market value of $20 trillion. And the market value of autonomous driving cars could be between $5 to $10 trillion.

Therefore, it is conceivable and possible for Tesla's market value to reach 10 times that of the current highest market value company. When I say we are starting to write a brand new history, I mean, this will be the best history, creating a brand new history.

We must ensure that these robots are very friendly to us, this is very important.

III Carbon Emissions and Factory Construction

In conclusion, with that said, let's begin my speech BlahBlahBlah, I think we should set a word limit for future Tesla shareholder proposals. We can't allow proposals as lengthy as novels, we can't allow proposals for Tesla to be written like a whole novel.

As you can see, our impact is accelerating, we are beginning to make a truly significant contribution in carbon emissions, and we have made great progress in achieving sustainable development goals.

One of the impacts brought by autonomous driving is that it will have a greater impact on reducing carbon emissions. Because I speculate that we will increase the usage of passenger cars from about 10 hours per week, which is about 1.5 hours per day, to one-third of the time per week, possibly 50 or 55 hours.

This means that the frequency of use of the same car will increase fivefold.

We need a certain amount of resources to produce cars, but their utilization rate will increase by five times. This is very likely to happen, which will have a much greater impact on carbon emissions.

We attach great importance to sustainable production, our factories are beautiful. You will know when you walk around the factory, in fact, contrary to media reports, the atmosphere in the factory is very good, people are smiling, happy, and friendly.

Yes, at Tesla, we focus on doing the right thing. We cannot be perfect, but we focus on doing the right thing. Our vehicles are water-saving, energy-saving, we use renewable energy sources as much as possible to power the factory, and we make every effort to do the right thing.

Our batteries have a longer lifespan, which is great.

The safety conditions in the factory are good.

Four Electric Vehicles

We are increasing the affordability of electric vehicles. Compared to Mode Y and BMW X3, the actual cost of ownership per month is much lower.

Last year, we ranked first in the global electric vehicle field.

![](https://wpimg-wscn.awtmt.com/ca0ba639-b37a-411b-81cf-6b347167955c? Congratulations to the Tesla team for creating a historical milestone of producing a total of 6 million cars! The Tesla fleet is gradually becoming extremely large, reaching over 7 million by the end of this year, exceeding 7 million.

We have a giant car factory in the San Francisco Bay Area, which is actually a very crazy thing. It is not the most cost-effective place to operate a car factory, it's like the Swiss Alps, but we can still produce excellent cars at a high volume. This also proves how excellent our team in Fremont is, congratulations to the Fremont team!

Many people used to say that the Cybertruck was fake and would never come to market, but now we are delivering a large number of Cybertrucks. We set a record of 1,300 units per week.

I think the Cybertruck is really special.

Sometimes people have different opinions about the Cybertruck, but if you really want to know if something is cool, if it's a good product, just show it to children without any filtering. Children won't filter their opinions. Ask a five or six-year-old, even a three-year-old, what car they like?

Cybertruck!

Then you will understand. Finally, it makes the future look like the future, and it's easy to drive, comfortable, and a great product. I think this is our best product.

It has many innovations, ultimately replacing the 12-volt with a 48-volt low-voltage system, an 800-volt drive system, the world's largest castings, bulletproof, and all the cool features. Its towing capacity surpasses the F-350 diesel pickup, which is an impressive pickup. It pulled a Porsche 911 and beat it in a quarter-mile race, it's insane.

Think about it, how many companies can create products that are truly special and captivating? It's so rare, I think the Cybertruck is one of them.

We have launched a refreshed version of the Model 3, which is a great car. I recommend everyone to try it, it's really a great car.

After saving on gasoline costs, the total monthly ownership cost is only $216. Compared to similar gasoline cars, its monthly cost is basically just over $200. It is indeed cost-effective and also a fun car. The performance of the high-performance version surpasses that of the Porsche 911, it is truly a great car.

Of course, the Model Y has become the best-selling car globally We predicted this result. As I mentioned before, in 2022, in terms of sales, the Model Y will become the best-selling car globally; then by 2023, it will become the car with the highest single-unit sales volume, and indeed it has. This year, it will once again become the best-selling car globally.

Next is the Tesla Semi, we are in the process of small-scale production. Just last week, I approved the mass production plan for the Tesla Semi, and we will produce a large number of Semi trucks. Yes, I believe it will eventually have a financial impact, which is no small matter.

The feature of commercial vehicles is that companies using commercial vehicles are very objective. Obviously, we want it to have a cool style and be a great car, but for companies making purchasing decisions for commercial vehicles, they will only look at it and ask, what about the numbers? Is its transportation cost lower or higher compared to diesel trucks?

The most important thing is that its economy is much better than diesel trucks.

This is basically a no-brainer. If you are a transportation company and you don't choose to use the Tesla electric Semi, you are losing money, why would you choose that? Don't you like money? Right. If you like money, I suggest you use the Tesla Semi.

I believe the actual sales volume of this vehicle will be surprising, it will drive economically and bring many benefits. In addition, it can also reduce carbon dioxide emissions, help sustainable development, because Semi trucks are always on the road, even though there are far fewer vehicles, they still contribute significantly to reducing total carbon emissions.

Obviously, under the white cloth cover, there are some new products we are developing. I think these products will be very special, some of them may not seem so magical at first, but wait and see, they will be in the end.

V Supercharging Network

Our Supercharging network is still expanding. Rumors of the demise of Supercharging have been greatly exaggerated, in fact, our Supercharging network continues to grow significantly. This year, the additional deployment of available Superchargers will exceed the total of other companies in the industry, just for reference.

However, we will pay more attention to where deploying Superchargers can bring higher capital efficiency. But one thing is certain, wherever there is congestion, wherever there are missing parts on the map, we will install Superchargers there.

In the remaining time this year alone, we expect to spend about $500 million on Supercharger deployment. This will be a considerable and worthwhile $500 million expenditure, not a small expense.

In fact, this does not yet show the series of new markets we will be expanding into later this year and next year. We hope Tesla will go global, and in many countries, the number of Teslas is zero or very few. Obviously, we want Tesla to be everywhere, so we will open many new markets later this year and next year We will also open up super charging stations to other companies.

That's a clunky adapter, and our goal is to help other car companies. We hope to help other car companies electrify through our supercharging network. We provide them with adapters so they can use the Tesla supercharging network. We believe that doing this is better for the development of electric vehicles than building high walls to suppress them.

Six 4680 Battery Cells, Lithium Refining

In addition, we have made a lot of innovations in battery production, and 4680 is produced here. This is a challenge, you know, many companies only produce battery cells, but we are making good progress, all the Cybertrucks you see are using Tesla's 4680 battery cells.

We believe that from a production efficiency perspective, 4680 may be the most competitive battery cell. But it has to be said, this is a challenge, mastering the production of battery cells is very brain-intensive.

But I think, once the geopolitical situation changes, it will allow Tesla to adapt freely. Having a certain degree of independence in battery cell manufacturing is very good.

Well, we are making steady progress, we also have a cathode refining factory right behind the main factory. You can see it in this picture, that's the cathode refining factory.

Then, we also have a lithium refining factory in South Texas.

We are just ensuring that we have the complete puzzle. Watch the video, the way Tesla refines cathodes and lithium is completely different from other companies in the industry. You can eat at the Tesla refining factory, but I wouldn't recommend doing that at other factories.

We are undergoing deep vertical integration, which is a wise investment that will yield higher returns than people currently realize.

Seven Energy Storage

This year, we will also complete a large amount of energy storage deployments, with a significant increase in deployment volume. Our energy storage deployments and fixed battery packs seem to be on track for annual year-on-year growth of 200% to 300%, which is a huge increase.

The real constraint lies in whether we can produce more megapacks and more powerwalls. We are ramping up the production capacity of Powerwall 3, which is a disruptive product for individual users.

Any technology usually needs three iterations to achieve the effect of "okay, now it's really awesome." Powerwall 3 is an epic product.

I think Megapack 2 is also an epic product. Megapack 3 may still be a few years away, through it, we will start to include more and more substation power electronic equipment. I think we want to develop megapack to the point where you can directly plug in high-voltage wires, no substations needed, just plug in the high-voltage wires, deploy it, plug it in, and it's ready to use By the way, this is simply incredible for the power industry. They would say, what?

Yes, as long as you plug in the wire, it can operate at very high voltage.

In any case, the work of the energy storage team is excellent.

The services and other businesses are also very good. We are actually making money now in services and other businesses, which is crucial.

Eight Software Development

Tesla is obviously not just a car company. We have developed a lot of software at Tesla, probably about half, or a large part of the engineering work we do is actually software engineering. Therefore, I believe that Tesla's degree as a software company is not inferior to a hardware company.

This is important because car companies are usually not software companies, so this is important.

For example, our auto-bidder energy storage software, all software that controls cars, megapacks, powerwalls, and solar energy. Obviously, artificial intelligence and FSD are important, as well as insurance and collision services.

Tesla has also developed a lot of software internally, which we call "Tesla Operating System," which is much better than any other company's internal software, I think possibly better than any Fortune 500 company. Tesla's internal software is much better. Yes, this far exceeds what people usually imagine for a car company.

Nine Artificial Intelligence Software and Hardware

Tesla is also a leader in real-world artificial intelligence, which is important.

Tesla is ahead of any company like Google, Meta, OpenAI in real-world software. We can accept videos and make decisions based on videos, which is unparalleled. I dare say that over time, our artificial intelligence will become stronger and stronger.

It is worth mentioning that Tesla is also quite outstanding in chip design.

For the artificial intelligence inference chips used in Tesla cars, we previously equipped them with the HW3 artificial intelligence inference chip. In the past year, our cars have been equipped with HW4. We have just completed the design of HW5 - now called AI 5.

Because in fact, there is still no chip from NVIDIA - I have great respect for NVIDIA - or any other company that gives us the motivation to replace our current onboard chips, that can match our onboard chips.

In terms of chip design, we started from scratch, just like we developed artificial intelligence software from scratch. We went from nothing to having the best real-world artificial intelligence software in the world, and the best artificial intelligence inference chips, which is important The capability of the in-vehicle chip is enormous.

Currently, we are running HW4 in simulation mode based on HW3. We will continue to make significant progress on HW3, but later this year, we will split off to continue training for HW3, and then conduct separate training for HW4.

That is the training cluster we built on the south side of this super factory, which will be dedicated to HW4 video and inference.

The camera resolution of HW4 will be four to five times higher than HW3. Depending on the calculation method, HW4 will be three to eight times better than HW3.

But what you are seeing now is still only HW3. We still have a long way to go before reaching the limits of HW3. I believe HW3 will still have amazing performance, but HW4 may be up to 5 times better than HW3.

HW5 is expected to be launched in about 18 months, with performance 10 times that of HW4. If you are familiar with NVIDIA hardware, it will be equivalent to a B200-level computer.

The ongoing work we are doing is to improve the reliability of cars by how many nines.

Of course, this will also be applied to Optimus. HW4 is already installed in Optimus, and AI5, namely HW5, will be installed in Optimus and all cars in about 18 months. Its computing power is extremely amazing and very energy-efficient.

In mobile applications, such as humanoid robots or cars, you cannot consume 10 kilowatts of power like in a data center, you must be very energy-efficient.

Both HW3 and 4 are only a few hundred watts. Yes, it's very challenging. The power of HW5 may reach 700 or 800 watts, but its power will increase depending on the complexity of the scenario. If it is stationary in a parking lot, then you don't need to think too much, just like a person. If it is in a complex traffic scenario, you have to think more than when driving on an open road.

Ten Inference Computing Services

But I think at some point in the future, the number of Tesla vehicles in the fleet may exceed one billion, which is very interesting.

If each car has a kilowatt of efficient inference computing power, I think there is an opportunity here similar to Amazon AWS services. Because if one billion cars all have a kilowatt of efficient inference computing power, then you have 100 gigawatts of computing power. 100 gigawatts of computing power is extremely huge and distributed worldwide.

When the car is not in autonomous driving mode—I think maybe 50 to 60 hours a week are spent on robotaxi work, but there may be 100 hours a week when it is stationary. In this way, there will be 100 gigawatts of inference computing power for 100 hours a week, and I think we should make use of it Why not?

You see, look at Amazon, it started out as just an online book retailer. Now it has grown into an amazing site where you can buy anything.

And then, all the computers for Amazon's AWS services sometimes experience peak usage. What if it's not peak time? Sometimes, Amazon servers shut down by 10%. Then, the Amazon AWS service becomes more valuable than all other businesses of Amazon.

In conclusion, I think this is a great opportunity that Tesla will have in the future. Similarly, no one really considers this, but I think this opportunity is quite substantial.

The bottleneck of FSD in November

In terms of training, we are no longer limited by computing power.

I discussed with the team, in terms of training and reasoning, what methods can speed up progress.

Currently, this is not a limiting factor. The current limiting factor is that the distance between two interventions is too long. It takes a considerable amount of time to determine which version is more effective because they do not require much intervention. If the intervention interval mileage reaches thousands of miles, or ten thousand miles before an intervention, then, on average, a person drives about ten thousand miles a year.

In urban environments, with an average speed of 20 miles per hour, to be honest, our professional test drivers find it boring. They would say, okay, I drove for a whole week without any intervention. And the highlight of this week is, yes, finally there was an intervention!

It has come to this point.

In this regard, having a large fleet of vehicles is particularly important. Because we can deploy a new FSD model and run it in shadow mode to see how it performs. Compare the performance of human-driven vehicles with the new FSD version, and then analyze the differences in shadow mode. Then, evaluation can be done by driving billions of miles in a short time with a large fleet of vehicles.

Fundamentally, this data engine is extremely helpful. In fact, it is impossible to solve the problem of autonomous driving without millions of vehicles on the road.

In conclusion, finding clever ways to test the performance of the next FSD version is actually a current limiting factor.

Of course, as I said, we are building a new training data center here specifically for the training of HW4. We will later, later this year, divert the training of HW3 and HW4. We will continue to improve HW3, but also fully unlock the capabilities of HW4.

I think most of you have tried the v12 version. We did say that unsupervised FSD will become the v12 version. We actually just arbitrarily designated the version number as 12, and then called it 12.3, 12.4, 12.5, but in fact, it is more like v13, v14 In short, this is just a randomly assigned version name. Version 12.4 is completely different from 12.3, and 12.5 is completely different from 12.4. Between two adjacent versions, you will see huge improvements, sometimes even tenfold.

As I said, for the existing fleet, you can have your car join or leave the fleet as needed. Then, there will also be cars owned by Tesla. This will be a combination of Uber and Airbnb models.

After a software update, the entire fleet suddenly becomes available, which will be very wild. Suddenly, there are 7 million cars, eventually tens of millions of cars that can drive automatically, no longer just using 10 hours a week, but 50 or 60 hours.

Twelve Optimus and the Road to a $25 Trillion Market Cap

A humanoid robot walks along the workstations in Palo Alto, which is quite crazy.

In a very short time, we have made significant progress on Optimus. From dancing in a robot suit to robots that can perform useful tasks in the factory today.

We have two Optimus robots in the Fremont factory, basically performing the task of taking the battery cells off the end of the production line and putting them into a container.

Yes, there are many such robots walking around in our Palo Alto office.

I think we will have a major hardware upgrade, which should be completed by the end of this year or early next year. Then, we will start limited production of Optimus next year.

Limited production is to be able to use it in our factory to test this robot. As the saying goes, use your own products. But I think next year we will have over a thousand, maybe several thousand Optimus robots working at Tesla.

From then on, the scale will expand rapidly. We will solve problems, and the level of automation will also greatly increase. By then, you just need to tell it, please complete this task; or say, let me demonstrate it to you, and then follow what I do. You can even let it watch a human video, learn from watching the video, and complete the task. It will be amazing.

And I believe the ratio of humanoid robots to humans will exceed one to one. By then, there will be over 10 billion humanoid robots in the world, maybe even over 20 billion.

And Tesla will be far ahead.

You will see many robot startups, but I think the challenge of producing Optimus as a robot startup is huge. Because we found that to make Optimus work, you have to design every part of the robot from scratch using first principles: motors, gearboxes, transmissions, power electronics, communication systems, everything from scratch.

We found that there is basically no supply chain. Although there are many motors in the world, there is no supply chain that can provide the motors, sensors, and gearboxes needed for humanoid robots What you are seeing now is the hand and arm of our current generation of robots, which has 11 degrees of freedom. Our next-generation product will have 22 degrees of freedom, and it will also be able to play the piano, wow!

Of course, as I said, we must ensure that a Terminator-like situation does not occur, this is very important, the safety of humanoid robots is very important.

But because it requires a lot of basic design, designing each motor, gearbox, sensor, and power electronics device from scratch, it is very difficult for startups, even impossible.

But at Tesla, we have the best electrical engineering capabilities in the world. I believe that in terms of gearboxes, motors, and power electronics, we have the best mechanical engineering capabilities in the world. We have enough resources to do this, and they are well applied.

And then, you also need a brain. You need a high-performance reasoning computer, which our cars are equipped with, and Optimus will also use. You need to be a leader in real-world artificial intelligence, and Tesla is the leader in real-world artificial intelligence.

To produce attractive robots, you need all these powerful conditions.

And you also need to be very good at large-scale manufacturing. In order to prevent the cost of robots from soaring to tens of thousands of dollars, and to produce them at a cost of one or two thousand dollars, you actually need to design manufacturing systems and be very good at manufacturing.

In my experience, prototype design is easier than mass production. Relatively speaking, prototypes are easy, production is difficult.

Tesla has production capabilities, engineering capabilities, and artificial intelligence software and hardware capabilities. Even the most optimistic estimates I have seen for Optimus—the optimists about Optimus—I think underestimate the capabilities of this robot.

As mentioned at the beginning of the speech, I agree with ARK Invest's analysis that autonomous driving will bring about a market value of 5 to 7 trillion dollars.

I believe that Optimus will bring about a market value of 25 trillion dollars.

I do not want to underestimate the effort required to achieve this goal. To reach this goal, a lot of work needs to be done, and the difficulty is extremely high. But we are rapidly advancing on this path, and we will definitely achieve this goal, thank you!

(Above)