news

text: web3 sky city · city lord

at the all-in summit held recently, elon musk, as always, became the focus of everyone's attention. the roundtable discussion covered everything from his challenges at x (twitter), to the latest progress in mars exploration, to breakthroughs in the field of ai. since the acquisition of twitter for $44 billion, musk's business strategy and personal style continue to spark heated discussions. whether it is freedom of speech, government regulation, or the shaping of the future by technology, musk has been outspoken in sharing his vision and thoughts.

this interview shows musk's many sides - from the technical leader of tesla and spacex, to the controller of social platforms, to the promoter of global artificial intelligence progress. his personal charm, belief in the future, and strategy for solving global problems have made him one of the most controversial and influential entrepreneurs of our time.

liberty and x company:musk wittily discussed his struggles in the area of ​​free speech, especially the challenges after acquiring twitter. he firmly believes that platforms should follow the laws of each country, but also emphasizes the importance of free speech. the conflict with brazil in particular was one of his focuses, as company x felt it was being asked to violate brazilian law.

regulatory challenges:musk has harshly criticized the red tape of the u.s. regulatory system. he believes that government inefficiency hinders innovation, especially in major infrastructure projects. he also mentioned the huge difference between california and texas in the speed of approval and project advancement, saying that texas' policies are more conducive to promoting rapid development of enterprises.

ai and the future:regarding artificial intelligence, musk said that the development speed of ai is far beyond expectations, and predicted that ai will bring unprecedented prosperity to the world in the future. he also talked about the progress of autonomous driving technology and tesla's dojo supercomputer project.

optimus robot project: in the interview, musk talked in detail about tesla's humanoid robot optimus project. he expects that in the next decade, the cost of robot production will drop significantly, and everyone will be able to afford robots, and the number of robots may even exceed that of humans. he believes that optimus will be able to do human work and become the core driver of future economic growth.

mars and the dream of multi-planet civilization:musk reiterated his ultimate vision of turning humans into a multi-planet species. he revealed that spacex's starship project is about to reach a key milestone, and the next flight will be carried out after regulatory approval. he firmly believes that spacex will achieve its goal of sending humans to mars in the next few years.

optimistic outlook for the future:musk has mentioned his vision for the future many times, saying that the goal of humanity in the future is not just to solve existing problems, but to create an exciting future, such as multi-planet colonization and the advancement of ai technology. he believes that a future full of technological progress will bring hope to mankind and encourage society to think about how to maximize the benefits of technology from a long-term perspective.

voiceover:

musk's career as a ceo has been a whirlwind after buying twitter for $44 billion. since he took over twitter, the stock price of his other big company, tesla, has plummeted by more than 30%. often, his next move is unclear. it can even be said that he has exhibited some uncharacteristic behavior.

elon musk's cooperation or relationship with other countries deserves scrutiny. the biden administration has just announced a second investigation into elon musk, and it's less than a week. whether it's tesla or spacex, they have their own product roadmaps, and whether elon is in the company does not affect their plans.

people said he would never get a rocket to space, he did. people said the roads here would never complete deliveries, he did. people said he would never complete a hundred, he completed two hundred.

“as an entrepreneur, you can’t listen to noise, especially from people who have accomplished nothing but are fanatically following you.”

"we are among the stars, we are a multi-planet species across multiple planets and multiple galaxies. that's a great future, and that's what we should be striving for."

almost every vc i've talked to, every ceo, has looked at what elon has done and said, this is an example of how you can challenge your team to achieve the impossible in a very difficult environment.

"you can see on the left screen those grid fins spinning and turning, guiding the booster. and that landing gun. the landing gun just went live. and you can see down there in the water, we have the blast zone. what an incredible sight."

he's truly a visionary like nothing i've ever seen. how could you bet against him? it seems like elon is going to be not only the richest man in the world, but the first trillionaire in the world. elon has overcome amazing challenges over the last 10 to 15 years. he's probably had to deal with things that most of us would have crumbled over, and he's come through it. he's basically blown all the haters to pieces. i think it's just incredible.

host:

elon musk!

i'm going to start. okay. thanks for your time. how are you doing, bro? are you busy?

Elon：

yeah. i mean... it's rare that a week is slow. i mean, in this world. any given week it seems like things are getting a little crazier.

host:

this is definitely a simulation. we have now agreed to that.

Elon：

if we were in some kind of alien netflix series, i imagine the ratings would be high.

host:

how is the battle for free speech going? you have been fighting for two years. free speech is not cheap.

Elon：

i think it's about $44 billion. that's pretty crazy. there's a weird movement around the world to suppress free speech, and it's something we should be very concerned about.you have to ask, why is the first amendment a priority? it's like number one.

host:

are you worried? i think the message resonates with this audience. i think we've always thought that the west was the exception. we know there are authoritarian places in the world, but we thought that in the west we would have free speech. and we've seen, as you said, that this seems to be a global movement. in the uk, teenagers have been sentenced to prison for internet memes.

Elon：

like if you like a facebook post, you can go to jail. people are literally being sent to jail for making obscure comments on social media. not even troll posts. it's crazy.

pavel was recently put in jail. i was really shocked. i was thinking, what kind of major crime is this?

host:

pavel is in france, and of course, we have judge voldemort in brazil. that seems to be the one that's affecting you the most. what's the latest development?

Elon：

i think we're trying to find a reasonable solution in brazil. the concern is, i think just to make sure that this is understood correctly. funny pictures aside, the essence of the concern is that, at least at xcorp, we have a sense that we are being asked to do things that violate brazilian law.

obviously, as an american company, we can't impose american laws and values ​​on other countries. we wouldn't get very far if we did that. but we do think that if a country's laws are a certain way and we're asked to break those laws and keep quiet about it, then obviously that's not ok. i just want to be clear on that.

sometimes it looks like there's some crazy stuff going on, whatever it is, with billionaires making outrageous demands on other countries. and while that's true, there are other things that i think make sense.

i think anything we do at company x has to be explainable in broad daylight without feeling like it's dishonorable or like we did something wrong. that's the nature of the problem.

so we are actually in discussions with the brazilian judicial authorities to hopefully get to the bottom of what happened. what exactly happened? if we were asked to break brazilian law, then obviously that shouldn't be accepted by the brazilian judicial authorities. if we weren't asked to break the law and we got it wrong, we want to understand where we went wrong. i think that's a pretty reasonable position.

host:

as your friend, i'm a little concerned that you're going to go to one of these countries and i'm going to wake up one day and find out that you've been arrested and i need to bail you out or something. it feels very urgent. this is not a joke. they're really saying, it's not just biden saying we have to investigate that person. now this has become very literal.

this…who is the person who just wrote that…is that the guardian article, this is a trend.

Elon：

the guardian? what are they protecting? ? right. censorship? censorship.

host:

but the premise is that you bought this thing, this online forum, this communication platform, and you allowed people to express themselves on it. therefore, you have to be locked up. i don't understand the logic here. what do you think they are afraid of right now? what is their motivation?

Elon：

if someone is trying to sell the world a false premise, that premise can be disproven through public dialogue. and then they're going to oppose public dialogue based on that premise because they want that false premise to come out. so i think that's the problem. if they don't like the truth, then we're going to suppress it. what we try to do at xcorp is ... i'm going to differentiate it from my son, who's also named x.

host:

yes.

you have your parents' goals, you have your company's goals, and basically everything is called x, and it's very hard to tell the difference.

Elon：

car, son... yes, everything is x.

so what we try to do is comply with the laws of each country. so if something is illegal in the united states, or illegal in europe, or brazil, or somewhere else, then we remove it or suspend the account because we are not here to make laws. but if the speech is not illegal, then what are we doing? okay, now we are acting as a censor, so when does it stop, who decides? so where does this road lead to? i think it leads to a bad place.

so if the people of a country want the law to be different, they should change the law. but other than that, we follow the law in every jurisdiction. right, that's it. it's nothing more complicated than that. we're not trying to flout the law. let me be clear about that. we're trying to follow the law. if the law changes, we'll change. if the law doesn't change, we'll stay the same. we really just want to follow the law. it's that simple.

if someone believes that we are not following the law, then they can file a lawsuit.

host:

i mean, there are some european countries that don't want people promoting nazi propaganda. yes, they're sensitive about it. it's illegal. in those countries, it's illegal. yes, it's illegal. in those countries, if someone posts that kind of content, you need to take it down. yes. but they usually file a complaint and ask for it to be taken down.

Elon：

no, in some cases, it's clearly illegal. that is, if something is clearly illegal, you don't need to file a lawsuit. we can totally read the law. this is a violation of the law. anybody can see that. if someone is stealing something, you don't need to... let me look up the law. no, they are stealing something.

host:

let's talk about this. we have jd vance here this morning. he's doing a great job. and, there's a picture on the x, which is basically you, bobby, trump, and jd like the avengers, i guess. and then there's another meme of you sitting at a desk that says doge (department of government efficiency). yeah, yeah. i posted that one.

Elon：

i made it with the grok image generator, and then i posted it. that's my profile.

i think it's very difficult. there hasn't been a serious attempt to reduce the size of government and get rid of ridiculous regulations in a long time. the last time there was a real concentrated effort on this was under reagan in the early '80s. it's been 40 years since the last serious attempt to get rid of regulations that don't serve the greater good and reduce the size of government.

if we don't do that, regulations and laws will accumulate every year, and eventually everything will become illegal. that's why we can't get major infrastructure projects done in the united states. for example, if you look at the ridiculous situation with the california high speed ​​rail, i think they spent $7 billion on a 1,600-foot section, and there's no rail on that section. that's the cost of 1,600 feet of concrete.

sometimes i can be overly optimistic about schedules, but if i weren't an optimist i wouldn't be doing what i'm doing. at current trends the california high speed ​​rail might not be finished until the next century. maybe not, by then we'll have teleportation. by then ai will do everything.

you can think of the us and many countries (arguably worse than the eu) as being like gulliver tied up by a million little strings. any one specific regulation isn't that bad, but you have a million of them, literally millions of them. you just can't get anything done at the end of the day. it's a huge tax on consumers, on people. it's just that people don't realize this huge tax in the form of irrational regulations.

i'm going to give you a recent example, and this is just ridiculous. spacex was fined $140,000 by the epa for dumping potable water on the ground. we're at starship base, in a tropical thunderstorm area, and this stuff is falling from the sky all the time. and it's not causing any actual harm. it's just for cooling the water on the launch pad during launch. it's not causing any harm at all, and they agreed that it wasn't causing any harm. we said, well, there wasn't any harm, and you want us to pay a $140,000 fine? and they said, yes, because you didn't have a permit. we didn't know we needed a permit, in an area where there's fresh water falling from the sky all the time, and there's fresh water on the ground, and there's zero harm.

close to the ocean, because there's a little water there too. sometimes it rains so hard that the roads are flooded. so is there any reason for us?

and then they said, if you don't pay this $140,000, we're not going to process your launch applications, the starship launch applications. so they blackmailed us. we paid the $140,000, but we felt that there was no point in continuing like this. at this rate, we would never get to mars.

host:

what's confusing about this is that we're acting against our own interests. when you look at it, we do have to make some sacrifices, let's not talk about fresh water, but rocket launches make a lot of noise. i'm sure there are some complaints about noise from time to time. but sometimes you want to have a party, or you want to make some progress, and there's going to be a little noise. so we're trading a little noise for a lot of progress or even fun. so when did we become unable to make these trade-offs?

talk about the difference between california and texas, where you and i live now. in texas, you were able to build the gigafactory. i remember when you got the land. it seemed like less than two years from the time you got the land to the time you had a party to celebrate.

Elon：it took 14 months from start to completion.

host:is there any place in the world that is faster than this? for example, is china faster than this?

Elon：it took china 11 months.

host:so texas, china, 11 months and 14 months respectively. how many months does california take?

Elon：

to give you an idea of ​​the scale, the tesla gigafactory in china is three times the size of the pentagon. there are bigger buildings in the us, but the pentagon is the biggest one. yeah, or it used to be. in terms of pentagons, that's about three. three pentagons and counting.

it takes two years to get regulatory approval in california alone. so that's the problem.

host:

in what ways do you think regulation would be helpful? for example, some would say we need some checks and balances. because for every good actor like you, there is also a bad actor. so where is that line?

Elon：

instead of having a reasonable deregulation and reduction in the size of government, we make it public. if the public is very interested in a rule and wants it to stay, we keep it. if it turns out that the rule is bad, we restore it immediately. problem solved. it's easy to add rules, but we don't actually have a process for removing rules. that's the problem. there is no garbage collection of rules.

host:

when we were watching you work, the first month at twitter, everybody was pulling together, and you were doing zero-based budgeting, and you got costs under control very quickly. and then, miraculously, everybody was saying the site was going to crash, and you added 50-plus features. so maybe explain that, because it was the first time. there were a lot of articles like, "twitter is dead, there's no way it could survive." it was almost like the media was expecting you to fail. rest in peace, let's write this obituary. they were all saying goodbye to each other on twitter, remember? they were all leaving, saying goodbye, because the site was going to crash. all the journalists were gone.

if you want to hang out with a bunch of hall monitors, oh my god, threads is awesome. every time i go there and post something, they get really excited.

Elon：

if you like being berated over and over for no apparent reason, threads is the place to be. it really is the most miserable place on earth.

host:

disney is the happiest place and this is the anti-disney.

but if we were to go into government, let's say you go into the department of education, or pick any department, you work with a lot of departments. you can't go in and use zero-based budgeting, okay, we get that. but if you could cut these organizations by two, three, four, five percent, what would the impact be?

Elon：

i think we need to do more than that. ideally, we'd be growing at 2 or 3 percent compounded annually, which would be better than what we have now.

if trump wins, i guess there are different opinions on whether this should happen, but we do have a once-in-a-lifetime opportunity to deregulate and reduce the size of government. except for the regulation thing, on top of all this, the united states is going bankrupt very fast. everyone seems to be turning a blind eye to this problem.

host:

they were all grabbing for silver. before the titanic went down, everybody was stuffing their pockets with silver.

Elon：

the defense department budget is a very large budget, a trillion dollars a year, the defense department, the intelligence community, a trillion dollars a year. the interest payments on the national debt just exceed the defense department budget. they are over a trillion dollars a year in interest alone, and it's rising. we are adding a trillion dollars to the debt every three months, and that debt will be paid back in some way by our children and grandchildren. soon, that cycle will be shortened to once every two months, and then once a month. eventually, we will only be able to pay the interest. it's like a person who accumulates too much credit card debt, it's not going to end well. so we have to reduce spending.

host:

let me ask a question, because i bring this up a lot. the counterargument that i hear, and i don't agree with it, but it's the counterargument that a lot of politicians have, is that if we cut spending, right now federal, state, and local government spending combined is 40 to 50 percent of gdp. so nearly half of our economy is supported by government spending, and nearly half of americans are dependent on government checks, either directly or indirectly, either through government-paid contractors or employed by government entities. if we cut spending quickly and aggressively, we're going to have significant contraction, job losses, and a recession.

so what is the balanced strategy, elon? think about it realistically, because i agree with you 100 percent. however, let's say trump wins and you become chief doge...doge...do, like double g. i think the challenge is how quickly can we move? how quickly can things change? without all the contraction and unemployment.

so my question is, how do you actually solve this problem when so much of the economy and so many people’s jobs and livelihoods depend on government spending?

Elon：

i do think it's a false dichotomy. it's not that there should be no more government spending. what you really need to ask is, is it at the right level? and remember, in any organization, if a person is in a less efficient organization than they are in a more efficient organization, their contribution to the economy, their net output of products and services will be reduced.

you have a few obvious examples, like east germany and west germany, north korea and south korea. south korea's economy is really amazing. this is the compounding effect of productivity growth, it's a world of difference.

then there's the comparison between east and west germany. in west germany, just in terms of cars, you had bmws, porsches, audis, and mercedes. in east germany, it was just a line on a map. the only car you could buy was the trabant, which was basically a lawn mower with a shell on it, and it was extremely unsafe, and there was a 20-year waiting list. you had to put your children on a list when they were conceived. even then, i think only one in four people could get one of these inferior cars. it's just an interesting example of how basically the same people behave under different operating systems.

west germany was not some capitalist paradise, in fact, it was quite socialist. so when you look at west germany, it was maybe a semi-government role, and in east germany it was 100% government role. again, the difference in living standards was 5 to 10 times higher, and even qualitatively much better.

obviously, a lot of people have an amazing debate in modern society about which system is better. i'll tell you which system is better: the one that doesn't have to build a wall to keep people in. that's a clear sign. if you have to build a barrier to keep people in, it's a bad system. it wasn't the wall that west berlin built, and anyone who wanted to escape west berlin could leave at will.

host:

you've thought about this carefully. wait, so your point is that if we cut government spending in half, jobs will be created. jobs will be created fast enough to make up for the...

Elon：

i'm not suggesting that people be dumped right away, with no severance pay, and no way to pay their mortgages. then you see some reasonable exit options...

a reasonable exit scenario is that they still receive a salary but have a year or two to look for a job in the private sector, which they will find. then they will be in a different operating system. again, you can see the difference. east germany was incorporated into west germany, and the standard of living in east germany rose dramatically.

host:

so, in four years, if you could reduce the size of government under trump, what would be a good goal?

Elon：

just in general terms. do you want me to be assassinated before this happens?

host:

no, it isn't. pick a lower number.

Elon：

there's an old saying, "mad mail." they probably will. so we'll keep the post office. i need a lot of security, guys. the number of disgruntled workers, former government employees, is pretty scary. i might not make it.

host:

i'm saying low single digits per year would be acceptable over four years.

Elon：

but the thing is, if it's not done, like you have a once in a lifetime, once in a generation opportunity, but you don't take serious action, and then you have four years to do it. if it's not done, then...

host:

how serious is trump about this? you talked to him about this.

Elon：

yes, i think he's very serious about this. in fact, if we eliminate pointless regulations and move people out of government and into the private sector, we're going to have tremendous prosperity. i think there's going to be a golden age in this country, and it's going to be fantastic.

host:

can we talk about spacex? you have a bunch of big milestones coming up.

Elon：

yes, actually, there's going to be a very exciting launch tonight. weather permitting, i'm going to leave here and head to cape canaveral for the polaris dawn mission, which is a private mission funded by jared eisenman. he's an amazing guy. it's going to be the first commercial spacewalk, and it's going to reach the highest altitude since apollo, so it's the farthest from earth.

host:

what happens next? let's assume the mission is successful.

Elon：

i really hope so. the safety of the astronauts is the most important. if i had all the wishes, i would say that's the one i want to achieve. space is dangerous. the next milestone will be the next flight of starship. the next flight of starship is ready to fly, and we are waiting for regulatory approval. it really shouldn't be the case that a giant rocket can be built faster than a file can be moved from one desk to another. the staff is really working hard.

host:

have you seen that movie zootopia? there's a sloth in it. they accidentally told a joke and i was like, oh no, this is going to take a long time.

Elon：

funny thing is, i went to the dmv, about a year after zootopia, to renew my driver's license. and the guy, very self-aware, had a picture of the sloth from zootopia taped to his cubicle. and he was actually pretty prompt.

host:

there is one imperative: defeat the sloths, defeat the personal agents.

Elon：

no, sometimes people think the government is more competent than it actually is. i'm not saying there aren't competent people in the government. they're just operating in an inefficient operating system. once you move them into a more efficient operating system, their output increases dramatically. as we saw when east germany reunited with west germany, the same people became much more prosperous. basically a semi-capitalist operating system.

for many people, their most direct experience with government is the department of motor vehicles (dmv). the important point is that government is the dmv on a large scale. with that image in mind, how big do you want to make it?

host:

sorry, can you go back to chamath's question about starship? you just announced the other day that starship will go to mars in two years, and you plan to launch humans in the next window, which is about four years. how much is the government involved?

Elon：

based on our progress so far on starship, we have successfully reached orbital velocity twice and were able to achieve a soft landing of the booster and spacecraft in the water, even though half of the spacecraft's flaps were blocked. you can see the video on the x platform, which is very exciting. so, we think we will be able to launch reliably and frequently, and at a very fast rate.

the fundamental holy grail breakthrough in rocketry, the fundamental breakthrough that is critical to making life multi-planetary, is a fast reusable, reliable rocket. starship is the first rocket designed to be fully reusable where success is one of the possible outcomes. for any project, you have to say, this is a circle. we have a graph, this is a circle, and it represents success. the successful points within the circle represent success in the set of possible outcomes. it sounds obvious, but often there are projects where success is not in the set of possible outcomes. so not only is starship fully reusable across all possible outcomes, but it's proving that with every launch. i'm sure we'll succeed, it's just a matter of when.

if we can improve on the speed of regulation, we can actually move faster, that would be very helpful. to be clear, i'm not talking about anything that's unsafe, just safe things that can keep up with the speed of rocket construction, not slower than it. then we can become a spacefaring civilization, a multi-planet species, and eventually reach out to the stars in the future. it's extremely important to have things that inspire us, things that you look into the future and say the future is going to be better than the past, things to look forward to. kids are a great way to assess that. what are kids passionate about? if you can tell them that they can be astronauts on mars, or even one day go beyond the solar system, that we can make star trek and starfleet academy a reality. that's an exciting and inspiring future. you need things that touch your heart.

host:

yes, absolutely. let's do it.

Elon：

life can’t just be about solving frustrating problems one after another, you need something to look forward to too.

host:

do you feel like you might have to move it to another jurisdiction to move it forward more quickly?

Elon Musk：

i've always thought that rocket technology is considered advanced weapons technology, so we can't just do it in another country.

host:

it's interesting to think that if we don't do it, other countries might. although they're far behind us, in theory, there's a national security reason for it. if someone could use their brains to think, like, do we want the technology that you're building, that your team is working hard on, to be stolen by other countries? then, maybe they don't have as much red tape.

Elon：

i hope that nobody tries to steal it. so, nobody tries to steal it. basically, it's crazy.

host:

yeah, it's crazy.

elon, what do you think happened that caused boeing to build the starliner this way? they were able to launch it, but they didn't finish it. they couldn't finish it. i don't understand it. now, they have to go up and finish it.

Elon：

i think boeing is a company that does a lot of business with the government, and they have a little bit of an impedance match with the government. so, they're basically one level away from the government. they're not far away from the government from an efficiency perspective because most of their revenue comes from the government.

a lot of people think that spacex is very government dependent, but it's not, the majority of our revenue is commercial. at least until recently, maybe they had a new ceo who actually showed up at the factory. the ceo before that, i think, had an accounting degree and had never been to a factory and didn't know how a plane flew. so i think if you're in charge of a company that makes airplanes fly and spacecraft go into orbit, you can't be completely unaware of how they work.

so i think if someone is running coke or pepsi and they're good at marketing or whatever, that's fine because it's not a technology-dependent business. or if they're running a financial advisory business and their degree is in accounting, that makes sense. but i think if you're a cavalry captain, you should know how to ride a horse. and it's really disturbing if the cavalry captain pulls the reins and is afraid of horses. he's afraid of horses and can't motivate the team. if he says, "sorry, i'm afraid of horses," and he gets on the horse backwards, i'm like, oops.

host:

turning to ai, peter mentioned here before that the only company that's really making money from ai so far is nvidia with its chips. do you have a sense of where the main applications of ai will be? will it be driving autonomous driving? will it be driving robotics? will it be transforming industries?

i think we are still in the early stages of ai's big business impact. do you feel that's the case right now?

Elon：

i think ai spending has probably already exceeded revenue, and it does now. there's no doubt about that. but the rate of improvement in ai is the fastest i've ever seen in any technology, far faster than before.

for example, the turing test used to be a standard. now, with a basic open source random llm, you can probably pass the turing test running on a simple raspberry pi. so, i think actually, the bright future of ai is an era of prosperity, an era of abundance of goods and services. everyone can have anything they want, unless it is something that we artificially define as scarce, like some special works of art.

but any manufactured goods or services provided, i think as ai and robotics advance, the cost of goods and services will approach zero. i'm not saying they'll literally cost zero, but everybody will be able to have anything they want. that's the future. of course, in my opinion, it's about an 80% chance. look on the bright side, it's only a 20% chance of destruction.

host:

what does that 20% look like?

Elon：

frankly, i do have to exercise some level of deliberate suspension of disbelief about ai in order to sleep well. even so, because i think the actual question, most likely the question, is how do we find meaning in a world where ai can do everything we can do, but better? that's probably the bigger challenge.

although, now i know more and more people who are retired, and they seem to enjoy that life. but i think there may be a crisis in some sense. because computers can do everything you can do, but better. so, maybe that will be a challenge.

but really, you need that end effector. you need self-driving cars, and you also need that humanoid or general purpose robot. but once you have general purpose humanoids and self-driving vehicles, really, you can build anything. and i think there's actually no upper limit to how big the economy can be. obviously, the mass of the planet is going to be a limiting factor. but the economy is really just productivity per capita times the number of people, that's the nature of the economy. if you have humanoids that can perform tasks, and there's no real limit to the number of them, and they can operate very intelligently, then there's really no real limit to the economy.

host:

you just launched colossus, which is currently the largest private computing cluster in the world, i guess a gpu cluster.

Elon：

this is the most powerful supercomputer of any kind.

host:

this also reinforces what david and peter said, which is that a lot of the economic value of ai so far has been captured by nvidia. but there are alternatives, and you actually provide an alternative. now, your situation is very special.

Elon：

because dojo deals primarily with images and large images, huge videos. so the tesla problem is different from the problem of large language models. the nature of intelligence and the important part of ai has to do with the long-term nature of the context in the tesla case. you have gigabytes of context.

why are we just mentioning this now? this is not some tens of billions of tokens of context. this is a very large amount of context because you have seven cameras, if you have several high-resolution cameras shooting for a minute, that's several gigabytes. so you need to compress it. the problem with tesla is that you have to compress that huge amount of context into pixels that actually matter, and condense it over a period of time. you have to compress the pixels in both the temporal dimension and the spatial dimension, compress the pixels in space, compress the pixels in time, and then do the inference on a relatively small computer. a small one that uses a few hundred watts of power. this is a tesla-designed ai inference computer that is probably still the best. we can't buy anything better from the vendors.

host:

the ai ​​inference computer that tesla designed is in the car is much better than what we can buy from any vendor. by the way, the ai ​​chip team at tesla is amazing. there is a technical paper and a powerpoint presentation by someone on your tesla team that was amazing to me about your design process. you designed your own transport control layer, like on ethernet. you said, ethernet is not good enough for us. you have this thing called ttcue and said, we are going to reinvent ethernet and string these chips together. it's really astounding.

Elon：

tesla's chip design team is excellent.

host:

so, over time, other companies that need video or graphics applications, in theory, you can say, why not? i have some idle compute cycles here, and that makes you a competitor to nvidia.

Elon：

this is not intentional, but we do have two projects, training and inference. at tesla, we have dojo, which is the training computer, and our inference chip, which is the inference computer in each car.

right now we only have dojo 1. dojo 2 is expected to be in volume production by the end of next year. we think it will be similar to the training system of the b200 type system. this has the potential to be used as a service. i have some improved confidence in dojo, but i think we may not really know how good dojo is until the third version. it usually takes three major iterations of a technology to get to a level of excellence. we will not have the second major iteration until next year, and the third iteration will probably be around the end of 2026.

host:

how is the optimus project going? i remember the last time we talked, you also publicly said that it was doing some light testing inside the factory. yes, it is actually useful.

what is the bill of materials, when you're mass producing it? when you start producing it like we're producing the model 3 now, millions of them coming off the line every year, how much are they going to cost? $20,000, $30,000, $40,000, what do you think?

Elon：

i've found that anything that's produced in high volumes asymptotically approaches its bill of materials cost. some things are limited by intellectual property costs, like paying royalties and things like that. a lot of things in chips are like paying royalties and depreciation of chip factories, but the actual marginal cost of chips is very low.

optimus is clearly a humanoid robot. it's much lighter and smaller than a car. so you can expect that, in mass production, you'll need maybe three more production versions of optimus. you'll need to make at least three major improvements to the design and then scale up production to a million-plus units per year. i think by then, the labor and materials cost of optimus will probably be no more than $10,000.

it's a 10-year journey, and basically, optimus will cost less than a car. at scale, there are three major technology iterations. if a car costs $25,000, then for about $20,000 you can have an optimus, a humanoid robot that's like a combination of r2-d2 and c3po but better, and can be your companion.

honestly, i think people get really attached to their humanoid robots. you look at r2-d2 and c3po in star wars, i love those characters, they're great. even though r2-d2 can only make beeps and can't speak english, and c3po translates those beeps.

host:

if each iteration took two to three years or more, is this a journey that will take ten years to reach some scale?

Elon：

i think the big iteration cycle is less than two years, about five years, maybe six years to reach a million units per year. at that price point, everyone can afford it. it will be a one-to-one, two-to-one situation.

host:

what do you think, ultimately, if we sit here 30 years from now, what will the number of robots on the planet be like compared to humans?

Elon：

yes, i think the number of robots will greatly outnumber the number of humans. a lot, yes. who wouldn't want a robot companion? everyone wants a robot companion, especially if it can take care of your life, like take your dog for a walk, mow the lawn, and watch and teach your children.

we could also send it to mars and do the work that would be needed to transform it into a habitable planet. mars is already a robotic planet, there are robots there, like rovers and helicopters, just robots. so, i think the opportunity for useful humanoid robots is the biggest opportunity ever.

i think the ratio of humanoid robots to humans will be at least 2 to 1, maybe 3 to 1. everyone will want one, and then there will be a lot of robots that you can't see producing goods and providing services.

host:

do you think this is a general purpose robot that learns how to do different tasks?

Elon：

yeah. we're a generic non-robot, we're just made of meat. we're a meat puppy, a generic meat puppy. yeah, i'm operating my meat puppy. so yes, we actually gradually learned more and more about why humans are shaped the way they are as we were designing optimus. why we have five fingers, why the pinky finger is smaller than the index finger. obviously, why we have opposable thumbs, but also, for example, the main muscles that operate the hand are actually in the forearm. your fingers are primarily controlled by tendons, which are like little strings. the vast majority of the strength in the fingers actually comes from the forearm.

the current version of the optimus hand has the actuators in the hand, and it only has 11 degrees of freedom. so it doesn't have all the degrees of freedom that a human hand has, which is about 25 degrees of freedom, depending on how you count it. in some ways it's not powerful enough because the actuators have to be in the hand. so the next generation optimus hand, which we already have prototypes of, the actuators are moved to the forearm, and like a human, the fingers are operated by cables. the next generation hand has 22 degrees of freedom, which we think is enough to do almost everything a human can do.

host:

speculatively, there are reports that x and tesla may be working together to provide services. if you just feed the bot a grok, the bot has a personality and can handle voice, video, images, all these things.

---

as we wrap up this interview, i think everybody talks about all the projects that you're working on, but people don't know that you have a great sense of humor. that's not true. you have a sense of humor. people don't see that, but i will say that for me, the funniest week of my life, or one of the funniest, was when you hosted saturday night live and i had the opportunity to tag along. maybe you saw that. maybe behind the scenes, you have some of the funniest memories of that crazy chaotic week where we were laughing for 12 hours a day. the first few days were kind of terrifying.

Elon：

yes, i was a little apprehensive at first because frankly, there was nothing funny about it. the first day was rough.

host:

just talking about the stuff that got cut... the funniest skits are the ones they don't let you do. the ones they don't let you do. you can name them so he doesn't... how much time do we have?

Elon；

are you all ready?

(some small talk about saturday night live, omitted here)