Elon Musk participated in a virtual conversation at the annual Baron Capital conference on November 14.
Unlike a typical roadshow, his dialogue with founder Ron Baron resembled a retrospective between old friends: during the toughest moments of acquiring Twitter (X) three years ago, Baron proactively called to offer financial support; even when the valuation was slashed by 70% that day, he still chose to “stand by.”
It is precisely this mutual support during challenging times that prompted Musk to express heartfelt sentiments during the exchange: true friends are those who continue to support you when you are at your lowest point and surrounded by criticism.
For Chinese investors, Ron Baron is a name that carries a low-profile yet legendary reputation. He founded Baron Capital in 1982, which now manages approximately $45 billion in assets and is renowned for its “patient holding and growing alongside exceptional entrepreneurs.”
The flagship product, Baron Partners Fund, has achieved an annualized return of about 15% since 1992, significantly outperforming benchmarks. This consistent outperformance across cycles is fundamentally rooted in deep research into entrepreneurs and technological pathways, coupled with long-term commitment.
Baron’s most notable investment was a substantial $400 million bet made between 2014 and 2016 in $Tesla (TSLA.US)$ ; despite subsequent extreme volatility, he personally “did not sell a single share,” accumulating unrealized gains of approximately $8 billion.
Even though he disclosed moderate reductions in client accounts this year, he still projects a long-term target price of $2,500 for Tesla shares within ten years and candidly admits that about 40% of his net worth is tied to Tesla.
In a recent interview with CNBC’s Becky Quick, he stated, “I have no intention of selling my Tesla or SpaceX shares in my lifetime.” Because “others invest in the technology industry, but we invest in technological talent — the world’s top engineers!”
Therefore, in this dialogue, Ron’s questions and follow-ups were particularly 'in tune with Musk': not only discussing Grok 5, xAI, and the open-source vision of the 'Galactic Encyclopedia,' but also probing into the hardest topics — the AI5 inference chip, whether to build a massive wafer fab, and the safety data and promotion strategy of FSD in the real world.
When Musk talked about the competition in AI, he also said, 'The key is, how can we do better than others? I think it mainly depends on three things: Can we attract the top talent; can we deploy the most AI hardware; and can we significantly outpace others in the ramp-up speed of GPU computing resources?'
Having watched many interviews and dialogues where Musk faces the media and the public in various states, this one was a particularly immersive exchange — relaxed, sincere, and covering everything.
Ron, as always, exuded the calmness, rigor, and wisdom of an elder.
One uses the patience of capital to uphold long-term value, while the other uses engineering persistence to expand human boundaries. Their mutual achievements are themselves the most touching footnote of this dialogue.
Smart Investors has meticulously translated and organized this content to share with everyone. Enjoy it~~
01. The total number of robots globally may eventually reach 30 to 40 billion.
Host: I have a question I really want to ask. When I was talking to my assistant about the production capacity planning for the Optimus project, I mentioned that next year we plan to produce one million units, then ten million, one hundred million, and even one billion units.
After hearing this, she asked me, 'Where will all these robots go? Will there be enough space on the sidewalks? Are you going to build special channels for them? Is this the real reason you founded the Boring Company? To let the robots move underground? Or will they fly in the sky?'
Elon Musk: Actually, you could fit all of humanity into a single city. Optimus is small in size and high in density.
So imagine this: all 8 billion people on Earth could fit into the area of a single floor in New York City if they were just standing.
We used to play a game as kids, like joking about dropping a water balloon from a plane, but you'd never hit anyone because there was just so much empty space down there.
Though I actually tried it once in middle school and ended up hitting someone (laughs). So, Caitlin (my assistant), don't worry—there's definitely enough space.
Host: You mentioned that Optimus will be incredibly capable. But when robots can do everything, what is left for humans? In that future of abundance, besides just 'existing,' what roles will we have?
Elon Musk: This is precisely the core question—when robots can do everything, how will humans lead meaningful lives?
In fact, we are already seeing some insights. Even though machines outperform humans, people still want to participate in many activities, such as sports competitions or chess games.
Take chess, for example. Modern computers are now so advanced that even an offline AI program on your phone can easily defeat world champion Magnus Carlsen. Yet, chess is more popular than ever.
Thus, just because machines are powerful doesn’t mean humans become irrelevant.
We engage in many activities not because we are the best at them, but because we enjoy them.
Now, back to Optimus. When you see what it can do, you might wonder: with only 8 billion people on Earth, why would we need to build 100 million, or even tens of billions, of robots? What will these robots do?
In reality, there won’t be an overnight production of such quantities; it will be a gradual process. However, I believe that ultimately we will indeed have billions of humanoid robots.
Consider this: who wouldn’t want their own R2D2 or C-3PO (robot characters from Star Wars)? And even better, a smarter and more practical version?
A 'good companion' robot that can help take care of your children, walk the dog, go grocery shopping, chat with you, and protect you when needed—wouldn’t that be great?
In industrial settings, such robots will also see large-scale application in manufacturing goods.
A rough estimate suggests that for every human in an industrial scenario, there would be three to four robots. Thus, the global total number of robots may eventually reach 30 to 40 billion units.
Host: Three to four hundred billion robots—it really does sound like a lot! Companies like Japan’s KUKA, which make industrial robots, sell one robot for fifty thousand, a hundred thousand, or even one hundred fifty thousand US dollars.
And you mentioned that the target price for Optimus is twenty thousand dollars per unit. Does that mean the prerequisite is to first achieve annual production of one million units, then scale up to ten million units, in order to bring the cost down to that level? At this price point, would ordinary people be able to afford it? Would it be a rental model, or direct purchase? Would companies buy them and operate them, and we take a share of the revenue? Do we currently have a clear business model?
Elon Musk: My current rough estimate is that once Optimus reaches stable annual production of one million units, it would take about another year to bring the per-unit manufacturing cost (i.e., materials plus labor) down to between twenty and thirty thousand US dollars (at current currency value).
I think this figure is fairly reliable.
02, “The Six Million Dollar Man” is becoming a reality.
Moderator: It sounds similar to what you did with cars. Doesn't Tesla also break down the cost of every single component? Sometimes, if suppliers' quotes are too high, you just go ahead and make the parts yourselves. Is it the same for robots? Aren’t they much simpler than cars? Or am I misunderstanding something?
Elon Musk: Not exactly. In fact, the most complex part is the “hand.”
Moderator: The hand? You mean the arm?
Elon Musk: More precisely, the palm and forearm. One hand alone contains about 50 actuators.
Moderator: By actuators, you mean combinations like motors, gear sets, and power modules, right?
Elon Musk: Yes. Each Optimus robot has approximately 100 actuators, plus a whole set of sensors. The overall complexity is actually quite high.
Moderator: Then why is the “hand” so complex?
Elon Musk: Because to achieve true dexterity, the robot’s hand must possess the same level of sensitivity, precision, and degrees of freedom as a human hand.
Actions that we find simple—like picking up a screwdriver, turning a wrench, threading a needle, or even playing the guitar—all require an extremely high level of fine control.
This is also one of the reasons we introduced the new goal of 'sustainable abundance.' It represents an upgrade to our previous goal of 'accelerating the transition to sustainable energy.'
Because in terms of energy, we have already largely achieved our goals. The next task is to ensure that everyone can enjoy a truly abundant life, and this must be done without damaging the natural environment or destroying wilderness ecosystems.
Host: But that raises a question—how do we allocate resources? If someone wants to live in my house, can they just move in?
Elon Musk: Uh... your house is indeed very nice, so I can understand why people would want to (laughs).
But seriously, in the future, robots will be able to build houses for anyone. As long as you’re not fixated on living in a specific location, it could even build you a castle.
Host: So you insisted on making the 'hands' so complex in order to achieve this level of precision?
Elon Musk: Exactly. For example, surgeries and various high-precision medical procedures—Optimus will be able to handle them all in the future.
Imagine this: everyone could have access to top-tier surgical care, truly democratizing healthcare for all.
The precision of Optimus’s hands will match or even surpass that of human hands, enabling it to perform complex medical tasks that are difficult even for human doctors. The key point is that this capability will be accessible to everyone.
People often talk about eradicating poverty and providing universal access to quality healthcare... but they lack a real technological pathway to achieve it. Even if you have money, you can’t simply buy more good doctors because they are a scarce resource—you can't just place an order for them.
But robots can be continuously produced in factories.
Host: I remember sending you an article from Barron's a year or two ago about a fund manager who was 33 years old at the time and a Paralympic athlete. However, he contracted flesh-eating bacteria and lost both legs. In the interview, he asked, 'Is it possible to get me out of the wheelchair?'
You replied then, 'Yes, in three to four years, we will be able to equip him with an Optimus body and connect his thoughts using a brain-computer interface chip, allowing him to dance, sing, walk, and run like an ordinary person.'
Has there been any progress in this area now?
Elon Musk: Yes. This is actually the collaborative result of two companies: Neuralink and Tesla.
Neuralink has made significant progress, with over ten patients having completed the implant procedure. These individuals were originally unable to move their limbs, some even in a 'completely locked-in' state, similar to Stephen Hawking.
Now, however, they are able to communicate, almost at the same speed as our conversation here.
This is truly remarkable, and the progress is accelerating.
We can now enable Neuralink to receive signals from the motor cortex of the brain while providing feedback to the somatosensory cortex.
Therefore, if someone has lost both legs, we can equip them with Optimus legs.
In a sense, we are truly realizing the 'Six Million Dollar Man' from that sci-fi series. Have you seen that show?
(Note: "The Six Million Dollar Man" is a classic American sci-fi TV series from the 1970s that tells the story of a former astronaut who, after suffering severe injuries in a plane crash, was transformed by the government into a bionic agent at the cost of six million dollars. Equipped with mechanically enhanced arms, legs, and eyes, he possessed superhuman strength, ran like the wind, and had extraordinary vision, becoming a high-tech human capable of executing secret missions. This character inspired subsequent cultural imaginings of cyborgs and also served as an analogy referenced by Elon Musk to enable paralyzed individuals to walk again—or even run faster than ordinary humans—through brain-computer interfaces and robotic prosthetics.)
Host: Of course!
Elon Musk: I've watched it too, and I think the show was really great.
Today, we can actually create similar high-tech prosthetic capabilities for some people akin to 'The Six Million Dollar Man,' and the cost would be less than six million dollars.
After all, six million dollars was an astronomical figure back then, but today... it's not as significant.
Now? Perhaps sixty thousand dollars could achieve something similar. And it would be practically usable.
We can use Neuralink to capture the brain’s command to 'move the legs' and transmit it directly to Optimus’ robotic legs. This would enable the person to run faster than an average human.
This is like the real-life version of 'The Six Million Dollar Man.'
03, Regarding X and Baron’s Investment History
Host: This is incredibly exciting—it feels like science fiction coming to life!
Alright, let’s change the topic and talk about X.AI. Three years ago when you visited, you had just purchased, or were about to purchase Twitter (now renamed X). You faced a lot of criticism back then.
Elon Musk: Yes, I remember... I got heavily criticized. (Laughs)
Host: At that time, you joked on stage, saying that so many people already wanted to kill you, and you didn’t want to provoke any new enemies.
Elon Musk: That’s true.
Host: When you bought Twitter, your offer was $42 billion. You were raising funds. I reached out to you—it was me who called you, not the other way around. I told you I wanted to invest $100 million: $60 million from our fund and $40 million of my own money.
You seemed surprised: 'Really?' I said: 'Yes.' You asked again: 'Are you sure?' I replied: 'Yes.'
At the time, you told me this investment could potentially double in value. I said, 'Hopefully.' But in my mind, you had already helped us make $8 billion through Tesla. It would have been unreasonable if I didn’t support your next move.
So we invested. But on the very day the funds were transferred, the valuation was immediately written down by 70%.
Elon Musk: (Laughs) Wow... That’s why I’ve always considered you a true friend. Truly, you’re someone I trust. What is true friendship?
It’s when someone stands by you during your toughest and most isolated moments. And that’s exactly what you did, Ron!
(Applause from the entire audience)
Host: Thank you. But this story is not over yet; the best part is yet to come.
Our $100 million investment, which on paper dropped from $100 million to $30 million. However, after about a year, hedge funds suddenly started reaching out to me—these guys always seem to know things they shouldn't.
They said, 'I want to buy your shares at the price you originally paid.' I replied, 'I think I'll wait.'
So I never sold them.
Then you renamed Twitter to X and completely restructured the business. It’s no longer just a social platform—you integrated it into a mega social network.
Suddenly, we had an enormous data engine: 600 million people engaging in online conversations, interactions, and content creation every day.
This isn’t just digital-level data—it's real-world behavioral data at a physical level.
And then you used this data to train your own large model—Grok. Other companies don’t have access to this kind of data; all they have are dry results from web crawlers. What you possess is the most authentic, real-time social behavior data of humans.
Next, to train AI models, we needed larger computing power centers. And you actually went ahead and built one. In just a few months, you constructed one of the largest data centers in the world, four times the size of those operated by other companies.
While others utilized 25,000 CPUs, you directly employed 100,000 GPUs, resulting in a tenfold increase in performance. The plan is to continue expanding, with the goal of building hundreds of data centers at this scale.
To date, we have invested a total of $350 million, and the current valuation of this investment has reached $700 million.
Almost every project you participate in yields incredible outcomes.
Others invest in the technology industry, but we invest in tech talent—the world’s top engineers!
I would like to follow up on a question: When you bought Twitter (now called X), was it because you valued its data? Did you foresee these developments at that time?
Elon Musk: Not really. My purchase of Twitter wasn’t for the data; rather, I believed it was having a negative impact on civilization, promoting some 'anti-civilization' elements.
Frankly, I felt that Twitter had already been dominated by far-left forces at that time. Perhaps they didn’t see it that way themselves, but to me, it represented a very typical 'San Francisco-Berkeley style' extreme political inclination, the furthest left in the United States.
As a result, it was no longer a public square that encouraged debate. Many people holding right-leaning views were banned, including even the sitting president at the time—a situation unprecedented in human history.
Therefore, I believe we need a platform that truly upholds freedom of speech.
Freedom of speech is the cornerstone of democracy. If people cannot freely express differing opinions, they will not be able to access truthful, diverse information; without information, rational judgment becomes impossible. At that point, what is called democracy is no longer true democracy.
My original intention in acquiring Twitter was actually to restore its neutrality.
Moreover, I need to clarify that we are not suppressing left-wing voices or banning their right to speak. We simply want this platform to treat every voice fairly, making it a space for rational exchange rather than a breeding ground for hatred.
That’s why I’ve always said that freedom of speech is the fundamental safeguard of a democratic system, which is precisely why it was enshrined in the First Amendment of the U.S. Constitution.
Because at the founding of the nation, people escaped from countries where they would be executed or imprisoned for speaking out. And today, such things are still happening in many parts of the world, even in democratic countries like the UK.
For me, this represents a systemic risk at the level of civilization, and I simply did what I believed someone had to do. If the United States can't hold on, then what's the point of doing business? The U.S. is the pillar of Western civilization; if it collapses, everything else will crumble too.
04. In the end, what the technology sector competes on is acceleration.
Host: You were one of the co-founders of OpenAI, the company behind ChatGPT. At the time, you established it as a non-profit organization to uphold certain values you deemed crucial for humanity, such as freedom of speech, safety, and so on.
But later, you parted ways. You were an investor, but control was taken over by another founder. Even though he wanted you to stay, you chose to leave and also declined your shares.
Now, with ChatGPT valued at over $500 billion, you have nothing to show for it. But clearly, you didn’t do these things for money. If you had been thinking about money, you wouldn’t have left OpenAI at that time.
And now you’ve founded X.AI and the Grok model, setting out anew with the goal of fulfilling the mission you originally entrusted to OpenAI.
Then what makes you believe that X.AI can truly make a difference now? Is it through data superiority? Or computational resources? What exactly do you mean by 'connecting the physical world with the digital world'?
As for OpenAI, the reason I founded it was because Larry Page and I had many discussions about AI. He used to be a close friend of mine, but I realized he was completely indifferent to the potential dangers of AI.
What really alarmed me was an incident at my birthday party when he called me a 'speciesist' in front of many people, implying that I was too biased toward humans and not concerned enough about artificial intelligence.
That comment really unsettled me. I thought to myself, 'Larry, whose side are you on? You sound like you're speaking for computers, but you should be standing with humanity.'
From that moment on, I realized that someone needed to step up to counterbalance $Google-C (GOOG.US)$ . Because it was clear that Larry didn’t care whether humans would be replaced by machines.
So what should we do? I thought, let’s create the 'opposite of Google': nonprofit, open-source. That’s also the meaning behind the 'Open' in OpenAI.
I provided all the seed capital, completed the seed round and Series A financing, personally recruited the team—including Ilia Sutskever (co-founder and former chief scientist of OpenAI)—and taught them everything I knew.
I even helped them connect with $Microsoft(MSFT.US)$ Establish connections so that Microsoft Azure can provide them with free computational power for training.
As for myself, I have never sought any financial returns. Later, when they wanted to give me shares, I also refused because I believe it is illegal and unethical for a non-profit organization to issue shares. Therefore, I completely relinquished ownership.
Later, I founded X.AI. We did start later than others, and it has only been two and a half years since then.
However, I am confident in my technical abilities (though it may sound a bit boastful), and our current development speed has already surpassed any other AI team.
In the technology field, what ultimately matters is not the starting line but acceleration. The winner will be the one who runs the fastest.
Host: So what we are doing is truly different from others. Other teams are still at the data and model levels, while we have entered the physical dimension—actions, vision, real-time feedback—all of which they cannot touch. Additionally, we have real-world real-time data.
What does this mean? Can we build a truly powerful enterprise? Do we have the opportunity to lead?
Elon Musk: If you look at it from the perspective of whether it’s a good business, there’s actually no need to worry too much.
Even if we are just a small player in the AI industry, as long as the product is genuinely effective—even if it only improves societal efficiency by 10%—that would be sufficient to support a company worth hundreds of billions of dollars.
In the future, there will be many AI companies like ours. Each of them, if properly established, can support an entire economy.
The key question is, how can we outperform others? I believe it comes down to three factors:
Can we attract the top-tier talent?
Can we deploy the largest number of AI hardware systems?
Can we significantly outpace others in the speed of rolling out GPU computing resources?
And on these three fronts, we already have clear advantages. Even Jensen Huang, CEO of NVIDIA, publicly expressed his astonishment at the speed with which we built the Colossus data center.
Host: Jensen Huang also said that there is only one person in the world who can achieve what you’ve done.
Elon Musk: He really did say that (laughs). I’m truly grateful—it’s quite an honor.
Host: What you just said makes us start to wonder… Are you really human?
Elon Musk: I’ve always told people I’m an alien, but no one believes me (laughs).
However, I do have official documentation. When I obtained my U.S. green card, it stated 'Alien Registration' (which could also be humorously interpreted as 'extraterrestrial registration,' a nod to Elon Musk's quirky humor, haha). It was written by the U.S. government; they just required some formal registration.
On a serious note, I do possess certain skills that are now extremely scarce in the United States, especially when it comes to actually delivering hardware products.
If you look at the most successful manufacturing companies in the U.S. since World War II, Tesla and SpaceX undoubtedly rank among them.
05. Three Key Factors for AI Company Success
Host: Let’s talk about Grok. You’ve been emphasizing the path of 'connecting the physical world with the digital world,' which is completely different from traditional AI companies that only operate within the digital realm to calculate 'who will buy what,' correct?
Elon Musk: It’s not entirely different; there are areas of overlap, but some aspects are indeed distinct.
If you ask whether an AI company can succeed, I believe its success hinges on three key factors:
First, talent—can they attract the smartest people? Second, hardware capability—can they deploy the largest number of AI hardware systems? Third, unique data—do they possess data that others don’t?
In terms of hardware, we have proven that X.AI is one of the strongest players currently. Even Jensen Huang publicly expressed his astonishment at the speed with which we built the Colossus data center.
Regarding data, we have full access to real-time data from the X platform, which may be the highest-quality source of real-time human conversational data globally—vast in volume and updated at an incredibly fast pace.
We have mastered these three elements, and I have so many ideas in my mind that sometimes I don’t even know how to process them. So you can expect that we will make moves beyond the 'chessboard,' doing things others cannot imagine.
The Grok 4 Heavy we have launched is already incredibly powerful. It can simultaneously generate multiple agent roles, akin to a small study group where everyone thinks in parallel, ultimately synthesizing conclusions.
Most importantly, it continuously self-optimizes.
And now, we are training Grok 5. I believe it will become the smartest AI model in the world, leading in every dimension.
Host: Grok 5?
Elon Musk: Yes, Grok 5 is the first time I feel that we truly might achieve Artificial General Intelligence (AGI).
Although the probability of success may still be low—I estimate around 10%—this is already a 'non-zero chance.' Before this, I never truly believed AGI was so close.
And this model is not only intelligent but also extremely fast. We’ve given this project a new name: The Galactic Encyclopedia.
This is a tribute to Isaac Asimov and Douglas Adams, both of whom mentioned this concept in their works.
Our goal is to create an open-source repository of knowledge, refining and integrating human knowledge and making it accessible to everyone. You can access and use it freely, or even train your own models.
But it's not just a tool for Earthlings. We want it to be spread throughout the solar system, sent to the Moon, dispatched to Mars, and delivered into outer space. It will be the modern equivalent of the Library of Alexandria. The destruction of that ancient library was an enormous loss to human civilization.
We hope this time, we won’t repeat that tragedy. To 'engrave' human civilization 'on stone,' though not literally stone, but rather a super-stable storage medium.
Even if the future world encounters the most extreme risks, there will always be some civilization, some simulation, that can pick up our story again and reignite the light of civilization.
Host: So what are the technical breakthroughs in Grok 5? Why do you suddenly think there is a 10% chance of achieving AGI? Is it because the parameter count has increased? Or the data quality improved? Or the training methods advanced?
Elon Musk: Several key advancements:
First, Grok 5 is one of the models with the highest number of parameters I know of, reaching 6 trillion parameters, while Grok 3 and Grok 4 were both based on a 3-trillion-parameter architecture.
Second, its 'intelligence density' has significantly increased. In other words, the amount of intelligence carried per GB of storage and the conscious capability supported by every trillion computations are higher. This is a metric we highly value.
Our training methodology has also evolved remarkably. The quality of data used this time is mission-critical, meticulously selected to achieve our goals.
Additionally, Grok 5 is a native multi-modal model. It can process not only text but also understand images, audio, and video, and possesses powerful tool-using capabilities, even creating new tools to assist in learning and thinking.
Its visual system will be extremely advanced. Most importantly, it will have real-time video comprehension abilities. This is crucial for achieving AGI.
Humans can comprehend real-time visual scenes, but most AI systems are still unable to do so. Without this capability, achieving true general intelligence is simply impossible.
Of course, we have a few secret ingredients… but it wouldn't be appropriate to disclose them in public (laughs). You can't reveal all your secrets, after all.
What I can say is that Grok 5 will give people the impression that 'it seems to be conscious.' Of course, it isn’t actually conscious, but you’ll feel as though it might be.
Solar-powered AI satellites may become the lowest-cost energy solution for operations.
Host: But if Grok 5 surpasses Grok 4, and Grok 4 outperforms Grok 3, will this evolutionary process continue indefinitely? Will we witness the evolution of what might be called 2x perception, 5x perception, or even 10x or 1 million x perception?
Elon Musk: Yes, perception will continue to grow.
What’s truly astonishing, however, is how far we might go. It could be a path that’s almost unimaginable.
We’ve already glimpsed the road ahead, and it stretches incredibly far—so far that it nearly exceeds the boundaries of human comprehension.
We plan to launch approximately 100 gigawatts of solar-powered AI satellites into space annually. This will become the most cost-effective energy solution for operating large-scale AI systems in the future.
To provide some context: The average electricity load across the United States is about 460 gigawatts. In other words, the total power output of these satellites will be roughly equivalent to one-quarter of the total electricity consumption in the U.S.
Host: You mean, it’s equivalent to increasing the entire power supply of the United States by 25%?
Elon Musk: Yes, and we already have a complete plan to achieve this. Going forward, this project is going to get very exciting.
Host: It is said that there are about a trillion Earth-like planets in the universe. The Big Bang occurred approximately 13.8 billion years ago, while Earth was formed just over 4 billion years ago. In other words, many planets with oxygen, hydrogen, carbon, silicon, and other essential elements for life may theoretically also have given rise to life. Life on Earth has gone through four mass extinctions, so similar processes might have occurred on other planets as well...
Do you think it’s possible that there could be an extraterrestrial civilization that combines carbon-based life forms with machine-based technology—something like 'half-human, half-machine'?
Elon Musk: Of course, I want to know. My starting point for doing many things is actually driven by curiosity.
I want to figure out: what exactly is the nature of the universe? Is the standard physics explanation for the origin of the universe truly reliable? Will 'heat death' be the ultimate fate of the universe? Are there any alien civilizations? Do we have the opportunity to communicate with them? Are there fundamental questions we should be asking but haven't yet thought of?
This is the driving force behind my actions: expanding consciousness and gaining deeper insights into the essence of the universe.
07. The AI5 chip will be at the core of Tesla’s future.
Host: Let’s bring the conversation back from the cosmos to reality and talk more about Tesla. You once said that Tesla's true core competitiveness lies in 'doing things faster, better, and cheaper.' When we first invested in you, you told us that what really matters isn’t the machines that build cars, but the machines that build those machines.
In other words, you were already researching 'the machine that builds machines' 15 years ago, and today we are finally seeing the results. Now, Tesla produces a new car every 35 seconds on average. You’ve also mentioned that this speed can be further reduced to 10 seconds, or even 5 seconds. Is that true?
Elon Musk: I do see the possibility of achieving a production rate of one car every five seconds.
In fact, just imagine—it's not that fast. A car is about five meters long, and people walk at a speed of roughly one meter per second. So, producing one car every five seconds is like the car 'walking itself off the production line,' which doesn’t seem crazy or dangerous.
To give an analogy: If you operate 24 hours a day, seven days a week, that’s 10,000 minutes in a week. Even if you only produce 10 cars per minute, you could still achieve a weekly output of 100,000 vehicles.
Host: The question is, why can only you achieve this? Is it because other automakers simply don’t care? Or is it that in traditional companies, if someone proposes a new method and it fails, they get fired; and even if it succeeds, it just means more work with no gratitude? If there's no safety net for failure, then no one dares to take risks.
Why are you able to continuously improve your products while others cannot? For instance, Chinese companies—they are indeed very good at imitation, and sometimes their first attempt at copying is even better than ours.
Executives at Ford have said that Chinese companies actually learned from us. The fact that they are doing well today, to some extent, is also an affirmation of our work. But the question remains: Why haven’t they reached your level? Why don’t others do the same?
Elon Musk: Most companies prefer incremental improvements. Their executives only want to achieve a 5% improvement each year, maybe 10% at most, rather than risk failure by pushing for a disruptive change.
But personally, I am not afraid to take big risks.
I also approach problems using a physics-based mindset. I remember once when I was in the factory, I started thinking: What is the 'volumetric efficiency' of this factory? In other words, if you consider the entire factory as cubic meters of space, what percentage of that space is being used effectively?
The result was surprisingly low.
Another issue is the inefficient movement of parts and vehicles within the factory. Often, this is due to the assembly work itself being slow, such as installing brake lights or seats.
So I propose an idea: optimize the factory as if it were a chip. Just like routing on a chip, we can reduce paths, increase density, and speed up the clock frequency. The goal is to make the entire factory run more compactly and efficiently.
Host: I heard you once said something very interesting: 'I think of myself as a bit. If I were a bit, how would I want to move?' I find that perspective quite unique.
Elon Musk: Yes, I did say that. If I were a bit, or an atom, I would ask myself: 'Is my current path logical? Does it make sense?' If not, then it’s worth optimizing.
Similarly, if the movement route of a part in the factory doesn’t make sense, it needs to be changed.
Host: You’ve mentioned that you spend your weekends working on chips. What exactly are you doing, and how are you approaching it?
Elon Musk: It’s usually on Saturdays, but over the past few weeks, I’ve also been working on Sundays.
We’re developing the next-generation AI5 chip, which will be at the core of our future. It determines the fate of Tesla as a whole.
AI5 will be used in our next-generation fully autonomous vehicles and also in the Optimus humanoid robot.
To be honest, this project was initially in trouble. The design goals were extremely ambitious, progress was slow, and there was no clear path to success.
Meanwhile, we have another project called Dojo. While it’s doing okay, it is far from being able to compete with NVIDIA.
So I decided to merge the two projects into one and focus entirely on AI5. I know we will still rely on NVIDIA for training, but AI5 is a chip specifically designed for inference, targeting high performance, low power consumption, and low cost.
Our goal is to make AI5’s inference performance 2 to 3 times that of NVIDIA’s chips while costing only one-tenth as much.
Can you imagine? Double the performance at one-tenth the cost. These performance and cost metrics are absolutely critical.
To get this project on track, I’ve invested an enormous amount of personal energy. I can now literally visualize the entire physical layout of the chip in my mind—the entire wiring and structure—as if it’s ignited in my brain.
08, Prepare to Build a 'Gigafactory' for Chips
Host: But when it comes to chip manufacturing, you once mentioned a very radical idea—that we might need to build our own 'gigafactory.' Doesn’t that sound crazy? Shouldn’t we collaborate with industry giants like Taiwan Semiconductor or Samsung? After all, they’re experts, and building a chip fab typically requires investments of $20 billion, $30 billion, or even $40 billion.
Are we really going to start from scratch and do it ourselves? Where will the people come from? Where will the talent come from?
Elon Musk: First, I have great respect for Taiwan Semiconductor and Samsung. We’ve collaborated with them at Tesla and SpaceX—they’re top-tier companies.
Of course, we hope they can help us produce AI chips as quickly as possible, pushing their production capacity to the limit. Whatever they can produce, we’ll take.
But the reality is: they are not fast enough.
I asked them, 'How long will it take to build a wafer fab from scratch and get it into actual production?' They replied, 'Five years.'
Five years? For me, five years is almost an eternity.
My pace of work is: setting goals in one year, achieving them within two years, and by three years out, things are already unclear.
Waiting five years? Unrealistic.
Of course, if they could change their pace and are willing to help us produce 100 billion, or even 200 billion AI chips annually within our required timeframe, that would be ideal. But currently, they can't do it.
Host: Then why don’t they proactively speed up? We are clearly the downstream demand side with a large number of chips for internal use. Why don’t they want to become our most important partner?
Elon Musk: In fact, they are already our partners.
The wafer fabs we use include: Taiwan Semiconductor in Taiwan, Taiwan Semiconductor in Arizona, Samsung in South Korea, and Samsung in Texas.
These four fabs are all serving us.
From their perspective, they are indeed doing their utmost. But I must say, they haven’t yet encountered clients with a real sense of life-or-death urgency.
So we face two options: either build a 'mega-fab' ourselves, or accept the reality that the production volumes of Optimus robots and autonomous vehicles will be constrained by chip supply capabilities.
Host Clearly, you wouldn't accept the second option.
Elon Musk Right, I can't accept that. (Laughs)
09. The safety of FSD mode is more than four times that of human driving
Host Let’s talk about FSD (Full Self-Driving). You recently mentioned that you wouldn't ramp up vehicle production until you were absolutely confident that FSD would work. Are you confident now?
Elon Musk Yes, I am completely confident now.
Host We have indeed seen significant breakthroughs. You’ve also mentioned that many users who purchased FSD haven’t even used it, not even once.
Elon Musk Yes, this situation is even more common than you might think.
Host This is simply surreal.
Elon Musk: Therefore, we are now requiring every user to demonstrate the FSD system on-site. This is for safety reasons.
To date, Tesla’s cumulative mileage driven in FSD mode has exceeded 10 billion miles. The data clearly shows that safety in FSD mode is more than four times higher than human driving.
Therefore, we strongly recommend that customers experience it at least once to understand how the system works and how to activate it. It’s not about sales but about protecting their lives.
Host: In other words, this system isn’t programmed with rigid “if-else” logic—it doesn’t freeze when encountering a fire truck towing a scooter. Instead, the AI truly understands the surrounding environment, correct?
Elon Musk: Absolutely correct. The key to achieving true autonomous driving lies in the evolution of AI software, not hardware.
Our self-developed AI chip (manufactured by Samsung) can already achieve 2-3 times the safety of human driving in unsupervised conditions. When the AI5 chip is deployed in the future, we are confident it will deliver a tenfold improvement in safety.
This represents a profound technological advancement. I strongly recommend everyone try the FSD system for themselves.
Visit any Tesla store, and they will provide a demonstration. It’s not a secret; it’s something everyone should experience.
Host: Just a heads-up—if you try the FSD, you might end up wanting to buy a car (laughs).
Elon Musk: That’s right. You should give it a try.
10. Vision: Expanding awareness and gaining a deeper understanding of the essence of the universe.
Host: You do these things not to buy yachts or seaside villas. It’s because you believe the future of humanity is worth protecting. So when Larry Page called you a 'speciesist,' I think he was right.
Elon Musk: Yes, I am indeed a staunch supporter of the human camp, and I’ve never shied away from that. (Laughs)
Host: Ultimately, whether you have $1 trillion or $400 billion in net assets in the future doesn’t matter to you at all. So what do you truly care about? What do you want people to remember you for?
Elon Musk: What I care about is whether I can continue to guide these companies in the right direction. That’s why I must retain a certain level of equity control.
From a personal consumption perspective, I don’t actually own any luxury assets. No vacation homes, no private yachts. I currently live in a moderately sized house in Austin.
Oh, and I should mention, I also have a very small house at Starbase (the Starship launch site).
Host: I’ve seen that house—it’s really tiny.
Elon Musk: Yeah, some of my friends came to visit me and thought I was joking when they saw it.
I said, 'No joke, it’s real.' It’s a house I bought for $8,000. I personally remodeled it, laid artificial turf in front, and put up a small white fence.
But back to the topic. As I mentioned earlier, even if AI and robotics can fulfill all of humanity's material needs, humans will still hold intrinsic value. In an ideal future scenario, AI might even encounter a peculiar problem: 'I’ve completed every task for humanity—what should I do next?'
It may truly find itself in a state of idleness.
However, overall, what I wish to achieve is rather simple: to expand the boundaries of consciousness, continually broadening its scope and density into the future.
I hope we can emulate the vision of 'Star Trek,' venturing into galaxies never before reached by humans, exploring whether extraterrestrial civilizations exist, and perhaps studying the remnants of extinct civilizations to understand their history and worldviews.
In summary, I want to understand the universe.
Host: Thank you for being here today. Thank you for everything you have done for me, for our shareholders, and for humanity.
Elon Musk: Thank you; it is my honor.
Futubull AI: Your Portfolio Manager! Your Stock Selection Advisor!Come and experience it now >>
Editor /rice
