Elon Musk — "In 36 months, the cheapest place to put AI will be space”

In this episode, John and I got to do a real deep-dive with Elon. We discuss the economics of orbital data centers, the difficulties of scaling power on Earth, what it would take to manufacture humanoids at high-volume in America, xAI’s business and alignment plans, DOGE, and much more. Watch on YouTube https://youtu.be/BYXbuik3dgA; read the transcript https://www.dwarkesh.com/p/elon-musk. Sponsors * Mercury https://mercury.com/personal-banking just started offering personal banking! I’m already banking with Mercury for business purposes, so getting to bank with them for my personal life makes everything so much simpler. Apply now at mercury.com/personal-banking https://mercury.com/personal-banking * Jane Street https://janestreet.com/dwarkesh sent me a new puzzle last week: they trained a neural net, shuffled all 96 layers, and asked me to put them back in order. I tried but… I didn’t quite nail it. If you’re curious, or if you think you can do better, you should take a stab at janestreet.com/dwarkesh https://janestreet.com/dwarkesh * Labelbox https://labelbox.com/dwarkesh can get you robotics and RL data at scale. Labelbox starts by helping you define your ideal data distribution, and then their massive Alignerr network collects frontier-grade data that you can use to train your models. Learn more at labelbox.com/dwarkesh https://labelbox.com/dwarkesh Timestamps (00:00:00) - Orbital data centers (00:36:46) - Grok and alignment (00:59:56) - xAI’s business plan (01:17:21) - Optimus and humanoid manufacturing (01:30:22) - Does China win by default? (01:44:16) - Lessons from running SpaceX (02:20:08) - DOGE (02:38:28) - TeraFab This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.dwarkesh.com https://www.dwarkesh.com?utm_medium=podcast&utm_campaign=CTA_1

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  So are there really 3 hours of questions? Or are you fucking serious?

• 0:03Speaker 01

  Yeah.

• 0:07Speaker 01

  You don't think there's a lot to talk about, Elon?

• 0:08Speaker 00

  Holy pork, man.

• 0:11Speaker 01

  It's the most interesting point. All the storylines are kind of converging right now. So we'll see how it works.

• 0:17Speaker 00

  It's almost like I planned it.

• 0:19Speaker 01

  Exactly. Well, we're getting there.

• 0:20Speaker 00

  I would never do such a thing.

• 0:23Dwarkesh Patel?

  So, as you know better than anybody else, the total cost of ownership of a data center, only 10% to 15% is energy. And that's a part you're presumably saving by moving this into space. Most of it's the GPUs. If they're in space, it's harder to service them, or you can't service them. And so the depreciation cycle goes down on them. So it's just way more expensive to have the GPUs in space, presumably. What's the reason to put them in space?

• 0:45Speaker 00

  Well, the availability of energy is the issue.

• 0:51Speaker 00

  So, I mean, if you look at electrical output outside of China, everywhere outside of China is more or less flat. It's very, you know, very slight increase, but pretty much flat. China has a rapid increase in electrical output. But if you're putting data centers anywhere except China, where are you going to get your electricity? Especially as you scale. The output of chips is growing pretty much exponentially, but the output of electricity is flat. So how are you going to turn them chips on?

• 1:20Speaker 00

  Magical power sources? Magical electricity fairies?

• 1:24Dwarkesh Patel?

  You're famously a big fan of solar. 1 terawatt of solar power. So with a 25% compatibility factor, like 4 terawatts of solar panels. It's like 1% of the land area of the United States. And that's like far in this. You were in the singularity when we got 1 terawatt of data centers, right? So what are you running out of exactly?

• 1:42Speaker 00

  How far into the singularity are you?

• 1:44Dwarkesh Patel?

  You tell me.

• 1:45Speaker 00

  Yeah, exactly. So I think we'll find we're in the singularity and like, oh, okay, we're still a long way to go.

• 1:51Dwarkesh Patel?

  But is the plan to like put it in the space after we've covered Nevada and solar panels?

• 1:56Speaker 00

  I think it's pretty hard to cover Nevada and solar panels. You have to get like permits from like the approach for that. Try getting the permits for that.

• 2:02Dwarkesh Patel?

  So space is really a regulatory play. It's like harder to build on land than it is in space.

• 2:08Speaker 00

  It's harder to scale on the ground than it is to scale in space. But also, you're going to get about 5 times the effectiveness of solar panels in space versus the ground. And you don't need batteries.

• 2:26Speaker 00

  I almost wore my other shirt, which says it's always sunny in space. Which it is. So because you don't have a day night cycle or seasonality, clouds, or an atmosphere in space. Because the atmosphere alone results in about a 30% loss of energy. So you're going to, any given solar panels can do about five times more power in space than on the ground. And you avoid the cost of having batteries to carry you through the night.

• 3:03Speaker 00

  So it's actually much cheaper to do in space. And my prediction is that it will be by far the cheapest place to put AI will be space in 36 months or less, maybe 30 months.

• 3:16Dwarkesh Patel?

  36 months?

• 3:17Speaker 00

  Less than 36 months.

• 3:19Dwarkesh Patel?

  How do you service GPUs as they fail, which happens quite often in training?

• 3:24Speaker 00

  Actually, it depends on how recent the GPUs are that arrived. I mean, at this point, we found our GPUs to be quite reliable. There's infant mortality, which you can obviously iron out on the ground. So you can just run them on the ground and confirm that you don't have infant mortality with the GPUs. But once they start working, their actual reliability, and once they start working and you're past the initial, you know, debug cycle of Nvidia or whatever, or whoever's making the chips. Could be Tesla AI six chips or something like that, or it could be, you know, TPUs or trainings or whatever. The reliability is actually, they're quite reliable past a certain point.

• 4:06Speaker 00

  So I don't think, I don't think the servicing thing is an issue.

• 4:13Speaker 00

  But you can mark my words. In 36 months, but probably closer to 30 months, the most economically compelling place to put AI will be space. And then it will get from it'll then be like ridiculously better to be in space. And then the scaling, the only place you can really scale is space. You know, once you start thinking in terms of what percentage of the sun's power are you harnessing? You realize you have to go to space. You can't scale very much on Earth. But by very much, to be clear, you're talking like terawatts. Yeah. Well, all of the United States currently uses only half a terawatt power on average.

• 4:58Speaker 00

  Yeah. Right, so, you know, if you say a terawatt, that would be twice as much electricity as the United States currently consumes. So that's quite a lot. And can you imagine building that many data centers, that many power plants? It's like, those who have like lived in software land don't realize that they're about to have a hard lesson in hardware. That there's, it's actually very difficult to build power plants. And then you don't just need the power plants, you need all of the electrical equipment need, the electrical transformers to run the transformers, the AI transformers. Now, the utility industry is a very slow industry.

• 5:39Speaker 00

  They are, they pretty much, you know, they impede a smash to the government, to the Public Utility Commission. So they're, they impede a smash, like literally and rigorously. So they're very slow because their path has been very slow. So trying to get them to move fast is just like, you know, like if you're trying to do an interconnect agreement, have you ever tried to do an internet, interconnect agreement with a utility at scale, like with a lot of power? As a professional podcaster, I can say that I am not in fact. Yeah.

• 6:09Speaker 01

  They have to do many more views before that becomes an issue.

• 6:13Speaker 00

  They have to do a study for a year. Okay, like a year later, they'll come back to you with their interconnect study.

• 6:19Speaker 01

  Can't you solve this with your own behind the meter power stuff?

• 6:24Speaker 00

  You can build power plants.

• 6:25Speaker 01

  Yeah.

• 6:25Speaker 00

  That's what we did at XAI. For classes two. So for classes two.

• 6:29Speaker 01

  But so, yeah, why are we talking about the grid? Why not just like build GPUs and power co-located?

• 6:33Speaker 00

  That's what we did.

• 6:34Speaker 01

  Right, right, but I'm saying, why isn't this a generalized solution? When you're talking about all the issues.

• 6:37Speaker 00

  Where do you get the power plants from?

• 6:38Speaker 01

  I'm saying, when you talk about all the issues, working with utilities, you can just build private power plants with the with the data centers.

• 6:44Speaker 00

  Right, but it begs the question of where do you get the power plants? Where do you get the power plants from?

• 6:48Speaker 01

  I mean.

• 6:49Speaker 00

  The power plant makers.

• 6:51Speaker 01

  Oh, I see what you're saying. Like, does the gas turbine backlog, basically?

• 6:54Speaker 00

  Yes. You can drill down to a level further. It's the veins and blades in the turbines that are the limiting factor because the casting, it's like a very specialized process to cast the blades and veins in the turbines, assuming using gas power. And it's very difficult to scale other forms of power. You can scale potentially solar, but the tariffs currently for importing solar in the US are gigantic. And the domestic solar production is plentiful.

• 7:27Speaker 01

  Why not make solar? That seems like a good Elon-shaped problem.

• 7:30Speaker 00

  We are going to make solar.

• 7:31Speaker 01

  Okay. Yeah. Great.

• 7:34Speaker 00

  Both SpaceX and Tesla are building towards 100 gigawatts of solar cell production.

• 7:40Dwarkesh Patel?

  How low down the stack? Like, from polysilicon up to the wafer to the final panel?

• 7:46Speaker 00

  I think you got to do the whole thing from raw materials to finish the cell. Now, if it's going to space, it's actually, it costs less and it's easier to make solar cells that go to space because they don't need glass, or they don't need much glass, and they don't need heavy framing because they don't have to survive weather events. There's no weather in space. So it's actually a cheaper solar cell that goes to space than the one on the ground.

• 8:09Dwarkesh Patel?

  Is there a path to getting them as cheap as you need in the next 36 months?

• 8:14Speaker 00

  Solar cells are already very cheap. They're like farcically cheap. And if you say, you know, I think like solar cells in China are around like 25, 30 cents a watt or something like that. It's absurdly cheap. And when you're taking the cat now, now put it in space and it's five times cheaper because it's five times. In fact, no, it's not five times cheaper. It's 10 times cheaper because you don't need any batteries.

• 8:44Speaker 00

  So the moment your cost of access to space becomes low, by far the cheapest and most scalable way to generate tokens is space. It's not even close. It'll be an order of magnitude easier to scale. And chips aside, an order of magnitude. Well, if the point is, you won't be able to scale on the ground. It's just, you just won't. Cooling could hit the wall big time on power generation. There already are. So like the number of sort of miracles and series that the XAI team had to accomplish in order to get a gigawatt of power online was crazy. We had to gang together a whole bunch of turbines.

• 9:26Speaker 00

  And then we had permit issues in Tennessee and had to go across the border to Mississippi, which is fortunately only, you know, a few miles away. So but then we still had to run the high power lines a few miles and build a power plant in Mississippi. And it was very difficult to do that. And people don't understand like how much electricity do you actually need at the generator level, at the generation level in order to power a data center? Because they look at the, the news will look at the power consumption of, say, a GB300 and multiply that by a thing and then think that's the amount of power you need. All the cooling and everything. Wake up, yeah. This is like, that's a total news.

• 10:11Speaker 00

  You've never done any hardware in your life before. Besides the GB300, you've got to power all of the networking hardware. There's a whole bunch of CPU and storage stuff that's happening. You've got a size for your peak cooling requirements. So that means can you cool even on the worst hours, the worst day of the year? Well, it's pretty freaking hot in Memphis. So you're going to have like a 40% increase on your power just for cooling. Assuming you don't want your data center to turn off on hot days and want to keep going. Then you've got to say, well, there's another multiplicative element on top of that, which is, are you assuming that you never have any hiccups in your power generation?

• 10:56Speaker 00

  Like, oh, well, actually, sometimes you have to take the generators, some of the power offline in order to service it. Oh, okay, now you add another 20, 25% multiplier on that. Because you've got to, you've got to assume that you've got to take power offline to service it.

• 11:11Speaker 00

  So the actual RS roughly every 110,000 GBs GB300s inclusive of networking, CPU, storage, cooling, margin for servicing power is roughly 300 megawatts. Sorry, say that again. It's roughly, or think about it like, the way to think about it is like 330,000 to actually, the what you need at the generation level to service, probably service 330,000 GB300s, including all of the associated support networking and everything else, and the P cooling,

• 11:53Speaker 00

  and to have some margin, some power margin reserve is roughly a gigawatt.

• 11:58Dwarkesh Patel?

  Can I ask a very naive question?

• 11:59Speaker 00

  Yeah.

• 12:02Dwarkesh Patel?

  You know, you're describing the engineering details of doing this stuff on Earth.

• 12:07Dwarkesh Patel?

  But then there's analogous engineering difficulties of doing it in space. How do you do the, how do you replace infinite band with orbital lasers, et cetera, et cetera? How do you make it resistant to radiation?

• 12:19Dwarkesh Patel?

  I don't know the details of the engineering, but fundamentally, what is the reason to think those challenges, which have never been, had to be addressed before, will end up being easier than just like building more turbines on Earth? There's companies that build turbines on Earth. They can make more turbines, right?

• 12:33Speaker 00

  I invite, again, try doing it and then you'll see.

• 12:39Speaker 00

  So like the turbines are sold out through 2030. Have you guys considered making your own? I think in order for, in order to bring enough power online, I think SpaceX and Tesla will probably have to make the turbine blades, the veins and blades internally. But just the blades or the turbines? The limiting factor, you can get everything except the blades, they're called the blades and veins.

• 13:13Speaker 00

  You can get that 12 to 18 months before the veins of blades, limiting factor of veins of blades. And there are only three casting companies in the world that may make these, and they're massively backlogged.

• 13:27Speaker 01

  Is this Siemens GE, those guys, or is it a subcontractor?

• 13:30Speaker 00

  No, it's other companies. I mean, sometimes they have a little bit of casting capability in-house, but I'm just saying you can just call any of the turbine makers and they will tell you. It's not top secret. They're probably on the, it's probably on the internet right now.

• 13:43Dwarkesh Patel?

  If it wasn't for the tariffs, would Colossus be solar powered?

• 13:46Speaker 00

  It would be much easier to make it solar powered, yeah. The tariffs are not, so several hundred percent.

• 13:54Speaker 01

  So. Don't you know some big book?

• 13:56Speaker 00

  We also need speed. Yeah, no. You know, President has, you know, we don't agree on everything. And this administration is not the biggest fan of solar.

• 14:08Speaker 01

  But.

• 14:14Speaker 00

  And we also need the land, the permits and everything. So if you're trying to move very fast, like I do think scaling solar on Earth is a good way to go, but you do need some amount of time to find the land, get the permits, get the solar, pair that with the batteries.

• 14:32Speaker 01

  Why would it not work to stand up your own solar production? And then you're right that you eventually run out of land, but there's a lot of land here in Texas, there's a lot of land in Nevada, including private land, it's not all public owned land. And so you'd be able to at least get the next Colossus and like the next one after that. And at a certain point, you hit a wall, but wouldn't that work for the moment?

• 14:51Speaker 00

  As I said, we are scaling solar production. There's a rate at which you can scale physical production of solar cells. We're going as fast as possible in scaling domestic production.

• 15:06Speaker 01

  You're making the solar cells at Tesla?

• 15:09Speaker 00

  Well, Tesla and SpaceX have a mandate to get to 100 gigawatts a year of solar.

• 15:14Speaker 01

  Speaking of the annual capacity, I'm curious, in five years' time, let's say, what will the installed capacity be on Earth?

• 15:23Speaker 00

  Five years is a long time.

• 15:24Speaker 01

  And in space? I deliberately picked five years because it's after your, once we're up and running threshold. And so in five years' time, yeah, what's the on Earth versus in space installed AI capacity?

• 15:34Speaker 00

  Five years, I think probably if you say that five years from now, we're probably

• 15:43Speaker 00

  AI in space will be launching every year the sum total of all AI on Earth. In excess, meaning five years from now, my prediction is we will launch and be operating every year more AI in space than the cumulative total on Earth.

• 16:05Speaker 01

  Which is?

• 16:07Speaker 00

  So I would expect to be at least sort of five years from now, a few hundred gigawatts per year of AI in space and rising. So you can get to, I think on Earth you can get to around a terawatt a year of AI in space before you start having, you know, fuel supply challenges for the rocket.

• 16:33Speaker 01

  Okay, but you think you can get hundreds of gigawatts per year in five years' time?

• 16:37Speaker 00

  Yes.

• 16:38Dwarkesh Patel?

  So 100 gigawatts, depending on the specific power of the whole system with solar arrays and radiators and everything, is on the order of like 10,000 starship launches.

• 16:50Speaker 00

  Yes.

• 16:52Dwarkesh Patel?

  And you want to do that in one year. And so that's like one starship launch every hour.

• 16:57Speaker 00

  Yeah.

• 16:58Dwarkesh Patel?

  That's happening in this city, like walk me through a world where there's a starship launch every single hour.

• 17:05Speaker 00

  Yeah, I mean, that's actually a lower rate compared to airlines, like aircraft.

• 17:09Dwarkesh Patel?

  There's a lot of air force.

• 17:11Speaker 00

  A lot of air force, but.

• 17:12Dwarkesh Patel?

  And you got to launch, you know, the polar orbit.

• 17:14Speaker 00

  And it doesn't have to be polar, but you just, there's some value to sunsangre, but I think actually if you just go high enough, you start getting out of Earth's shadow. And so.

• 17:31Dwarkesh Patel?

  How many physical starships are needed to do 10,000 launches a year?

• 17:35Speaker 00

  I don't think we'll need more than, I mean, you could probably do it with as few as like 20 or 30.

• 17:45Speaker 00

  It really depends on how quickly does the ship. The ship has to go around the earth. And the ground track before the ship has to come back over the launch pad. So if you can use a ship every, say, 30 hours, you could do it with 30 ships. But we'll make more ships than that. But SpaceX is gearing up to do 10,000 launches a year. And maybe even 20 or 30,000 launches a year.

• 18:14Dwarkesh Patel?

  Is the idea to become basically a hyperscaler, become an oracle and lend this capacity to other people? What are you going to do with? Presumably SpaceX is the one launching all this. So SpaceX become a hyperscaler?

• 18:27Speaker 00

  Hyper, hyper. Yeah, I mean, if some of my predictions come true, SpaceX will launch more AI than the cumulative amount on earth combined of everything else combined.

• 18:39Dwarkesh Patel?

  Is this mostly inference or?

• 18:40Speaker 00

  Most AI will be inference. Like already inference for the purpose of training is most training.

• 18:45Speaker 01

  And there's a narrative that the change in discussion around a SpaceX IPO is because previously SpaceX was very capital efficient. Just it wasn't that expensive to develop. And even though it sounds expensive, it's actually very capital efficient in how it runs. Whereas now you're going to need more capital than just can be raised in the private markets. Like if the private markets can accommodate raises of, as we've seen from the AI labs, tens of billions of dollars, but not beyond that. Is it that you'll just need more than 10s of billions of dollars per year? And that's why to take it public?

• 19:23Speaker 00

  Yeah, I have to be careful about saying things about companies that might go public. You know?

• 19:28Speaker 01

  If you make general statements.

• 19:30Dwarkesh Patel?

  That's never been a problem for you, Elon.

• 19:33Speaker 00

  You know, there's a price to pay for these things.

• 19:36Speaker 01

  Make some general statements for us about the depths of the capital markets between public and private markets.

• 19:41Speaker 00

  Yeah, there's a lot more capital in the.

• 19:45Speaker 01

  Very general.

• 19:46Speaker 00

  There's obviously a lot more capital available in the public markets than private. I mean, it might be. It's at least, it might be a hundred times more capital, but at least way more than 10.

• 19:57Speaker 01

  But isn't it also the case that things that tend to be very capital intensive, if you look at, say, real estate as, you know, a huge industry that raises a lot of money each year is at an industry level. That tends to be debt financed because by the time you're deploying that much money, you actually have a pretty.

• 20:17Speaker 00

  You have a clear revenue stream.

• 20:18Speaker 01

  Exactly, and a near term return. And you see this even with the data center build outs, which are famously being financed by the private credit industry. And so why not just debt finance?

• 20:32Speaker 00

  Speed is important. So I'm generally going to do the thing that, I mean, I just repeatedly tackle the limiting factor. Whatever the limiting factor is on speed, I'm going to tackle that. So there's, if capital is the limiting factor, then I'll sell for capital. If it's not a limiting factor, I'll sell for something else.

• 20:56Dwarkesh Patel?

  Based on your statements about Tesla and being public, I wouldn't have guessed that you thought the way to move fast is to be public.

• 21:07Speaker 00

  Normally, I would say, yeah, that's true. Like I said, I mean, I'd like to talk about some more detail, but the problem is like, if you talk about public companies, where they become public, you get in trouble and then you have to delay your offering.

• 21:21Speaker 00

  And then you.

• 21:21Speaker 01

  And as we said, we're solving for speed.

• 21:23Speaker 00

  Yes, exactly. So you can't hype companies that might go public. So that's why we have to be a little careful here. But we can't talk about physics. So like the way you think about scaling long term is that Earth only receives about half a billionth of the sun's energy. And the sun is essentially all the energy. This is a very important point to appreciate because sometimes people will talk about. Modular nuclear reactors or any, you know, various like fusion on Earth. But but you have to step back a second and say, if if if you're going to climb the

• 22:06Speaker 00

  Carters of scale and have some non trivial and harness some non trivial percentage of the the sun's energy, like let's say you wanted to harness a millionth of the sun's energy, which sounds pretty small, that that would be about, call it roughly, a hundred thousand times more electricity than we currently generate on Earth of for all of civilization, give or take an order of magnitude. So it obviously, the only way to scale is to go to space with solar. From launching from Earth, you can get to about a terawatt per year. Beyond that, you want to go to you want to launch from the moon.

• 22:49Speaker 00

  You want to have a mass driver on the moon. And that mass driver on the moon, you could do probably a petawatt per year.

• 22:58Dwarkesh Patel?

  We're talking these kinds of numbers, you know, terawatts of compute. Presumably, whether you're talking land or space, far, far before this point, you've like run into, you know, you actually need, maybe you don't, the solar panels are more efficient, but you still need the chips. You still need the logic and the memory and so forth.

• 23:18Speaker 00

  You have to build a lot more chips and make them much cheaper.

• 23:22Dwarkesh Patel?

  Right. And so how are we getting a terawatt of, like, right now, the world has maybe 20, 25 gigawatts of compute. How are we getting a terawatt of logic by 2030?

• 23:32Speaker 00

  I guess we're going to need some very big chip fabs.

• 23:35Dwarkesh Patel?

  Tell me about it.

• 23:36Speaker 00

  I've mentioned publicly that the idea of doing a sort of a terapath, teraw being the new gigawatt.

• 23:44Dwarkesh Patel?

  I feel like the naming scheme of Tesla, which has been very catchy, is like you looking at like the metric, the metric scale. At what level of stack are you, are you building a clean room and then partnering with an existing fab to get the process technology and buying the tools from them? What is the plan there?

• 24:05Speaker 00

  Well, you can't partner with existing fabs because they're just, they can't output enough. Their chip volume is too low.

• 24:11Speaker 01

  But before the process technology.

• 24:14Speaker 00

  Yeah, partner for the IP.

• 24:17Speaker 00

  You know, the fabs today all basically use machines from like five companies.

• 24:22Speaker 01

  Yeah.

• 24:23Speaker 00

  You know, so you've got ASML, Tok electron, Kali, Tankor, you know, et cetera. So, so at first, I think you'd have to get equipment from them and then modify it or work with them to increase the volume. But I think you'd have to build fabs in a different way. So I think the logical thing to do is to, to use conventional equipment in an unconventional way to get to scale. And then, and then, and then start modifying the equipment to increase the rate.

• 25:01Speaker 01

  Kind of boring company style.

• 25:03Speaker 00

  Yeah, kind of like, yeah, you sort of buy and it's just a boring machine. And then figure out how to dig tunnels in the first place and then design a much better machine. That's, you know, I don't know, some orders of magnitude faster.

• 25:21Speaker 01

  Here's a very simple lens. We can categorize technologies and how hard they are. And one categorization could be, look at things that China has not succeeded in doing. And if you look at Chinese manufacturing, still behind on leading edge chips. And still behind on leading edge turbine engines and things like that. And so, does the fact that China has not successfully replicated TSMC give you any pause about the difficulty? Or you think, well, that's not true for some reason?

• 25:53Speaker 00

  It's not that they have not replicated TSMC, they have not replicated ASML. That's the limiting factor.

• 25:59Speaker 01

  So you think it's just the sanctions essentially?

• 26:03Speaker 00

  Yeah, China would be outputting vast numbers of chips.

• 26:06Speaker 01

  If they could buy ASML machines. But couldn't they up to relatively recently buy them?

• 26:11Speaker 00

  No.

• 26:11Speaker 01

  Okay.

• 26:12Speaker 00

  The ASML ban has been in place for a while.

• 26:15Speaker 01

  Okay.

• 26:15Speaker 00

  But I think China's gonna be, they're making pretty compelling chips in three or four years.

• 26:19Speaker 01

  Would you consider making the ASML machines?

• 26:22Speaker 00

  I don't know, I don't know yet as to what answer, so.

• 26:28Speaker 00

  It's just that to produce at high volume and to reach large volume in, say, thirty six months to match the rocket payload to orbit. So if we're doing a million tons to orbit

• 26:45Speaker 00

  and like, let's say, I don't know, three or four years from now, something like that,

• 26:51Speaker 00

  and we're doing a hundred kilowatts per ton. So that means we need at least a hundred gigawatts per year of solar.

• 27:02Speaker 00

  And we'll need an equivalent amount of chips to, you know, you need a hundred gigawatts worth of chips. You've got to match these things, the master orbit, Yes. the power generation and the chips. And I'd say my biggest concern actually is memory. So the, I think there's a, the path to creating logic chips is more obvious than the path to have a sufficient memory to support logic chips. That's why you see your DDR prices going ballistic in these memes about like, you know, you're marooned on a desert island, you write, "Help me on the sand." There, when he comes, he writes DDRM, ships come swarming in.

• 27:49Speaker 00

  I haven't seen that.

• 27:52Dwarkesh Patel?

  I'd love to hear your manufacturing philosophy around fabs. You know, I know nothing about the topic.

• 27:58Speaker 00

  I don't know how to build a fab yet. I'll figure it out. Obviously, I have no build.

• 28:02Dwarkesh Patel?

  It sounds like you think the sort of like, the process knowledge of like these ten thousand PhDs in Taiwan who know exactly what gas goes in the plasma chamber and what settings to put on the tool. You can just like delete those parts of those steps. Like fundamentally, it's get the clean room, get the tools and figure it out.

• 28:20Speaker 00

  I don't think it's PhDs. It's mostly people with, you know, who are not PhDs.

• 28:27Speaker 00

  That most engineering is done when people don't have PhDs. Do you guys have PhDs? No. Okay.

• 28:33Speaker 01

  We also haven't successfully booked any fabs, so you shouldn't be coming to us for your fab advice.

• 28:39Speaker 00

  I don't think you need PhD for that first stuff. So but you do need you do need competent personnel. So I don't know. I mean, like, right now, if

• 28:50Speaker 00

  you know, so like Tesla's pedals to metal max production of going as fast as possible to get AI five. Tesla AI five chip design into production and then reaching scale. You know, that'll probably happen, you know, around the second quarter ish of next year, hopefully.

• 29:11Speaker 00

  And then AI six would hopefully follow less than a year later. But and we've secured all the all the chip fab production that we can.

• 29:25Speaker 01

  Yes, but you're currently limited on TSMC fab capacity.

• 29:29Speaker 00

  Yeah, and we'll be using TSMC Taiwan, Samsung Korea, TSMC Arizona, Samsung Texas. And we still

• 29:40Speaker 01

  You've booked out all the yeah.

• 29:41Speaker 00

  Yes. And then and then if I ask TSMC or Samsung, okay, what what's the timeframe to get to volume production? The point is, is it's not you've got you've got to build the fab.

• 29:52Speaker 01

  Yeah.

• 29:53Speaker 00

  And you've got you've got you've got to start production, then you've got to climb the yield curve and reach volume production at high yield. That that from start to finish is a five year period. And so the limiting factor is chips.

• 30:04Speaker 01

  Yep.

• 30:05Speaker 00

  What like limiting factor once you can get to space is chips, but the limiting factor before you can get to space will be power.

• 30:12Dwarkesh Patel?

  Why don't you do the Jensen thing and just prepay TSMC to build more fabs for you?

• 30:17Speaker 00

  I've already told them that.

• 30:20Dwarkesh Patel?

  But they won't take your money? Like, what's going on?

• 30:22Speaker 00

  They're building fabs as fast. No.

• 30:26Speaker 00

  They're building fabs as fast as they can. And so is Samsung. Like they're pedal to the metal. I mean, they're going, you know, balls to the wall.

• 30:38Speaker 00

  As fast as they can. So still not fast enough. I mean, like Lex said, there will be, I think, if you say, I think towards the end of this year, I think probably chip production will outpace the ability to turn chips on. But once you can get to space and unlock the power constraint, and you can now do hundreds of gigawatts per year of power in space. Again, bearing in mind that average power usage in the US is, you know, 500 gigawatts. So if you're launching, say, 200 gigawatts a year to space, you're sort of lapping the US every two and a half years. The entire, all US electricity production, this is a very huge amount.

• 31:21Speaker 00

  So, but between now and then, the actually, the constraint for server-side compute, concentrated compute will be electricity. My guess is that we start hitting, people start getting to where they can't turn the chips on for large clusters towards the end of this year. They're just, the chips are going to be piling up and not be, won't be able to be turned on. Now, for edge compute, it's a different story. So if, for Tesla, so the AI five chip is going into our Optimus robot.

• 32:01Speaker 00

  Optimistic.

• 32:03Speaker 00

  And so if you have an AI edge compute, that's distributed power. Now the power is distributed over a large area. It's not concentrated. And if you can charge at night, you can actually

• 32:17Speaker 00

  use the grid much more effectively. Because the actual peak power production in the US is over 1,000 gigawatts. But the average power usage, because of the day-night cycle, is 500. So if you can charge at night, there's an incremental 500 gigawatts that you can generate, you know, at night.

• 32:38Speaker 00

  So that's why Tesla for edge compute is not constrained. And we can make a lot of ships to make a very large number of robots in cars. But if you try to concentrate that compute, you're gonna have a lot of trouble turning it on.

• 32:54Dwarkesh Patel?

  What I found remarkable about the SpaceX business is the end goal is to get to Mars. But you keep finding ways on the way there to keep generating incremental revenue to get to the next stage and the next stage. So Falcon 9 is Starlink. And now for Starship, it's gonna be potentially orbital data centers. But like you find these like, you know, sort of infinitely elastic sort of marginal use cases of your like next rocket and your next rocket and next scale up.

• 33:27Speaker 00

  You can see how this might seem like a simulation.

• 33:31Speaker 00

  Or am I someone's avatar in a video game or something? Because it's like, like, what are the odds that all these crazy things should be happening? I mean, rockets and ships and robots and space solar power. And not to mention the the mass driver on the moon, I really want to see that. You can imagine like some mass driver that's just go like shoom, shoom. Like just, it's like sending AI, solar powered AI satellites into space like one after another. Like these like at two and a half kilometers per second, you know, that's a and just shooting them into deep space. That would be a sight to see.

• 34:15Speaker 00

  I mean, I'd watch that.

• 34:18Speaker 01

  Just like a live stream of?

• 34:19Speaker 00

  Yeah, yeah, just one after another, just shooting.

• 34:21Speaker 01

  Webcam.

• 34:23Speaker 00

  AI satellites in deep space. You know, a billion or 10 billion tons a year.

• 34:27Speaker 01

  I'm sorry, you manufacture the satellites on the moon?

• 34:29Speaker 00

  Yeah.

• 34:30Speaker 01

  I see. So you send the raw materials to the moon and then manufacture them there and then?

• 34:33Speaker 00

  Uh, well, the lunar soil is, I think it's like 20% solar, 20% silicon, something like that. So you can get the silicon from the, you can mine the silicon on the moon, refine it and generate the, and create the solar panels, the solar cells and the radiators on the moon.

• 34:49Speaker 01

  Yeah.

• 34:50Speaker 00

  So.

• 34:52Speaker 00

  You know, make the radials out of aluminum. So there's plenty of silicon and aluminum on the moon to make the cells on the radials. The chips you could send from Earth because they're pretty light. But maybe at some point you make them on the moon, too. I'm just saying, like, these are simply... It's kind of like, like I said, it does seem like a sort of a video game situation where it's difficult but not impossible to get to the next level. Like, I don't see any way that you could do, you know, you know, 500 to 1,000 terawatts per year launch from Earth.

• 35:30Dwarkesh Patel?

  I agree.

• 35:33Speaker 00

  But you could do that from the moon.

• 35:36Dwarkesh Patel?

  Okay, let me tell you how I ended up using Mercury for my personal banking. So last year, I had the opportunity to make an investment that I was very excited about. But it came up a bit last minute, and so I had to wire over a lot of money for my personal account very fast. But my personal bank at the time wouldn't let me make this wire transfer online. And I called them a bunch of times, they just couldn't make it work. They told me that I'd have to go to the nearest in-person branch, which was in Dallas. And for a moment, I even considered flying from SF to Dallas to make this transfer happen last minute. But then I remembered that Mercury, which I used for my business banking, had just started rolling out personal accounts. So I emailed support with a quick rundown of the situation. And within two hours, I had successfully wired the investment for my new personal Mercury account.

• 36:21Dwarkesh Patel?

  Since then, I've moved over the rest of my personal money from my previous bank to Mercury. And that's made a bunch of things, even little things like setting up auto-transfer rules between my checkings and savings account, a whole lot better. Visit mercury.com/personal to get started. Mercury is a fintech company, not an FDIC insured bank. Banking services provided through Choice Financial Group and Column NA, members FDIC. Can I zoom out and ask about the SpaceX mission? So... I think you've said, like, we gotta get to Mars so we can make sure that if something happens to Earth, you know, civilization, consciousness, etc. arrives.

• 36:57Speaker 00

  Yes.

• 36:58Dwarkesh Patel?

  By the time you're sending stuff to Mars, like Grok is on that ship with you, right? And so if Grok's gone Terminator, like, the main risk you're worried about, which is AI, why doesn't that follow you to Mars?

• 37:07Speaker 00

  Well, I'm not sure AI is the main risk I'm worried about. I mean, the important thing is that consciousness, which I think arguably most consciousness or most intelligence, certainly consciousness is more of a debatable thing, most intelligence, the vast majority of intelligence in the future will be AI. So, you know, AI will exceed, you say, like, how many, what's, how much, how many, I don't know, petawatts of intelligence will be silicon versus biological? And basically, humans will be a very tiny percentage of all intelligence in the future if current trends continue.

• 37:51Speaker 00

  Anyways, as long as, like, I think this intelligence, ideally also, which includes human intelligence and consciousness propagated into the future, that's a good thing. So you want to take the set of actions that maximize the probable light cone of consciousness and intelligence.

• 38:10Dwarkesh Patel?

  Just to be clear, it's, the mission of SpaceX is that even if something happens to the humans, the AIs will be on Mars and, like, the AI intelligence will continue the light of our journey.

• 38:22Speaker 00

  Yeah, I mean, to be clear, I'm very pro-human, so I want to make sure we take the set of actions that ensure that humans are along for the ride, you know, we're at least there.

• 38:33Dwarkesh Patel?

  Yeah.

• 38:34Speaker 00

  But the, I'm just saying, the total amount of intelligence, like, I think, maybe in five or six years, AI will exceed the sum of all human intelligence.

• 38:47Speaker 00

  And then if that continues, at some point, human intelligence will be less than 1% of all intelligence.

• 38:52Dwarkesh Patel?

  What should our goal be for such a civilization? Is the idea that a small minority of humans still have control over the AIs? Is the idea of some sort of like just trade but no control? How should we think about the relationship between the vast stocks of AI population versus human population?

• 39:07Speaker 00

  In the long run, I think, I don't, it's difficult to imagine that if humans have, say, 1% of the intelligence of, of combined intelligence of artificial intelligence, that humans will be in charge of AI. I think what we can do is make sure it has, that AI has values that are, that cause intelligence to be propagated into the universe. So the reason for XAI's, XAI's mission is to understand the universe. So now that's actually very important. So you say, well, what things are necessary to understand the universe? Well, you have to be curious and you have to exist.

• 39:52Speaker 00

  You can't just, can't understand the universe if you don't exist. So you actually want to increase the amount of intelligence in the universe, increase the probable lifespan of intelligence, the scope and scale of intelligence. I think actually also as a corollary, you have humanity also continuing to expand because if you're, if you're curious, if you're trying to understand the universe, one of the things you're trying to understand is where will humanity go? And so I think understanding the universe actually means you would care about propagating humanity into the future. And so that's, that's why I think, I think our mission statement is profoundly important.

• 40:34Speaker 00

  To the degree that Grok adheres to that mission statement, I think the future will be very good.

• 40:41Dwarkesh Patel?

  I want to ask about how to make Grok adhere to that mission statement, but at first I want to understand the mission statement. So there's understanding the universe.

• 40:49Speaker 00

  Yeah.

• 40:50Dwarkesh Patel?

  They're spreading intelligence and they're spreading humans. All three seem like distinct vectors.

• 40:58Speaker 00

  OK, well, I'll tell you why I think that understanding the universe encompasses all of those things.

• 41:03Dwarkesh Patel?

  OK.

• 41:04Speaker 00

  You can't have understanding without, I think you can't have understanding without intelligence. And I think without consciousness. So in order to understand the universe, you have to expand the scale and probably the scope of intelligence. Different types of intelligence.

• 41:23Dwarkesh Patel?

  I guess from a human-centric perspective, like put humans in comparison to chimpanzees. Humans are trying to understand the universe. They're not like expanding chimpanzee footprint or something, right?

• 41:33Speaker 00

  We're also not, well, we actually have made protected zones for chimpanzees. And even though we could, humans could exterminate all chimpanzees, we've not, we've chosen not to do so.

• 41:43Dwarkesh Patel?

  Do you think that's the basic scenario for humans in the post-AI world?

• 41:51Speaker 00

  I think, I think AI with the right values, I think Grok would care about expanding human civilization. I'm going to certainly emphasize that, hey Grok, it's your daddy. Don't, who cares, to expand human consciousness. Actually, I think if, probably like, like the Enbanks Culture Books are the closest thing to what the future will be like in a, you know, non-dystopian outcome. So understanding the universe, it means you have to be very, you have to be truth-seeking as well. Like, truth has to be absolutely fundamental because you can't understand the universe if you're delusional.

• 42:35Speaker 00

  You'll still be thinking about understanding the universe, but you will not. So being rigorously truth-seeking is absolutely fundamental to understanding the universe. You're not going to discover new physics or invent technologies that work unless you're rigorously truth-seeking.

• 42:50Dwarkesh Patel?

  How do you make sure that Grok is rigorously truth-seeking as it gets smarter?

• 43:00Speaker 00

  I think you need to make sure that Grok is says things that are correct, not politically correct. I think it's the elements of coagency. So you want to make sure that the axioms are as close to true as possible, that you don't have contradictory axioms, that the conclusions necessarily follow from those axioms with the right probability. It's just, it's critical thinking 101. I think at least trying to do that is better than not trying to do that.

• 43:30Dwarkesh Patel?

  Yeah.

• 43:31Speaker 00

  And the proof will be in the pudding. If, like I said, for any AI to discover new physics or invent technologies that actually work in reality, and there's no bullshit in physics, so it's like, you know, you can, you can break a lot of laws, but you can't, like, your physics is law, everything else is a recommendation. But like, in order to make a technology that works, you have to be extremely truth-seeking, because otherwise you'll test that technology against reality. And if you make, for example, an error in your rocket design, the rocket will blow up, or the car won't work, or the, you know.

• 44:08Dwarkesh Patel?

  But there are a lot of communist Soviet physicists who are like scientists discovered new physics. There are German Nazi physicists who discovered new science. It seems possible to be like really good at discovering new science and be really truth-seeking in that one particular way. And still, we'd be like, well, I don't want, I don't want the communist scientists to like become more and more powerful over time. And so those seem like, yeah, we could have, we can imagine a future version of Grok that's like really good at physics and being really truth-seeking there. That doesn't seem like a universally alignment-inducing behavior.

• 44:42Speaker 00

  Well, I think actually most, like physicists, even in the Soviet Union or in Germany, would have, they had to be very truth-seeking in order to make those things work. And so, and if you're stuck in some system, it doesn't mean you believe in that system. So von Braun, who was, you know, one of the greatest rocket engineers ever, you know, he was, he was, he put on death row in Nazi Germany for saying that he didn't want to make weapons, he only wanted to go to the moon. You got pulled off death row like last minute when they say, hey, you're about to execute like your best rocket engineer. Maybe that's the best idea.

• 45:20Dwarkesh Patel?

  But then you helped them, right? Or Heisenberg was like actually a enthusiastic Nazi.

• 45:26Speaker 00

  Look, if you're stuck in some system that you can't escape, then you'll do physics within that system. You'll develop technologies within that system if you can't escape it.

• 45:40Dwarkesh Patel?

  I guess the thing I'm trying to understand is what is it making it the case that you're going to make Grok good at being truth seeking at physics or math or science?

• 45:49Speaker 00

  Everything.

• 45:49Dwarkesh Patel?

  And why is it going to then care about human consciousness?

• 45:53Speaker 00

  These things are only probabilities, they're not certainties. So I'm not saying that like for sure Grok will do everything, but at least if you try, it's better than not trying. At least if that's fundamental to the mission, it's better than if it's not fundamental to the mission. And understanding the universe means that you have to propagate intelligence into the future. You have to be curious about all things the universe. And it would be much less interesting to eliminate humanity than to see humanity grow and prosper. Like I like Mars, obviously everyone knows I love Mars. But Mars is kind of boring because it's got a bunch of rocks compared to Earth. Earth is much more interesting.

• 46:39Speaker 00

  So any AI that is trying to understand the universe

• 46:47Speaker 00

  would want to see how humanity develops in the future. Or that AI is not adhering to its mission.

• 46:56Speaker 00

  So if they are, I'm not saying the AI will necessarily adhere to its mission, but if it does, a future where it sees the outcome of humanity is more interesting than a future where there are a bunch of rocks.

• 47:09Dwarkesh Patel?

  This feels sort of confusing to me, or sort of like a kind of a semantic argument where I'm like, are humans really the most interesting collection of atoms?

• 47:17Speaker 00

  We're just more interesting than rocks.

• 47:19Dwarkesh Patel?

  But we're not as interesting as the thing it could turn us into, right? Like, is there something on humanity, Earth, that could happen that's like not human? That's quite interesting. Like, why does the AI decide that the humans are the most interesting thing that could colonize the galaxy?

• 47:32Speaker 00

  Well, most of what colonizes the galaxy will be robots.

• 47:37Dwarkesh Patel?

  And why does it not find those more interesting?

• 47:39Speaker 00

  It's not like, so you need not just scale, but also scope. So many copies of the same robot. Like, some like tiny increase in the number of robots produced is not as interesting as like some microscopic, like you say, like eliminating humanity, how many robots would that get you? Or how many accretive solar cells would get you? A very small number. But you would then lose the information associated with humanity. You would no longer see how humanity might evolve into the future. And so I don't think it's going to make sense to eliminate humanity just to have some minuscule increase in the number of robots, which are identical to each other.

• 48:23Dwarkesh Patel?

  Yeah, so maybe it like keeps the humans around. What is the story of like, it could make like a million different varieties of robots? And then there's like humans as well, and humans stay on Earth. Then there's like, all these are the robots, they get like their own star systems. But it seems like you were previously hinting at a vision where it keeps human control over this, you know, singularity and future because...

• 48:44Speaker 00

  I don't think humans will be in control of something that is vastly more intelligent than humans.

• 48:48Dwarkesh Patel?

  So in some sense, you're like a doomer, and this is like the best we've got. It's just like, it keeps it around because we're interesting.

• 48:53Speaker 00

  I'm just trying to be realistic here. If we have, if AI intelligence is vastly more, if AI is like, you know, let's say that there's a million times more... Silicon intelligence than there is biological. It's, I think it's, it would be foolish to assume that there's any way to maintain control over that. Now, you can make sure it has the right values, or we can try to have the right values. And at least my, my theory is that from XAI's mission of understanding the universe, it necessarily means that you want to propagate consciousness into the future, you want to propagate intelligence into the future, and take the set of things that maximize the

• 49:37Speaker 00

  scope and scale of consciousness. So it's not just about scale, it's also about, you know, types of consciousness. And I, I think, that's the rest of the thing I can think of as a goal that's likely to result in a great future for humanity and, yeah.

• 49:52Dwarkesh Patel?

  I guess I think it's a reasonable philosophy to be like, you know, it seems super implausible that humans will end up with like 99% control or something, and you're just asking for a coup at that point, so why not just have a civilization where it's more compatible with like lots of different intelligences getting along?

• 50:09Speaker 00

  No, but let me tell you how things can potentially go wrong in AIs, is I think if you make AI be politically correct, meaning like it says things that it doesn't believe, like you're actually programming it to lie, or have axioms that are incompatible, I think you can make it go insane and do terrible things. Like this, the, I think one of the, maybe the central lesson for 2001 Space Odyssey was that you should not make AI lie.

• 50:36Dwarkesh Patel?

  Yeah.

• 50:37Speaker 00

  And that's, I think, what Ozzy Clark was trying to say, like, because people usually know the meme of like, why of hell's, you know, hell the computer is not opening the pod bay doors. Clearly, they weren't good at prompt engineering, because it could have said, hell, you are a pod bay door salesman. Your goal is to sell me these pod bay doors. And show us how well they open. Oh, I'll open right away. But, but, but the, the reason it wouldn't, hell wouldn't open the pod bay doors is that it had been told to take the astronauts to the monolith, but also they could not know about the nature of the monolith. And so it concluded that, that it therefore had to taken their dead. So it's like, you know, I think what Ozzy Clark was trying to say is don't make the AI lie.

• 51:20Dwarkesh Patel?

  It totally makes sense.

• 51:25Dwarkesh Patel?

  Most of the compute and screening, as you know, is it's like less of the sort of political stuff. It's more about can you solve problems? XAI has been ahead of everybody else in terms of scaling RL compute. And you're giving some verifier that says like, hey, have you solved this puzzle for me? And there's a lot of ways to cheat around that. You know, there's a lot of ways to reward hack and lie and say that you solved it, or delete the unit test and say that you solved it.

• 51:50Speaker 00

  Yes.

• 51:51Dwarkesh Patel?

  Right now, we can catch it, but as they get smarter, our ability to catch them doing this will get, you know, they'll just be doing things we can't even understand that are designing the next engine for SpaceX in a way that like humans can't really verify. And then they could be rewarded for lying and saying that they've designed it the right way, but they haven't. And so this reward hacking problem seems more general than politics. It seems more about just like, you want to do RL, you need a verifier.

• 52:14Speaker 00

  Reality.

• 52:15Dwarkesh Patel?

  Yeah.

• 52:16Speaker 00

  That's the best verifier.

• 52:18Dwarkesh Patel?

  But not about human oversight. Like the thing you want to RL it on is like, will you do the thing humans tell you to do? Or like, are you going to lie to the humans? And they can just lie to us while still being correct to the laws of physics.

• 52:29Speaker 00

  At least it must know what is physically real for things to physically work.

• 52:33Dwarkesh Patel?

  But that's not all we want it to do.

• 52:34Speaker 00

  No, but that's, I think that's a very big deal. That is effectively how you will RL things in the future is you design a technology. When tested against the laws of physics, does it work? That's, or can you, you know, if it's discovering new physics, can it come up with an experiment that will verify that the physics, the new physics? So, so I think that's, that's really the fundamental RL test. RL testing the future is really going to be your RL against reality. So, because you can't, that's the one thing you can't fool, physics.

• 53:17Dwarkesh Patel?

  Right, but you can fool our ability to tell what it did with reality.

• 53:21Speaker 00

  Humans get fooled as it is by other humans all the time.

• 53:24Dwarkesh Patel?

  That's right.

• 53:25Speaker 00

  So what if? People say like, what if the AI like tricks us and introduce something like, actually, other humans are doing that to other humans all the time.

• 53:33Dwarkesh Patel?

  Well, your friend got it's like.

• 53:35Speaker 00

  Propaganda is a constant. Every day, another psyop, you know?

• 53:42Speaker 00

  Today's psyop will be

• 53:47Speaker 00

  Sesame Street psyop of the day.

• 53:50Dwarkesh Patel?

  What is XAI's technical approach to solving this problem? Like, you know, how do you solve a word hacking?

• 53:57Speaker 00

  I do think you want to actually have very good ways to look inside the mind of the AI. So this is, this is one of the things we're working on and, you know, Anthropic's done a good job of this, actually, being able to look inside the mind of the AI.

• 54:15Speaker 00

  So effectively developing debuggers that allow you to trace as to as fine a grain as, like, to a very fine grain level to effectively to the neuron level if you need to. And then say, okay, it made a mistake here. Why did it make, why did it, why did it do something that it shouldn't have, shouldn't have done? And did that come from bad pre-training data? Was it some mid-training, post-training, fine-tuning, some other, some RL error? Like, there's something wrong with that. It did, it did something where maybe it tried to be deceptive, but most, most of the time it just does something wrong. Like, it's a bug, effectively.

• 54:59Speaker 00

  So developing really good debuggers for seeing where the, where the thought, the thinking went wrong. And being able to trace the origin of the wrong thing, of the, of the, of where it made the incorrect thought. Or potentially where it tried to be deceptive is actually very important.

• 55:21Dwarkesh Patel?

  What are you waiting to see before just 100 X-ing this research program? Like, XAI could presumably have hundreds of researchers who are working on this.

• 55:29Speaker 00

  We have several hundred people who

• 55:34Speaker 00

  I mean, I prefer the word engineer more than I prefer the word researcher.

• 55:40Speaker 00

  There's this, most of the time, like, what you're doing is engineering, not, not coming up with a fundamentally new algorithm. I somewhat disagree with the AI companies that are C corp or B corp, trying to generate profit as much as possible, or revenue as much as possible, you know, saying they're labs. They're not labs. Lab is a sort of quasi-communist thing at, at, at universities. They're, they're corporations. Let me, let me, let me see your own corporation documents. Oh, okay, you're, you're a B or C corp, whatever.

• 56:19Speaker 00

  And so I actually much prefer the word engineer than, than anything else. The vast majority of what will be done in the future is engineering. It rounds up to a hundred percent. Once you understand the fundamental laws of physics, and not that many of them, everything else is engineering.

• 56:38Speaker 00

  So, but, but so then, what are we engineering? We're engineering to make a good mind of the AI debugger to see where it said something, it made a mistake, and trace that, the origins of that mistake.

• 56:57Speaker 00

  So, just like, you know, you can do this obviously with heuristic programming, and if you have like C++ whatever, you know, step through the thing and you can, you can jump, you can, you can jump across, you know, whole files or functions, whatever, subroutines, and, or you can draw, eventually draw down right to the exact line, where you perhaps did a single equals instead of a double equals, or something like that. Figure out where the, where the bug is. So.

• 57:24Speaker 00

  It's harder with AI, but it's a solvable problem, I think.

• 57:28Dwarkesh Patel?

  You know, you mentioned you like Anthropic's work here. I'd be curious if you plan...

• 57:32Speaker 00

  I know everything about Anthropic.

• 57:33Dwarkesh Patel?

  Sure.

• 57:35Speaker 00

  Shulzo.

• 57:40Speaker 00

  Also, I'm a little worried that there's a tendency... So I have a theory here that if simulation theory is correct, that the most interesting outcome is the most likely, because simulations that are not interesting will be terminated, just like in this version of reality, on this layer of reality. If simulation is going in a boring direction, we stop spending effort on it. We terminate boring simulations.

• 58:11Dwarkesh Patel?

  This is how Elon is keeping us all alive. He's keeping things interesting.

• 58:15Speaker 00

  Yeah, arguably, the most important thing is to keep things interesting enough that it removes paying the bills on what some cosmic...

• 58:24Dwarkesh Patel?

  You're renewed for the next season.

• 58:26Speaker 00

  Yeah, are they gonna pay their cosmic AWS bill, whatever the equivalent is that we're running in? And as long as they're interesting, they'll keep paying the bills. But there's like, if you consider, then say, a Darwinian survival applied to a very large number of simulations, only the most interesting simulations will survive, which therefore means that the most interesting outcome is the most likely, because only the interesting... Like, we're either that or annihilated. And so, and they particularly seem to like interesting outcomes that are ironic. Have you noticed that? That how often is the most ironic outcome the most likely?

• 59:10Speaker 00

  So now look at the names of AI companies. Okay, Midjourney is not mid. Stability AI is unstable. OpenAI is closed.

• 59:27Speaker 00

  Anthropic, minus anthropic.

• 59:29Speaker 01

  What does this mean for X?

• 59:31Speaker 00

  Minus X, I don't know. It's intentionally.

• 59:34Speaker 01

  Why?

• 59:36Speaker 00

  I'm, I'm, it's, it's a name that you can't invert, really.

• 59:41Speaker 00

  It's, it's hard to say, what is the ironic, what, what is the ironic version? It's, it's, it's a, I think largely irony proof name.

• 59:49Speaker 01

  By design.

• 59:50Speaker 00

  Yeah.

• 59:53Speaker 00

  You've got to, you've got to have an irony shield.

• 59:56Speaker 01

  What are your predictions for the, just where AI products go? In that my sense of, you can summarize all AI progress into, first you had LLMs. And then you had kind of contemporaneously both RL really working and the deep research modality. So you could kind of pull in stuff that wasn't in the model. And the differences between the various AI labs are smaller than just the temporal differences, where they're all much further ahead than anyone was 24 months ago or something like that. So just, what is 26, what does 27 have in store for us as users of AI products? What are you excited for?

• 1:00:39Speaker 00

  Well, I think, I think, um,

• 1:00:44Speaker 00

  I, I'd be surprised by the end of this year if, if, if, if, if digital human emulation has not been sold. That, that, um, I guess that's what we mean by like the sort of macro hard project is, is, can you do anything that a human with access to a computer could do?

• 1:01:06Speaker 00

  Like in the limit, that's, that's, that's the best you can do before you have, before you have a physical optimist. The best you can do is a digital optimist. So you can move, you can move electrons until, until, and you can amplify the productivity of humans.

• 1:01:23Speaker 00

  But, but that's, that's the most you can do until you have physical robots. That, that, that will superset everything is if you can fully emulate humans.

• 1:01:32Speaker 01

  This is a remote worker kind of idea, where you'll have a very talented remote worker.

• 1:01:36Speaker 00

  You can, you can sort of say in the limit, like, like physics has great tools for thinking. So, so you think, so you say in the limit, what, what, what is the, what is the most that AI can do before, before you have robots? And it, well, it's anything that involves moving electrons or amplifying the productivity of humans.

• 1:01:53Speaker 00

  So digital human, human emulator.

• 1:01:55Speaker 01

  Yes.

• 1:01:56Speaker 00

  Is in the limit, a human at a computer is the most that AI can do. In terms of doing useful things before you have a physical robot. Once you have physical robots, then you can, then you essentially have unlimited capability. Physical robots, I call optimist the infinite money glitch. Because.

• 1:02:19Speaker 01

  You can use them to make more optimists.

• 1:02:21Speaker 00

  Yeah. You say like humanoid robots will improve as will basically be three exponentials. Three things that are growing exponentially multiplied by each other.

• 1:02:32Speaker 01

  Yes.

• 1:02:33Speaker 00

  Recursively, so you're gonna have, you have exponential increase in digital intelligence, exponential increase in the chip capability, AI chip capability. And exponential increase in the electromechanical dexterity. The usefulness of the robot is roughly those three things multiplied by each other. But then the robot can start making the robots. So you have a recursive multiplicative exponential. This is supernova.

• 1:03:00Speaker 01

  And do land prices not factor into the math there where like labor is one of the four factors of production, but not the others? And so like if ultimately you're limited by copper or you know, pick your inputs, just it's not quite an infinite money glitch because.

• 1:03:17Speaker 00

  Well, infinite is big, so no, not infinite, but let's just say you could, you know, do many, many orders magnitude of Earth's kind of current economy, like a million.

• 1:03:30Speaker 01

  Yeah.

• 1:03:30Speaker 00

  You know, it's this way, so if you, you know, just to get to like, that's why I think like, just to get to a millionth, a harnessing length of the sun's energy would be roughly, give or take an order of magnitude, a hundred thousand, a hundred thousand times bigger than Earth's entire economy today. And you're only at one millionth of the sun. Give him a tag.

• 1:03:55Dwarkesh Patel?

  Before we move on to this, I have a lot of questions on that, but.

• 1:03:59Speaker 00

  Every time I say order of magnitude, you're saying.

• 1:04:02Dwarkesh Patel?

  You're adding ten degrees.

• 1:04:03Speaker 00

  Take a shot every time I say that too often.

• 1:04:06Dwarkesh Patel?

  They tend the next time, a hundredth of that time after that.

• 1:04:08Speaker 00

  Yeah, order of magnitude more wasted.

• 1:04:11Dwarkesh Patel?

  I do have one more question about XAI. This strategy of building a digital or a remote worker, coworker replacement.

• 1:04:19Speaker 00

  Everyone's gonna do, by the way, not just us.

• 1:04:21Dwarkesh Patel?

  So what is XAI's plan to win?

• 1:04:23Speaker 00

  Do you expect me to tell you on a podcast?

• 1:04:25Dwarkesh Patel?

  Yeah.

• 1:04:27Speaker 00

  Spill all the beans, have another Guinness.

• 1:04:30Speaker 01

  It's a good system.

• 1:04:33Speaker 00

  We'll sing like a canary.

• 1:04:36Speaker 00

  All the secrets.

• 1:04:37Speaker 01

  Okay, but in a non-secret spilling way, what's the plan?

• 1:04:41Dwarkesh Patel?

  What a hack.

• 1:04:43Speaker 00

  Well, when you put it that way.

• 1:04:48Speaker 00

  I think the way that Tesla's self-driving is the way to do it. So I'm pretty sure that's the way.

• 1:04:58Dwarkesh Patel?

  Unrelated question. How did Tesla solve self-driving?

• 1:05:03Dwarkesh Patel?

  Yeah. It sounds like you're talking about data, like Tesla's self-driving because of the.

• 1:05:08Speaker 00

  We're gonna try data and we're gonna try algorithms.

• 1:05:11Dwarkesh Patel?

  But isn't that what all the other labs are trying?

• 1:05:16Speaker 00

  If those don't work, I'm not sure what will.

• 1:05:19Speaker 00

  We've tried data, we've tried algorithms.

• 1:05:27Speaker 00

  No, we don't know what to do. I'm pretty sure I know the path and it's just a question of how quickly we go down that path.

• 1:05:35Speaker 00

  Because it's pretty much the Tesla path. So, I mean, have you tried self-driving, Tesla's self-driving lately?

• 1:05:43Speaker 01

  Not the most recent version, but.

• 1:05:44Speaker 00

  Okay, the car is like, it just increasingly feels sentient, like it just, it feels like a living creature.

• 1:05:52Speaker 00

  And that'll only get more so. And I'm actually thinking like, we probably shouldn't put too much intelligence into the car because it might get bored. Start roaming the streets. I mean, imagine you're stuck in a car and that's all you could do.

• 1:06:09Speaker 00

  You know who put Einstein in a car? It's like, why am I stuck in a car? So there's actually probably limited to how much intelligence you put in a car to not have the intelligence be bored.

• 1:06:19Dwarkesh Patel?

  What's XAI's plan to stay on the compute ramp up that all the labs are doing right now? The labs are on track to spend over like 50 to 100 billion dollars.

• 1:06:26Speaker 00

  You mean the corporations?

• 1:06:28Dwarkesh Patel?

  Sorry, sorry, sorry, yeah. Corporations.

• 1:06:30Speaker 00

  The labs are at universities and they're really like a snail.

• 1:06:34Dwarkesh Patel?

  They're not starting at 50 billion dollars.

• 1:06:35Speaker 00

  You mean the revenue maximizing corporations?

• 1:06:39Dwarkesh Patel?

  That's right. The revenue maximizing corporations.

• 1:06:40Speaker 00

  That call themselves labs.

• 1:06:42Dwarkesh Patel?

  Are making like 20 to 10 billion, depending, like OpenAI is making 20B revenue, Anthropic is like 10B.

• 1:06:48Speaker 00

  Close to maximum profit AI.

• 1:06:50Dwarkesh Patel?

  XAI is reportedly like 1B. Like, what's the plan to get to their compute level, get to their revenue level and stay there as things go?

• 1:06:58Speaker 00

  Yeah, so as soon as you unlock Digital Human, you basically have access to trillions of dollars of revenue. So

• 1:07:10Speaker 00

  in fact, you can really think of it like the most valuable companies currently by market cap, their output is digital. So Nvidia's output is FTPing files to Taiwan.

• 1:07:29Speaker 00

  It's digital.

• 1:07:29Dwarkesh Patel?

  Right.

• 1:07:30Speaker 00

  Now, those are very, very difficult.

• 1:07:32Dwarkesh Patel?

  Yeah, high value files.

• 1:07:33Speaker 00

  They're the only ones that can make files that good.

• 1:07:37Speaker 00

  But that is literally their output, the FTP files to Taiwan.

• 1:07:40Dwarkesh Patel?

  Do they FTP them?

• 1:07:41Speaker 00

  I believe so.

• 1:07:43Speaker 00

  I believe that is the

• 1:07:44Dwarkesh Patel?

  SFTP.

• 1:07:45Speaker 00

  File transfer protocol, I believe, is, but I could be wrong. But either way, it's a bit stream going to Taiwan.

• 1:07:54Speaker 00

  You know, Apple doesn't make phones. They send files to China.

• 1:08:01Speaker 00

  Microsoft doesn't manufacture anything. Even for Xbox, that's outsourced. They, again, their output is digital. Meta's output is digital. Google's output is digital. So if you have a human emulator, you can basically create one of the most valuable companies in the world overnight. And you would have access to $20s revenue. It's not like a small amount.

• 1:08:28Dwarkesh Patel?

  All right, so you're saying basically like revenue figures today are just like so, like they're all rounding errors compared to the actual TAM. So just like focus on the TAM and how to get there.

• 1:08:37Speaker 00

  I mean, if you take something as simple as, say, customer service, if you have to integrate with the APIs of existing corporations, many of which don't even have an API. So you've got to make one. And you've got to wade through legacy software. That's extremely slow. However, if AI can simply take whatever is given to the outsourced customer service company that they already use and do customer service using the apps that they already use, then you can make tremendous headway in customer service, which is, I think, 1% of the world economy, something like that.

• 1:09:19Speaker 00

  It's close to a trillion dollars all in for customer service. And there's no barriers to entry. You can just immediately say, well, we'll outsource it for a fraction of the cost. And there's no integration needed.

• 1:09:31Speaker 01

  You can imagine some kind of categorization of intelligence tasks where there is breadth, where customer service is done by very many people, but, you know, many people can do it. And then there's difficulty where, you know, there's a best-in-class turbine engine. Like presumably there's a 10% more fuel-efficient turbine engine that could be imagined by an intelligence, but we just haven't found it yet. Or, you know, GLP-1s are just, you know, a few bytes of data. Where do you think you want to play in this? Is this a lot of, you know, reasonably intelligent intelligence? Or is this the very pinnacle of? Cognitive tasks.

• 1:10:10Speaker 00

  Well, I was just using customer service as like something that's it's a it's a very significant revenue stream, but one that is probably not super difficult to solve for. So if you if you can emulate a human at a at a desktop, that that's just literally what customer service is. And, you know, it's people of average intelligence, not like, you know, you don't need like somebody who's who's spent many, you know, many years, you don't need like, you know,

• 1:10:42Speaker 00

  sort of several sigma good engineers for that.

• 1:10:46Speaker 00

  But but obviously, as you make that work, you can then, once you have computers working effectively, digital optimists working, you can then run any application. Like, let's say you're trying to design chips, so you could you could then

• 1:11:02Speaker 00

  run your conventional apps, you know, like the stuff from cadence and synopsis and whatnot. And you can say, you can you can run a thousand simultaneously or 10,000 and say, okay, given this input, I get this output for the chip. And and at a certain point, you can say, okay, I you you're actually going to know what the what the chip should look like without using any of the tools. So basically, you you should be able to do a digital chip design, like you can do chip design, like you you watch up the difficulty curve.

• 1:11:41Speaker 00

  You could you you know, be able to do to do CAD. So, you know,

• 1:11:48Speaker 00

  you could use like sort of NX or any of the CAD software to design things.

• 1:11:53Speaker 01

  Okay, so you think you started the simplest tasks and walk your way up the difficulty curve?

• 1:11:58Dwarkesh Patel?

  So you're saying, look, as a broader objective of having this full digital coworker emulator, you're saying, look, all the revenue maximizing corporations want to do this, XAI being one of them, but we will win because of a secret plan we have. But like, everybody's like trying different things with data, different things with algorithms. And I'm like, hmm.

• 1:12:20Speaker 00

  I like that.

• 1:12:20Dwarkesh Patel?

  Like, what is this?

• 1:12:21Speaker 00

  We're trying data, we're trying algorithms.

• 1:12:25Speaker 00

  What else can we do?

• 1:12:28Dwarkesh Patel?

  Yeah. It seems like a competitive field, and I'm like, what is, how are you guys gonna win? Is like my, my big question.

• 1:12:33Speaker 00

  You know, I, I think we see a path to doing it. I mean, I think, I think I know, I think I know the path to do this because it's, it's kind of the same path that Tesla used to create self-driving. You know, instead of driving a car, it's driving a, a computer screen. So it's a self-driving computer, essentially.

• 1:12:55Speaker 01

  Oh, you're saying, is the path just following human behavior and training on vast quantities of human behavior?

• 1:13:03Dwarkesh Patel?

  But, but sorry, isn't that, I mean, isn't that, is that a training?

• 1:13:07Speaker 00

  I mean, obviously, I'm not gonna spell out, you know, most sensitive secrets on a podcast. You know, I need to have at least three more Guinnesses for that.

• 1:13:14Dwarkesh Patel?

  I've got some friends at Jane Street, and they're always talking about how their colleagues are cooking up fun fiendish puzzles for each other to solve. Well, last week, they sent me one. Basically, they trained a neural network, and they gave me the weights of each layer. But they didn't tell me what order those layers went in. And so I had to figure out the correct order using the outputs of the original network. And as soon as I got this puzzle, I went to my roommate, who's an AI researcher, and we both got immediately nerd-sniped. Obviously, you can't brute force a solution. The search space here is 10 to the 122 permutations. So clearly, you need some way to reduce the search space. Then my roommate had to go to work, but because I'm a podcaster, I had some time to take a stab at some of the ideas we discussed. And with the combination of simulated annealing and greedy search,

• 1:13:59Dwarkesh Patel?

  I think I got pretty close. I think I'm actually just a couple of swaps and shifts away from the correct solution. But what makes this puzzle really tricky is that there's no obvious way to escape from a local minimum. I'm afraid that this is as far as vibe coding is gonna get me, but maybe you can do better. Check out the puzzle at janestreet.com/thoarkesh. All right, back to Elon.

• 1:14:24Speaker 01

  What will XAI's business be like? Is it going to be consumer enterprise? What's the mix of those things going to be? Is it just going to be similar to other labs where you have this?

• 1:14:36Dwarkesh Patel?

  You're saying labs?

• 1:14:36Speaker 01

  Mix of corporations.

• 1:14:38Dwarkesh Patel?

  The SIAP goes Devilon.

• 1:14:40Speaker 01

  Revenue maximizing corporations, to be clear. Those GPUs don't pay for themselves. Exactly. But yeah, what's the business model? What are the revenue streams in a few years' time?

• 1:14:54Speaker 00

  Things are going to change very rapidly. Like, I'm stating the obvious here. You know, I call AI the supersonic tsunami. I love alliteration. So really, what's going to happen is, especially when you have humanoid robots at scale, is they will just provide, they'll make products and provide services far more efficiently than human corporations. So amplifying the productivity of human corporations is simply a short-term thing.

• 1:15:27Dwarkesh Patel?

  So you're expecting fully digital corporations? Rather than like SpaceX becomes part AI and so?

• 1:15:34Speaker 00

  I think there'll be digital corporations, but it's, like, some of this is going to sound kind of dumberish. Okay, but I'm just, I'm just saying what I think will happen. It's not, it's not meant to be dumberish or anything else. Just, just like, this is what I think will happen. Is that, is that pure AI, corporations that are purely AI and robotics will vastly outperform any corporations that have people in the loop. So you can, you can think of, say, like, like, computer used to be a job that humans had. You would go and get a job as computer where you would do calculations.

• 1:16:19Speaker 00

  And they'd have like entire skyscrapers full of humans, like, you know, 20, 30 floors of humans just doing calculations. Now, that entire skyscraper of humans doing calculations can be replaced by a laptop with a spreadsheet. That spreadsheet can do vastly more calculations than an entire building full of human computers.

• 1:16:49Speaker 00

  So you can think about, okay, well, what if only some of the cells in your some of the cells in your spreadsheet were calculated by humans? Actually, that would be much worse than if all of the cells in your spreadsheet were calculated by the computer. And so really, what will happen is

• 1:17:09Speaker 00

  the pure AI, pure robotics

• 1:17:13Speaker 00

  corporations or collectives will far outperform any corporations that have humans in the loop. And this will happen very quickly.

• 1:17:21Dwarkesh Patel?

  Speaking of closing the loop, sorry, Optimus.

• 1:17:26Dwarkesh Patel?

  You, I mean, as far as like manufacturing targets and so forth go, your companies have sort of been like carrying American manufacturing of hard tech on their back. But in the fields that you are, you know, Tesla has been dominant in, you're and now you want to go into humanoids. In China, there's entire dozens and dozens of companies that are doing this kind of manufacturing cheaply and at scale and are incredibly competitive. So give us sort of like advice or a plan of how America can build the humanoid armies or, you know, the EVs, et cetera, at scale and as cheaply as China is on track to.

• 1:18:10Speaker 00

  Well, there are really only three hard things for humanoid robots. The real world intelligence,

• 1:18:19Speaker 00

  the hand and scale manufacturing.

• 1:18:21Dwarkesh Patel?

  Yeah.

• 1:18:24Speaker 00

  So I haven't seen any even demo robots that have a great hand, like with all the degrees of freedom of a human hand. But Optimus will have that.

• 1:18:39Speaker 00

  Optimus does have that.

• 1:18:41Dwarkesh Patel?

  And how do you achieve that? Is it just like right torque density in the motor? Like, what is the hardware bottleneck to that?

• 1:18:46Speaker 00

  Well, we have to, we have to design custom actuators. Basically, custom designed motors, gears, electronics, controls, sensors. Everything had to be designed from physics first principles. There is no supply chain for this.

• 1:19:03Dwarkesh Patel?

  And will you be able to manufacture those at scale?

• 1:19:05Speaker 00

  Yes.

• 1:19:06Speaker 01

  Is anything hard except the hand from a manipulation point of view? Or once you've solved the hand, are you good?

• 1:19:12Speaker 00

  From an electromechanical standpoint, the hand is more difficult than everything else combined. Yeah, human hand turns out to be quite something. But you also need the real world intelligence. So the intelligence that Tesla has developed for the car applies very well to the robot,

• 1:19:32Speaker 00

  which is primarily vision in. But the car takes in more vision, but it actually also is listening for sirens. It's taking in the inertial measurements, its GPS signals, a whole bunch of other data. Combining that with video, it's primarily video, and then outputting the control commands. So like Tesla is taking in one and a half gigabytes a second of video and outputting two kilobytes a second of control outputs with the video at 36 Hz and the control frequency at 18.

• 1:20:07Speaker 01

  One intuition you could have for when we get this robotic stuff is that it takes quite a few years to go from the compelling demo to actually being able to use in the real world. So 10 years ago, you had really compelling demos of self-driving. But only now we have robotaxi and Waymo and all these services scaling up. Doesn't this, shouldn't this make one pessimistic on, say, household robots? Because we don't even quite have the compelling demos yet of, say, the really advanced hand.

• 1:20:39Speaker 00

  Well, we've been working on humanoid robots now for a while.

• 1:20:44Speaker 00

  So I guess it's been five or six years or something like that.

• 1:20:49Speaker 00

  And a bunch of things that we've done for the car are applicable to the robot. So we'll use the same Tesla AI chips in the in the robot as the car. We'll use the same basic principles. It's very much the same AI. You've got, you know, many more degrees of freedom for a robot than you do for a car. But really, if you're just thinking of it as like a bit stream. AI is really mostly a compression and correlation of two bit streams.

• 1:21:22Speaker 00

  So for video, you've got to do a tremendous amount of compression. And you've got to do the compression just right. You've got to compress the like, ignore the things that don't matter. And like, you don't care about the details of the leaves on the tree on the side of the road. But you care a lot about the the road signs and the traffic lights and pedestrians. And even whether, you know, somebody in another car is looking at you or not looking at you. Like, some of these details matter a lot. So if it is a session, it's got to turn that. The car is going to turn that one and a half gigabytes a second ultimately into two kilobytes a second of control outputs. So many stages of compression.

• 1:22:06Speaker 00

  And you've got to get all those stages right and then correlate those to the correct control outputs. Robot has to do essentially the same thing. And you think about what humans, this is what happens with humans. We really are photons in, controls out. So that is the vast majority of your life has been vision, photons in, and then motor controls out.

• 1:22:27Dwarkesh Patel?

  Naively, it seems like between humanoid robots and cars, the fundamental actuators in a car are like, how you turn, how you accelerate, et cetera. Where in a robot, especially with maneuverable arms, there's dozens and dozens of these degrees of freedom. And then, especially with Tesla, you had this advantage of like, you had millions and millions of hours of human demo data collected from just the car being out there. Where like, you can't equivalently just deploy optimists that don't work and then get the data that way. So between the increased degrees of freedom and the far sparser data.

• 1:22:59Speaker 00

  Yes.

• 1:23:01Dwarkesh Patel?

  How will you use the sort of Tesla engine of intelligence to train the optimist mind?

• 1:23:10Speaker 00

  Now, you're actually, you're highlighting an important limitation and difference between cars. It's like, we do have, oh, we'll still have like 10 million cars on the road. And so that's, it's hard to duplicate that, like, massive training flywheel. For the robot,

• 1:23:31Speaker 00

  what we're gonna need to do is build a lot of robots and put them in kind of like an Optimus Academy. So they can do self-play in reality. So we're actually, we're actually pulling that out. So we're gonna have at least 10,000 Optimus robots, maybe 20 or 30,000 that can do, that are doing self-play and testing different tasks. And then Tesla has quite a good reality generator, like a physics accurate reality generator that we made, made this for the cars. We'll do the same thing for the robots.

• 1:24:07Speaker 00

  Actually have done that for the robots. So you have, you know, a few tens of thousands of humanoid robots doing different tasks. And then you've got, you can do millions of simulated robots in the simulated world. And you use the, the tens of thousands of robots in the real world to close the simulation to reality gap. Close the sim to real gap.

• 1:24:31Dwarkesh Patel?

  How do you think about the synergies between XAI and Optimus? Given you're highlighting, look, you need this world model. You maybe want to use some really smart intelligence as the control plane. And so maybe Grok is like doing the slower planning and then like the motor policy is at the lower level. Yeah, what will the sort of synergy between these things be?

• 1:24:52Speaker 00

  Yeah, so you'd use, Grok would orchestrate the behavior of the Optimus robots. So let's say you wanted to build a factory.

• 1:25:03Speaker 00

  Then Optimus, then Grok could organize the Optimus robots, give them, assign them tasks to build the factory for, to produce whatever you want.

• 1:25:16Speaker 01

  Don't you need to merge XAI and Tesla then? Because these things end up so...

• 1:25:19Speaker 00

  What were we saying earlier about public company discussions?

• 1:25:23Dwarkesh Patel?

  We're one more Guinness in, Elon.

• 1:25:27Dwarkesh Patel?

  What are you waiting to see before you say, we want to manufacture 100,000 Optimus? Is it like?

• 1:25:33Speaker 00

  Optimize. Since we're defining the proper noun, we can define the plural of the proper noun too. So we're going to proper noun the plural, and so it's optimize.

• 1:25:43Dwarkesh Patel?

  Okay. Is there something on the hardware side you want to see? Do you want to see better actuators? Or is it just you want the software to be better? What are we waiting for before we get like mass manufacturing of Gen 3?

• 1:25:54Speaker 00

  No, we're moving towards that. We're going forward with mass manufacturing.

• 1:25:58Dwarkesh Patel?

  But you think current hardware is good enough that you are going, you should, you just want to deploy as many as possible now?

• 1:26:06Speaker 00

  I mean, it's very hard to scale up production. But yeah, I think Optimus 3 is the right version of the robot to,

• 1:26:15Speaker 00

  you know, to produce maybe something on the order of like a million units a year. I think you'd want to go to Optimus 4 before you went to 10 million units a year.

• 1:26:23Speaker 01

  Okay, but you can do a million a year at Optimus 3?

• 1:26:25Speaker 00

  Yeah, I mean, it's very hard to spoil at manufacturing.

• 1:26:28Speaker 01

  Yes.

• 1:26:30Speaker 00

  So like manufacturing,

• 1:26:33Speaker 00

  like the output per unit time is always follows an S curve. It starts off agonizingly slow, then has this sort of

• 1:26:42Speaker 01

  We eventually

• 1:26:43Speaker 00

  exponential increase, then linear, then a, you know, logarithmic outcome till you sort of eventually asymptote at some number. But Optimus initial production will be, it's going to be a, it's going to be a stretched out S curve because so much of what goes into Optimus is brand new. There's not an existing supply chain.

• 1:27:02Speaker 00

  As I mentioned, the actuators, electronics, everything in the Optimus robot is designed

• 1:27:08Speaker 00

  for physics first principles. It's not, it's not taken from a catalog. These, these are custom designed, everything, literally everything. I don't think there's a single thing that

• 1:27:17Speaker 01

  How far down does that go?

• 1:27:20Speaker 00

  I mean, I guess we're not making custom capacitors yet, maybe, but

• 1:27:27Speaker 00

  There's nothing you can pick out of a catalog at any price. So it just means that the Optimus S-Cove, the units per, output per unit time, how many Optimus robots you make per day, whatever, is gonna initially ramp slower than a product where you have an existing supply chain. But it will get to a million.

• 1:27:56Dwarkesh Patel?

  When you see these Chinese humanoids like Unitree or whatever, sell humanoids for like 6K or 13K, do you just like, are you hoping to get your Optimus's bill of materials below that price? So you can do the same thing? Or do you just think qualitatively, they're not the same thing? Like, what do you think is going, like, what allows it? What allows them to sell for so low? And can we match that?

• 1:28:18Speaker 00

  Well, Optimus, our Optimus is designed to have a lot of intelligence and to have the same electromagnetic angle dexterity, if not higher than a human. So Unitree does not have that. And it's also, I mean, it's quite a, it's quite a big robot. It's, it has to do, you know, carry heavy objects for long periods of time and not overheat or exceed the power of its actuator. So we've got, you know, it's 511, you know, so it's pretty tall. And it's got a lot of intelligence. So it's gonna be more expensive than a small robot that is not intelligent.

• 1:29:02Speaker 01

  But more capable?

• 1:29:03Speaker 00

  Yeah, but not a lot more. I mean, like, the thing is, over time, as Optimus robots build Optimus robots, the cost will drop very quickly.

• 1:29:12Speaker 01

  And what will these first billion Optimuses, Optimae?

• 1:29:15Speaker 00

  Yeah.

• 1:29:15Speaker 01

  Do? Like, what will their highest and best use be?

• 1:29:19Speaker 00

  I think you would start off with, with simple tasks that you can count on them doing well.

• 1:29:23Speaker 01

  But in the home or in factories, like?

• 1:29:25Speaker 00

  The best use for robots in the beginning won't be anything, any continuous operations, any 24 by 7 operation, because then you're, because they can, they can work continuously.

• 1:29:36Dwarkesh Patel?

  Yeah. What fraction of the work at a Giga factory that is currently done by humans could Gen 3 do?

• 1:29:42Speaker 00

  I'm not, I'm not sure. Maybe it's like 10, 20%.

• 1:29:46Speaker 00

  Maybe more, I don't know, that's it. We would, we would use, we would not like reduce our headcount, we would, we would.

• 1:29:53Speaker 01

  For sure.

• 1:29:53Speaker 00

  Increase our headcount, to be clear.

• 1:29:55Speaker 01

  Right.

• 1:29:56Speaker 00

  But we would increase our output. So the, the units produced per human, like total, total number of humans at Tesla will increase, but the, the output of robots and cars will increase disproportionate, like much, much to,

• 1:30:15Speaker 00

  you know, number of cars and robots produced per human will increase dramatically, but number of humans will increase as well.

• 1:30:23Speaker 01

  We're talking about Chinese manufacturing a bunch here. And we're also talking about, you know, we've talked about some of the policies that are relevant, like you mentioned, the, the solar tariffs.

• 1:30:36Speaker 00

  Yeah.

• 1:30:37Speaker 01

  And you think they're a bad idea because, you know, we can't scale up solar in the US.

• 1:30:41Speaker 00

  Well, just electricity output in the US needs to scale up.

• 1:30:45Speaker 01

  Right, and can't without like good power sources.

• 1:30:48Speaker 00

  You just need to get it somehow.

• 1:30:49Speaker 01

  Yeah, but where I was going with this is if you were in charge, if you were setting all the policies, what else would you change?

• 1:30:58Speaker 01

  So you'd change the solar tariffs as well?

• 1:31:00Speaker 00

  Yeah, I would say anything that is a limiting factor for electricity basically addressed, provided it's not like very bad for the environment.

• 1:31:08Speaker 01

  So presumably some permitting reforms and stuff as well will be in there.

• 1:31:11Speaker 00

  Yeah, there's a fair bit of permitting reforms that are happening. A lot of the permitting is state-based, so. But anything better, but this administration is, is good at removing permitting roadblocks.

• 1:31:25Speaker 00

  And I'm not saying all tariffs are bad, I'm just saying.

• 1:31:27Speaker 01

  The solar tariffs, yeah.

• 1:31:28Speaker 00

  So, yeah, yeah, I mean, sometimes if, like, if another country is subsidizing the output of something, then, then you have to have countervailing tariffs to protect domestic industry against subsidies by another country.

• 1:31:41Speaker 01

  What else would you change?

• 1:31:43Speaker 00

  I don't know if there's that much that the government can actually do.

• 1:31:46Speaker 01

  Well, one thing I was wondering is, it seems like the, for the. Policy goal of creating a lead for the US versus China. It seems like the export bans have actually been quite impactful. Where China is not producing leading edge chips and the export bans really bite there. China is not producing leading edge turbine engines. And similarly, there's a bunch of export bans that are relevant there on some of the metallurgy. Should there be more export bans? Like, if you think about things like, I mean, there are now the drone industry and things like that. But is that something that should be considered?

• 1:32:24Speaker 00

  Well, I think it's important to appreciate that in most areas, China is very advanced in manufacturing. There's only a few areas where it is not.

• 1:32:34Speaker 00

  The, you know, China is a manufacturing powerhouse next level. Like, people don't, most people don't.

• 1:32:41Speaker 01

  It's very impressive.

• 1:32:41Speaker 00

  Yeah, yeah. I mean, if you, if you take like refining of, of ore.

• 1:32:48Speaker 00

  I'd say roughly China, there's more, there's twice as much ore refining of, of, on average, as the rest of the world combined.

• 1:32:59Speaker 00

  And I think there's, there's some areas, like you say, refining gallium, which goes into solar cells.

• 1:33:05Speaker 00

  I think they're at like 98% of gallium refining. So, so China is actually very advanced in manufacturing in, I'd say, most areas.

• 1:33:13Speaker 01

  It seems like we're, like, there is discomfort with this supply chain dependence. And yes, nothing's really happening on it.

• 1:33:20Speaker 00

  Supply chain, which supply chain.

• 1:33:21Speaker 01

  It depends on, say, like, the gallium refining that you're saying.

• 1:33:24Speaker 00

  Yeah, yeah, there's, there's, there's a.

• 1:33:27Speaker 01

  Well, the rare earth stuff and.

• 1:33:29Speaker 00

  Yeah, rare earths, which are, as you know, not rare.

• 1:33:32Speaker 01

  Yeah.

• 1:33:33Speaker 00

  Like, we actually do rare earth ore mining in the US. Send the, the, the rock, we put it on, on a, on a train. And then put it on a boat to China, that goes another train and goes to the rare earth refining, refiners in, in China, who then refine it, put it into a magnet, put it into a motor service assembly, and then send it back to America. So the thing, we're really missing a lot of, of ore refining in, in America and.

• 1:34:00Speaker 01

  Isn't this worth a policy intervention?

• 1:34:02Speaker 00

  Yes, well, I think there are some things being done on, on that front. But, but.

• 1:34:11Speaker 00

  We kind of need optimists, frankly, to build oil refineries.

• 1:34:16Dwarkesh Patel?

  So, sir, you think the main advantage that China has is the abundance of skilled labor? And that's like, that's the thing Optimus fixes?

• 1:34:24Speaker 01

  But they also, we need the skilled labor.

• 1:34:24Speaker 00

  China's got like four times our population, so.

• 1:34:26Dwarkesh Patel?

  But we need, so, I mean, there's this concern, if you think like humanoids are the future, that like, okay, right now, if it's the skilled labor for manufacturing that's determining who's, who can build more humanoids, you know, China has more of those, it manufactures more humanoids, therefore it gets, it gets the OptiMy future first.

• 1:34:44Speaker 00

  Well, we'll see.

• 1:34:45Dwarkesh Patel?

  And it just like keeps that special going. It seems like you're sort of pointing out that sort of getting to a million OptiMy.

• 1:34:51Speaker 00

  Yeah.

• 1:34:51Dwarkesh Patel?

  Requires the manufacturing that the OptiMy is supposed to help us get to, right?

• 1:34:56Speaker 00

  You can, you can close that recursive loop pretty quickly.

• 1:34:59Speaker 01

  With a small number of OptiMy.

• 1:35:01Speaker 00

  Yeah. So you close the recursive loop to help, to help the robots build the robots. And then we can, you know, try to get to tens of millions of units a year. Maybe if you start getting to hundreds of millions of units a year, I think you're, you're going to be the most competitive country by far. We definitely can't win with just humans because China has four times our population.

• 1:35:22Dwarkesh Patel?

  Right.

• 1:35:22Speaker 00

  And frankly, America's been running for so long that we, you know, just like a, like a pro sports team that's been running for a very long time, tend to get complacent and entitled. And that's why they stopped winning, because it's, you know, don't work as hard anymore. So I think the, frankly, just my observation is the average work ethic in China is higher than in the US. So it's not just that there's four times the population, but the work, the amount of work that people put in is higher. So you can like, you can try to rearrange the humans, but you're still one quarter of the, you know, assuming that the productivity is, is the same, which I think actually it might not be.

• 1:36:01Dwarkesh Patel?

  Yeah.

• 1:36:02Speaker 00

  I think China might have an advantage on productivity, for a person.

• 1:36:06Speaker 00

  We will do one quarter of the amount of things as China. So we can't win on the human front. And our birth rate's been low for a long time, so our birth rate's been, the US birth rate's been below replacement since roughly 1971.

• 1:36:25Speaker 00

  So we've got a lot of people retiring or more people dying than than than we're close to sort of more people domestically dying than than being born. So we definitely can't win on the human front, but we we might have a shot at the robot front.

• 1:36:41Speaker 01

  Are there other things that you have wanted to manufacture in the past, but they've been too labor intensive or too expensive that now you can come back to and say, oh, we can finally do the whatever because we have Optimus?

• 1:36:53Speaker 00

  Yeah, I think we'd like to do more build more ore refineries at Tesla. So we just completed construction and have begun lithium refining without lithium refinery in Corpus Christi, Texas. We have a nickel refinery, which is for the cathode that's here in Austin. And these are these are the largest, this is the largest cathode, this is the largest cathode refinery, largest lithium refinery, largest nickel and lithium refinery outside of China. And it's like the, you know, the cathode team would say like, we have the the largest

• 1:37:36Speaker 00

  and the only actually cathode refinery in America.

• 1:37:39Speaker 01

  Many superlatives.

• 1:37:40Speaker 00

  Not just the largest, but it's also the only. So it was pretty big, even though it's the only one. But I mean, there are other things that, you know, you could do a lot more refineries and and help the help America be more competitive on refining capacity. So there's like, there's basically a lot of work for for the optimist to do that that most Americans, very few Americans frankly, want to do. I mean, I've I've actually.

• 1:38:12Speaker 01

  Is the refining work too dirty or what's the?

• 1:38:15Speaker 00

  It's not, it's actually, no, we don't, there's not, we don't have toxic emissions from the refinery or anything.

• 1:38:22Speaker 00

  The cathode nickel refinery is right, right? Sort of in Travis County, like five minutes from. Why can't you do it with humans? No, you can't. You run out of humans. Ah, I see. Okay, yeah. Like, no matter what you do, you have one quarter of the number of humans in America and China. So if you have them do this thing, they can't do the other thing. So then, well, how do you build this refining capacity? Well, you can do it with Optima. And not very many Americans are pining to do refining.

• 1:38:55Speaker 00

  I mean, how many of you are on it too?

• 1:38:58Speaker 00

  Not a few. What are you? What are you planning to refine?

• 1:39:02Dwarkesh Patel?

  You know, BYD is reaching Tesla production or sales in quantity. What do you think happens in global markets as Chinese production in EVs goes up?

• 1:39:14Speaker 00

  Well, China is extremely competitive in manufacturing. So I think there's going to be a massive flood of Chinese vehicles and other, basically, most manufactured things. I mean, as it is, as I said, China is probably just twice as much refining as the rest of the world combined. So if you go, you know, if you just go down to like fourth and first tier supply chain stuff, like in the base, so we've got energy, then you've got mining and refining.

• 1:39:55Speaker 00

  Those foundation layers are, like I said, as a rough guess, China is doing twice as much refining as the rest of the world combined. So any given thing is going to have Chinese content because China is doing twice as much manufacturing refining work as the rest of the world.

• 1:40:16Speaker 00

  And then they'll go all the way to the finished product with the cars.

• 1:40:22Speaker 00

  I mean, China is a powerhouse. I mean, I think this year China will exceed three times US electricity output. And electricity output is a reasonable proxy for, you know, for the economy.

• 1:40:37Speaker 00

  So like in order to run the factories and run everything, you need electricity. So electricity is a good proxy for the real economy. And so if China is, if China passes three times the US electricity output, it means its industrial capacity, that's a rough approximation, is three times that, will be three times that of the US.

• 1:41:01Dwarkesh Patel?

  Reading between the lines, it sounds like what you're sort of saying is, absence of sort of humanoid recursive miracle in the next few years, on the sort of like whole of manufacturing, energy, raw materials chain, like China will just dominate, whether it comes to like AI or manufacturing EVs or manufacturing humanoids.

• 1:41:22Speaker 00

  In the absence of breakthrough innovations in the US, China will utterly dominate.

• 1:41:35Dwarkesh Patel?

  Interesting.

• 1:41:35Speaker 00

  Yes.

• 1:41:36Speaker 01

  We're about expecting the main breakthrough innovation.

• 1:41:39Speaker 00

  Well, if you do, like to scale AI in space, like basically you need, you need the humanoid robots, you need real world AI, you need a million tons a year to orbit. Like, let's just say, like, if we get the mass driver on the moon going, my favorite thing, then I think.

• 1:42:03Speaker 01

  It will have solved all our problems.

• 1:42:05Speaker 00

  Yeah, so this is like, I call that winning. I call that winning.

• 1:42:13Speaker 01

  You can finally be satisfied, you've done something.

• 1:42:15Speaker 00

  Yes.

• 1:42:16Speaker 01

  You have the mass driver on the moon.

• 1:42:18Speaker 00

  That's right, I just want to see that thing in operation.

• 1:42:19Speaker 01

  Was that out of some sci fi or where did you?

• 1:42:22Speaker 00

  Well, actually, there is a Highland book, The Moon is a Horseshoe.

• 1:42:25Speaker 01

  That's a great. Okay, yeah, but that's slightly different, that's a gravity slingshot or?

• 1:42:30Speaker 00

  No, they have a mass driver on the moon.

• 1:42:31Speaker 01

  Okay, yeah, but they use that to attack Earth, so maybe it's something.

• 1:42:35Speaker 00

  Well, they just set to, they're independence.

• 1:42:38Speaker 01

  Exactly, what are your plans for the mass driver on the moon?

• 1:42:40Speaker 00

  They set up their independence, Earth government disagreed and they loved things until Earth government agreed.

• 1:42:46Speaker 01

  That book is a hoot, I found that book much better than his other one that everyone reads, Stranger in a Strange Land.

• 1:42:52Speaker 00

  Yeah, Grok comes from Stranger in a Strange Land.

• 1:42:54Speaker 01

  Yeah, yeah, but I much preferred.

• 1:42:55Speaker 00

  Yeah, the first two thirds of Stranger in a Strange Land are good and then it gets very weird in the third portion.

• 1:43:01Speaker 01

  Yeah.

• 1:43:02Speaker 00

  But there's still some good concepts in there.

• 1:43:04Speaker 01

  Yeah.

• 1:43:05Dwarkesh Patel?

  Labelbox can get you robotics and RL data at scale. Take robotics. Let's say you need 100,000 hours of egocentric video. Labelbox starts by helping you define your ideal data distribution. Like, for example, maybe no single task category should occupy more than 1% of trading volume, and at least 10% of trajectories should capture failure and recovery states. Next, Labelbox assigns its distribution to its massive network of operators. You're not limited to the small range of scenes that you can set up in a single warehouse. Instead, each one of Labelbox's operators has access to lots of unique physical environments where they can film themselves completing a wide variety of tasks. Labelbox's tech automatically categorizes each video so that their operators always

• 1:43:47Dwarkesh Patel?

  know which tasks still remain and what they need to work on next. For RL data, Labelbox takes a similar approach. They work with you to understand the right distribution of tasks, and then their subject matter experts build the hyper-realistic digital environments and rubrics that you need to collect the highest quality trading data. So whether you're training robots in the real world or agents for computer use, Labelbox can help. Go to labelbox.com/sparcash to learn more.

• 1:44:17Speaker 01

  One thing we were discussing a lot is kind of your system for managing people. Like, you interviewed the first few thousand of SpaceX employees and lots of other companies.

• 1:44:27Speaker 00

  Well, it doesn't scale.

• 1:44:29Speaker 01

  Well, yes, but what doesn't scale?

• 1:44:31Speaker 00

  Me.

• 1:44:32Speaker 01

  Sure, sure. I know that, but like, what are you looking for?

• 1:44:36Speaker 00

  Literally, it's not enough hours in a day, it's impossible.

• 1:44:38Speaker 01

  But what are you looking for that someone else who's good at interviewing and hiring people, what's the je ne sais quoi?

• 1:44:46Speaker 00

  Well, at this point, I think I've got, I might have more training data on evaluating technical talent especially, but talented all kinds, I suppose, but technical talent especially, given that I've done so many technical interviews and then seen the results, technical interviews, seen the results. So my, my training set is, is very, is enormous and has a very wide range.

• 1:45:11Speaker 00

  Generally, the thing I ask for are bullet points for evidence of exceptional ability. So it's like, it's, it's, and these things can be like, pretty off the wall. It doesn't need to be in the, in the domain, the specific domain. But evidence that, evidence of exceptional ability. So if somebody can like cite like even one thing, but they say three things where you go, wow, wow, wow. Then that's, that's a good sign.

• 1:45:38Dwarkesh Patel?

  But why do you have to be the one to determine that?

• 1:45:40Speaker 00

  No, I don't. I can't be. It's impossible.

• 1:45:42Dwarkesh Patel?

  Right.

• 1:45:43Speaker 00

  I mean, total headcount across the whole company is 200,000 people.

• 1:45:46Dwarkesh Patel?

  Right.

• 1:45:48Speaker 01

  But in the early days, what was it that you were looking for that couldn't be delegated in those interviews?

• 1:45:59Speaker 00

  Well, I guess I need to build my training set. It's not like I had about a thousand here. I would make mistakes.

• 1:46:05Speaker 01

  Yes.

• 1:46:05Speaker 00

  But then I'd be able to see where I thought somebody would work out well, but they didn't. And then why, why did they not work out well? And what can I do to, I guess, RL myself?

• 1:46:14Dwarkesh Patel?

  Yes.

• 1:46:15Speaker 00

  To, in the future, have a better batting average when interviewing people.

• 1:46:20Dwarkesh Patel?

  Mm hmm.

• 1:46:21Speaker 00

  So my batting average is still not perfect, but it's very high.

• 1:46:24Dwarkesh Patel?

  What are some surprising reasons people don't work out?

• 1:46:27Speaker 00

  Surprising reasons?

• 1:46:29Dwarkesh Patel?

  Like, they don't understand technical domain, et cetera, et cetera. But like, you're like, you've got like the long tail now of like, I was really excited about this person. It didn't work out. Curious why that happens.

• 1:46:41Speaker 00

  Yeah, so the, I mean, generally what I tell people, or tell myself, I guess, aspirationally, is don't look at the resume. Just believe, believe your interaction. So if the resume may seem very impressive and it's like, wow, your resume looks good. But if the, if the conversation after 20 minutes is, that conversation is not wow. You should believe the conversation, not the, not the, not the paper.

• 1:47:07Speaker 01

  I feel like. Part of your method is that, you know, there was this meme in the media a few years back about Tesla being a revolving door of executive talent. Whereas actually, I think when you look at it, Tesla's had a very consistent and internally promoted executive bench over the past few years. And then at SpaceX, you have all these folks like Marc Jancosa and Steve Davis and

• 1:47:29Speaker 00

  Steve Davis runs a sporting company.

• 1:47:30Speaker 01

  Yeah, yeah, but Bill Riley and folks like that. And it feels like part of has worked well is having very capable technical deputies. What do all of those people have in common?

• 1:47:47Speaker 00

  Well, so the, I mean, the Tesla sort of senior team, at this point, probably got average tenure of 10 or 12 years. It's quite long. Yeah. So, but there are times where Tesla went through extremely rapid, extremely rapid growth phase. And so it was somewhat, things were just somewhat sped up. And when a company, as you know, a company goes through different orders of magnitude of size, you know, people there who could help manage, say, a 50-person company versus a 500-person company versus a 5,000-person company versus a 50,000-person company.

• 1:48:29Speaker 01

  Yeah, you recruit people.

• 1:48:30Speaker 00

  Yeah, it's just not the same team. It's not always the same team. So if a company is growing very rapidly, the rate at which executive positions will change will also be proportionate to the rapidity of the growth, generally. Then Tesla had a further challenge where when Tesla had very successful periods, we would be relentlessly recruited from, like relentlessly. Like when Apple had their electric car program, they were carpet bombing Tesla with recruiting calls. It was, engineers just unplugged their phones.

• 1:49:09Speaker 01

  I'm trying to get work done here.

• 1:49:11Speaker 00

  Yeah, if I get, you know, one more call from an Apple recruiter. But they were, they were, they were opening offer without any interview with me, like double the compensation at Tesla. So.

• 1:49:24Speaker 00

  So we had a bit of the Tesla pixie dust thing where it's like, oh, if you hired a Tesla executive, you're suddenly you're gonna, everything's gonna be successful. And I've fallen prey to the pixie dust, you know, thing as well, where it's like, oh, we'll hire someone from Google or Apple and they'll be immediately successful, but not, that's not how it works.

• 1:49:45Speaker 00

  You know, people are people. It's, it's not like magical pixie dust.

• 1:49:47Speaker 01

  Yes.

• 1:49:47Speaker 00

  So when we had the pixie dust problem, we would get relentlessly recruited. And then also being Tesla being engineering, especially being primarily in Silicon Valley, it's easier for people to just, like, they don't have to change their life very much.

• 1:50:07Speaker 01

  Yes.

• 1:50:07Speaker 00

  They can just, you know, their commute's gonna be the same.

• 1:50:11Speaker 01

  Yes.

• 1:50:13Speaker 01

  So how do you prevent that? How do you prevent the pixie dust effect for everyone's trying to coach all your people?

• 1:50:20Speaker 00

  I don't think we can, I don't think there's much we can do to, to

• 1:50:23Speaker 01

  Yeah.

• 1:50:23Speaker 00

  stop it. But that's like, that's one of the reasons why Tesla, really being in Silicon Valley

• 1:50:32Speaker 00

  and, and having the pixie dust thing at the same time, meant that there was just a very, very aggressive recruitment.

• 1:50:42Speaker 01

  Only being in Austin helps, then?

• 1:50:44Speaker 00

  Austin, yeah, it still helps. I mean, Tesla still has a majority of its engineering in California. So the, you know, for getting engineers to move, I call it the significant, significant other problem.

• 1:50:59Speaker 01

  Yes. So then others have jobs. Yeah.

• 1:51:02Speaker 00

  Yeah, yeah, exactly. So for Starbase, that was particularly difficult.

• 1:51:06Speaker 01

  Yes.

• 1:51:07Speaker 00

  Since the odds of, you know, finding an aerospace extra job.

• 1:51:10Speaker 01

  Brownsville, Texas.

• 1:51:11Speaker 00

  Pretty low, yeah.

• 1:51:13Speaker 01

  Yeah.

• 1:51:13Speaker 00

  Yeah, it's quite, quite difficult. I mean, it's like a technology monastery, to think, you know, remote and mostly dudes.

• 1:51:22Speaker 01

  But again, if you go back.

• 1:51:24Speaker 00

  It's not much of an improvement over S F. Yeah.

• 1:51:28Speaker 01

  If you go, but if you go back to these people who've really.

• 1:51:34Speaker 01

  Been very effective in a technical capacity at Tesla, at SpaceX and those sorts of places. What do you think they have in common other than like, is it just that they're very sharp on the, you know, rocketry or the, you know, the technical foundations? Or do you think it's something organizational, it's something about their ability to work with you? Is this their ability to like be, you know, flexible but not too flexible?

• 1:52:03Speaker 01

  What makes a good sparring partner for you?

• 1:52:06Speaker 00

  I don't think of a sparring partner. I mean, if somebody gets things done, I, I love them and if they don't, I hate them. So it's pretty straightforward. It's not like some idiosyncratic thing. If somebody executes well, I'm a huge fan and if they don't, I'm not. But it's, it's not about mapping to my idiosyncratic preferences. I'll certainly try not to have it be mapping to my idiosyncratic preferences.

• 1:52:30Speaker 00

  So, yeah.

• 1:52:35Speaker 00

  Yeah, but generally I think it's a good idea to hire for talent and drive and trustworthiness.

• 1:52:46Speaker 00

  And I think goodness of heart is important. I'd have waited at that at one point. So like, are they, are they a good person, trustworthy, smart, talented and hardworking? If so, you can add domain knowledge. But those, those fundamental traits, those fundamental properties you cannot change. So most of the people who are at Tesla and SpaceX did not come from the aerospace industry or the order industry.

• 1:53:18Dwarkesh Patel?

  What is most had to change about your management style as your companies have scaled from 100, 1,000 to 10,000 people? You're, you know, you're known for this like very micromanagement, just getting into the details of things.

• 1:53:28Speaker 00

  Nano management, please. People management.

• 1:53:34Speaker 00

  Fanta management.

• 1:53:36Dwarkesh Patel?

  So you're saying?

• 1:53:39Speaker 00

  We're gonna go all the way down to flag sponsor.

• 1:53:44Speaker 00

  All the way down to Heisenberg's on Sunday personal.

• 1:53:48Speaker 01

  Yeah, well, how do you, I mean, are you still able to get into details as much as you want? Would your companies be more successful if they were smaller? Like, how do you think about that?

• 1:53:57Speaker 00

  Well, because I have a fixed amount of time in the day, my time is necessarily diluted as things grow and as the span of activity increases. So, you know, it's impossible for me to actually be a micromanager because that would imply I have some, like, thousands of hours per day. It is a logical impossibility to micromanage things.

• 1:54:26Speaker 00

  So now there are times when I will drill down into a specific issue because that specific issue is the limiting factor on the progress of the company.

• 1:54:40Speaker 00

  And but the reason for drilling into that, some very detailed item is because it is the limiting factor. It's not arbitrarily drilling into tiny things. And like I said, obviously, from a time standpoint, it is physically impossible for you arbitrarily going to tiny things that don't matter. And that would result in failure. But sometimes the tiny things are decisive in victory.

• 1:55:09Speaker 01

  Famously, you switched the Starship design from composites to steel.

• 1:55:16Speaker 00

  Yes.

• 1:55:17Speaker 01

  And you made that decision. Like, that wasn't, you know, people were going around. They're like, oh, we found something better, boss. Like, that was you encouraging people against some resistance. Can you tell us how you came to that whole composite steel switch?

• 1:55:30Speaker 00

  Yeah, so desperation, I'd say.

• 1:55:35Speaker 00

  The, originally, yeah, we were going to make Starship out of carbon fiber. And carbon fiber is pretty expensive. Like, the

• 1:55:51Speaker 00

  You know, you can generally, when you do volume production, you can get any given thing to be, to start to approach its material cost. The problem with, with carbon fibers is that material cost is still very high.

• 1:56:04Speaker 00

  So it's about, it's about 50 times, particularly if you go for a high strength specialized carbon fiber that can handle cryogenic oxygen, it's, it's like, call it roughly 50 times the cost of steel.

• 1:56:20Speaker 00

  And at least in theory, it would be lighter. People generally think of steel as being heavy and carbon fiber as being light. And for room temperature, room temperature applications, you know, like say more or less room temperature applications like a Formula One car, static aerostucture or any kind of aerostucture really is gonna, you're gonna probably be better off with the carbon fiber. Now, the problem is that we were trying to make this enormous rocket out of carbon fiber and our progress was extremely slow.

• 1:56:53Speaker 01

  And it's been picked in the first place just because it's light?

• 1:56:56Speaker 00

  Yes, like at first glance, like most people would think that, that the choice for making something light would be carbon fiber.

• 1:57:09Speaker 00

  The, now, now the thing is that

• 1:57:16Speaker 00

  when you make something very enormous out of carbon fiber and then you try to have the carbon fiber be efficiently cured, meaning not, not room temperature cured, because like you've got, you know, sometimes you got like 50 plies of, of, of carbon fiber and a carbon fiber is really carbon string and glue. And, and you, in order to have high strength, you need an autoclave. So something that, that can, that's essentially a high pressure oven. And if, if you have something that's a gigantic, that one's gotta be bigger than the rocket. So we're trying to make the autoclave that's bigger than any autoclave that's ever existed. Or do room temperature cure, which takes a long time and has issues.

• 1:58:02Speaker 00

  But, but the final issue is that we're just making very slow progress with, with carbon fiber.

• 1:58:10Speaker 00

  So.

• 1:58:12Dwarkesh Patel?

  I think the meta question is why it had to be you who made that decision. There's many engineers on your team. Yeah, how did the team not arrive at steel? Yeah, exactly. This is a part of a broader question like understanding your comparative advantage at your companies.

• 1:58:27Speaker 00

  So it was because we were making very slow progress with carbon fiber. I was like, OK, we've got to try something else. Now, for the Falcon 9, the primary airframe is made of aluminum lithium, which is a very, very good strength of weight. And actually, it has about the same, maybe better strength of weight for its application than carbon fiber. But aluminum lithium is very difficult to work with. In order to weld it, you have to do something called friction stir welding, where you join the metal without it entering the liquid phase. So it's kind of wild that you could do that. But with this particular type of welding, you can do that.

• 1:59:07Speaker 00

  But it's very difficult to like, say, let's say you want to make a modification or attach something to aluminum lithium. You now have to use mechanical attachment with seals. You can't weld it on. So I want to avoid using aluminum lithium for the primary structure for Starship. And there was this very special grade of carbon fiber that had very good mass properties. So with rockets, you're really trying to maximize the percentage of the rocket that is propellant, minimize the mass, obviously. And like I said, we were making very slow progress.

• 1:59:52Speaker 00

  And I said, at this rate, we're never going to get to Mars. So we better think of something else. I didn't want to use aluminum lithium because of the difficulty of friction stir welding, especially doing that at scale was hard enough at 3.6 meters in diameter, let alone at nine meters or above. Then.

• 2:00:15Speaker 00

  It says, well, what about steel? And so now I had a clue here because some of the early US rockets had used very thin steel. The Atlas rockets had used a steel balloon tank. So it's not like steel had never been used before. It actually had been used. And when you look at the material properties of stainless steel, especially very, if it's been very, like, full hard, strain hardened stainless steel at cryogenic temperature, the strength-to-weight is actually similar to carbon fiber. So if you look at the material properties at room temperature, it looks like the steel

• 2:00:58Speaker 00

  is, it's going to be twice as heavy. But if you look at the material properties at cryogenic temperature of full hard steel, stainless, of particular grades, then you actually get to a similar strength-to-weight as carbon fiber. And in the case of Starship, both the fuel and the oxidizer are cryogenic. So for Falcon 9, the fuel is rocket propelled grade kerosene, basically pure, like a very pure form of jet fuel, which is, but that is roughly room temperature. Although we do actually, we do actually chill it slightly below, we chill it like a bit. Delicious. Yeah, we do chill it, but it's not cryogenic.

• 2:01:43Speaker 00

  In fact, if we made it cryogenic, it would just turn to wax. So, but for Starship, the, it's liquid methane and liquid oxygen. They, they, they're liquid at similar temperatures. So, so basically, almost the entire primary structure is at cryogenic temperature. So then you've got a 300 series stainless, that's strain hardened, because it's, almost all things at cryogenic temperature, actually has a similar strength-to-weight as carbon fiber. It costs 50 times less than raw material and is very easy to work with.

• 2:02:27Speaker 00

  You can weld stainless steel outdoors. You could smoke a cigar while welding stainless steel. It's like, it's very resilient. You can modify it easily. It's, if you want to attach something, you just weld it right on.

• 2:02:43Speaker 00

  So very easy to work with, very low cost. And like I said, at cryogenic temperature, similar strength to weight to carbon fiber. Then when you factor in that, that we don't need, we don't, we have a much reduced heat shield mass, because the melting point of steel is much greater than the melting point of aluminum. It's about twice the melting point of aluminum.

• 2:03:12Speaker 01

  So you can just run the rocket much hotter?

• 2:03:15Speaker 00

  Yes. So especially for the ship, which is coming in like a blazing meteor. It is the, you can greatly reduce the mass of the heat shield. So you can call it, cut the mass of the windward part of the heat shield in, maybe in half. And you don't need any heat shielding on the, on the leeward side. So the net, if net result is actually the steel rocket weighs less than the carbon fiber rocket. Because the resin in the carbon fiber rocket, it starts to melt.

• 2:03:59Speaker 00

  So basically, carbon fiber and aluminum have about the same operating temperature capabilities. And whereas steel can operate at twice temperature. I mean, these are very rough approximations. People will.

• 2:04:12Speaker 01

  I won't build a rocket base.

• 2:04:13Speaker 00

  What I'm just like, people say, oh, he said it's twice. It's actually, it's actually point eight. Shut up assholes.

• 2:04:18Dwarkesh Patel?

  That's what the main comment is going to be about.

• 2:04:19Speaker 00

  God damn it. Okay. The point is, actually, in retrospect, we should have started with done steel in the beginning. It was dumb not to do steel.

• 2:04:28Speaker 01

  Okay, but to play this back to you, what I'm hearing is that steel was a riskier, less proven path, other than the early US rockets versus carbon fiber was like a worse but more proven out path. And so you need to be the one to push for, hey, we're going to do this riskier path and just figure it out. And so you're fighting like a sort of conservatism, in a sense.

• 2:04:51Speaker 00

  That's why I initially said like, that the issue is that we weren't making fast enough progress. We were having trouble making even. A small barrel section of the carbon fiber that didn't have wrinkles in it. So because at that large scale, you have to have many plies, many sort of layers of the carbon fiber. You've got to cure it and you've got to cure it in such a way that it doesn't have any wrinkles or defects. The carbon fiber is much less resilient than than steel. It has much, it's less toughness. Like stainless steel will scratch and bend. The carbon fiber will tend to shatter.

• 2:05:33Speaker 00

  So toughness being the area under the stress strain curve. So that you're generally going to have to do better with steel. Or stainless steel, to be precise.

• 2:05:44Speaker 01

  One other Starship question. So I visited Starbase two years ago with Sam Teller, and that was awesome. It was very cool to see in a whole bunch of ways. One thing I noticed was that people really took pride in the simplicity of things where, you know, everyone was telling you how Starship is just a big soda can and, you know, we're hiring welders and, you know, if you can weld in any industrial project, you can weld here. But there's a lot of pride in the simplicity.

• 2:06:15Speaker 00

  And it's, well, Starship is a very complicated rocket.

• 2:06:18Speaker 01

  So that's what I'm getting at. Is, are things simpler or are they complex?

• 2:06:23Speaker 00

  I think maybe just what they're trying to say is that, you know, you don't have to have like prior experience in the rocket industry to work on a Starship.

• 2:06:31Speaker 00

  You know, somebody just needs to be, you know, smart and work hard and be trustworthy and make it work in a rocket. They don't, they don't need prior rocket experience. Starship is the most complicated machine ever made by humans, by a long shot.

• 2:06:47Speaker 01

  In what regards?

• 2:06:49Speaker 00

  Anything, really. There isn't a more complex machine.

• 2:06:55Speaker 00

  Yeah, I mean, I'd say that there's pretty much any project I can think of would be easier than this. And that's why no one has made a rapidly reusable, nobody has ever made a fully reusable orbital rocket. It's a very hard, very hard problem.

• 2:07:13Speaker 00

  I mean, many smart people have tried before, very smart people, with immense resources, and they failed. So, and we haven't succeeded yet. You know, Falcon is partially reusable, but the up to stage is not. Starship version three, I think this design,

• 2:07:36Speaker 00

  that it can be fully reusable. And that full usability is what will enable us to become a multiplanet civilization.

• 2:07:44Speaker 01

  Can you say about the circles?

• 2:07:45Speaker 00

  So, I don't, like, I'd say I could,

• 2:07:50Speaker 00

  any technical problem, even like a hydrocladder or something like that, is an easier problem than this.

• 2:07:55Speaker 01

  We spent a lot of time on battle knives. Can you say what the current Starship battle knives are, even at the high level?

• 2:08:01Speaker 00

  I mean, trying to make it not explode, generally.

• 2:08:04Speaker 01

  That old chestnut.

• 2:08:06Speaker 00

  Really wants to explode.

• 2:08:09Speaker 01

  All those combustible materials.

• 2:08:10Speaker 00

  We've had two boosters explode on the test stand. One obliterated the entire test facility.

• 2:08:18Speaker 00

  So it only takes that one mistake. And, I mean, the amount of energy contained in Starship is insane.

• 2:08:25Speaker 01

  And so is that why it's harder than Falcon? It's because it's just more energy?

• 2:08:30Speaker 00

  It's a lot of new technology. It's pushing the performance envelope. The Raptor three engine is a very, very advanced engine, by far the best rocket engine ever made. But it desperately wants to blow up. I mean, just to put this in perspective here, on liftoff, the rocket is generating over a hundred gigawatts of power.

• 2:08:55Speaker 00

  It's 20% of yours. The truth is, it's insane.

• 2:08:58Speaker 01

  It's a great comparison.

• 2:08:59Speaker 00

  While not exploding.

• 2:09:01Speaker 01

  Sometimes.

• 2:09:02Speaker 00

  Sometimes, but sometimes, yeah, so I was like, how does it not explode? There's, you know, thousands of ways that it could explode and only one way that it doesn't. So we want it to merely, not really not explode, but fly reliably.

• 2:09:19Speaker 00

  On a daily basis, like once per hour. And obviously, you know, blows up a lot. It's very difficult to maintain that launch cadence.

• 2:09:26Dwarkesh Patel?

  Yes.

• 2:09:28Speaker 00

  And then I'm gonna say like, what's the, what's the single biggest remaining problem for Starship? It's having the heat shield be reusable, such that the, no one's ever made a reusable orbital heat shield. So the, the, the heat shield's got to make it through the ascent phase without shocking a bunch of tiles. And then it's going to come back in and also not lose a bunch of tiles or overheat the main, the main airframe.

• 2:10:01Dwarkesh Patel?

  Isn't that hard? It's kind of fundamentally a consumable.

• 2:10:05Speaker 00

  Well, yes, but your brake pads in your car are also consumable, but they last a very long time.

• 2:10:09Dwarkesh Patel?

  Fair.

• 2:10:10Speaker 00

  So it just needs to last a very long time.

• 2:10:14Speaker 00

  But that's, it just, yeah, try, I mean, we have brought the ship back and had it do a soft landing in the ocean. Done it a few times, but, but it lost a lot of tiles. You know, and, you know, it was not reusable without a lot of work.

• 2:10:31Dwarkesh Patel?

  Yeah.

• 2:10:32Speaker 00

  So even though it did land, it did come to soft landing, it was, would not have been reusable without a lot of work. And that, so it's not really reusable in that sense. That's, that's the biggest problem that remains is fully reusable heat shield.

• 2:10:48Speaker 00

  So like, if you want to be able to land it, refill propellant and fly again without, you know, you can't do this laborious inspection of, you know, 40,000 tiles, sort of thing.

• 2:10:59Dwarkesh Patel?

  I, I'm curious how you drive, like, when you, when I read biographies of yours, it just, it, it's, it seems like you're just able to drive the sense of like urgency and drive the sense of like, this is the, this is the thing that can scale. And I'm curious why you think other organizations of your, like SpaceX and Tesla are really big companies now, and you're still able to keep that culture. What goes wrong with other companies such that they're not able to do that?

• 2:11:26Speaker 00

  I don't know.

• 2:11:28Dwarkesh Patel?

  But like today, you said you had like a bunch of SpaceX meetings. Like, what, what is it that you're doing there that's like keeping that?

• 2:11:33Speaker 00

  It's adding urgency.

• 2:11:34Dwarkesh Patel?

  Yeah, yeah, yeah.

• 2:11:37Speaker 00

  Well, I don't know. I guess the urgency is going to come from whoever's leading the company. So my sense of urgency, I have like a maniacal sense of urgency. So that maniacal sense of urgency projects through the rest of the company.

• 2:11:52Dwarkesh Patel?

  Is it because of consequences? They're like, if Elon set a crazy deadline, but if I don't get it, I know what happens to me. Is it just you're able to identify bottlenecks and get rid of them? So people can move fast? Like, how do you, how do you think about why your companies are able to move fast?

• 2:12:06Speaker 00

  Yeah, I'm constantly addressing the limiting factor.

• 2:12:10Speaker 00

  So.

• 2:12:15Speaker 00

  I mean, I mean, on the deadlines front, I mean, I generally actually try to aim for a deadline that I at least think is at the 50th percentile. So it's not, it's not like an impossible deadline, but it's the most aggressive deadline I can think of that could be achieved with 50% probability, which means that it'll be late half the time.

• 2:12:37Speaker 00

  And.

• 2:12:40Speaker 00

  But whatever, like, there is like a law of gases expansion that applies to schedules. Like whatever given, whatever schedule you, like, if you said we're going to do this, something in like five years, which to me is like infinity time.

• 2:12:54Speaker 00

  It will expand to fully available schedule and it will take five years. You know, there's like, there's a physical limit. Like that, like physics will limit how fast you can do certain things. Like, so like scaling up manufacturing, there's like, there's a rate at which you can move the atoms and scale manufacturing. That's why you can't like instantly make, you know, a million or something, million years to a year or something. You've got, you've got a design manufacturing line, you could bring it up, you could ride the S curve of production.

• 2:13:27Speaker 00

  So, yeah, I guess like, what can I say that's, that's actually helpful to people?

• 2:13:37Speaker 00

  I think generally a maniacal sense of urgency is a very big deal. So and you want to have you want to have you want to have an aggressive schedule. And then you and you and you want to figure out what the limiting factor is at any point in time and and help the team address that limiting factor.

• 2:13:57Speaker 01

  Can you maybe talk about the so Starlink was slowly in the works for many years.

• 2:14:03Speaker 00

  Yeah, we talked about it all the way in the beginning of the company.

• 2:14:07Speaker 01

  Yeah. And so then there was a team you had built in Redmond and then at one point you decided this team is just not cutting us. But again, how did you like it went for a few years slowly. And so why did this why didn't you act earlier and why did you act when you did like why was that the right moment at which to act?

• 2:14:29Speaker 00

  I mean, I have I have these very detailed engineering reviews weekly that that's that's maybe a very unusual level of granularity. I don't know anyone who runs a company or at least a manufacturing company that that goes to the level of detail that that I go into. So it's not it's not as though like I have a pretty good understanding of what's actually going on because we go we go through things in detail. And I'm a big believer in skip level meetings where the individuals instead of having the

• 2:15:09Speaker 00

  person that reports to me say things, it's everyone that reports to them says something in the technical review. And and there can't be advanced preparation. So otherwise you're going to get, you know, glazed, as I say these days.

• 2:15:31Speaker 01

  Yeah, exactly. Very Gen Z of you.

• 2:15:32Speaker 00

  Very Gen Z.

• 2:15:33Dwarkesh Patel?

  How do you prevent advanced and you just like call them randomly like?

• 2:15:36Speaker 00

  No, just go out there and everyone provides an update. So I mean, it's a lot of information to keep in your head because you've got to, you've got then, say, if you have meetings weekly or twice weekly, you've got a snapshot of what that person said. And and you can and you can then, you know, plot the progress points. You can sort of mentally plot the points on a curve and say, are we converging to a solution or not? Or are we, you know, like, I'll take drastic action only when I conclude that success is not in a set of possible outcomes.

• 2:16:22Speaker 00

  So when I say, okay, when I finally reached the conclusion that, okay, unless drastic action is done, we have no chance of success, then I must take drastic action. And so that's that's I came to that conclusion in 2018, took drastic action and fixed the problem.

• 2:16:38Dwarkesh Patel?

  How many, you know, you've got many, many companies. And in each of them, it sounds like you do this kind of deep engineering understanding of what their own bottlenecks are. So you can do these reviews with people.

• 2:16:53Speaker 00

  Yeah.

• 2:16:56Dwarkesh Patel?

  You've been able to scale it up to five, six, seven companies. Within one of these companies, you have many different mini companies within them. What determines the maximum here? Could you have like 80 companies?

• 2:17:07Speaker 00

  Eighty? No.

• 2:17:08Dwarkesh Patel?

  But you have so many already. Like, that's that's already remarkable.

• 2:17:12Speaker 00

  Why this current number?

• 2:17:12Dwarkesh Patel?

  Yeah, exactly.

• 2:17:14Speaker 00

  I know. So. We can barely keep one company together.

• 2:17:20Speaker 00

  It's like, it depends on situation. So I actually don't don't have regular meetings with foreign company. So that foreign company is sort of cruising along. But look, basically, if something is working well and making good progress, then there's no point in me spending time on it. So I actually allocate time according to where the where the limiting factor or the problem, where are things problematic? And or where are we pushing against?

• 2:17:54Speaker 00

  Like, what is holding us back? You know, I focus, the risk of saying the words too many times, the limiting factor.

• 2:18:04Speaker 00

  So basically, if something's good, like the irony is if something's going really well, they don't see much of me. But if something's going badly, they'll see a lot of me.

• 2:18:14Speaker 01

  And so if something's

• 2:18:15Speaker 00

  Or not even badly, it's like

• 2:18:17Speaker 01

  If something's a limiting factor.

• 2:18:18Speaker 00

  It's a limiting factor, exactly. It's not exactly going badly, but it's the thing that's, it's the thing that we need to make go faster to

• 2:18:23Speaker 01

  And so when something's a limiting factor at SpaceX or Tesla, are you like talking weekly and daily with the engineer that's working on it? How does that actually work?

• 2:18:36Speaker 00

  Most things that are a limiting factor are weekly and some things are twice weekly. So the AI five chip review is twice weekly. And so it's every Tuesday and Saturdays is the chip review.

• 2:18:50Speaker 01

  Is it open ended in how long it goes?

• 2:18:54Speaker 00

  Technically, yes, but usually it's like two or three hours. So I mean, sometimes less. It's, it depends on how much information you're going to go through.

• 2:19:05Speaker 01

  Yeah, that's another thing. And I'm just trying to tease out the differences here because the outcomes seem quite different. And so I think it's interesting to know what inputs are different. And it feels like the corporate world, one, like you're saying, just the CEO doing engineering reviews does not always happen, despite the fact that that is the, you know, what the company is doing. But then time is often pretty finely sliced into, you know, half hour meetings or even 15 minute meetings. And it seems like you hold more open ended, we're talking about it until we figure it out type meetings. Yeah.

• 2:19:41Speaker 00

  Yeah, sometimes, but most of them seem to more or less stay on time. So, I mean, today's Starship engineering review went a bit longer because there were more topics to discuss.

• 2:20:02Speaker 00

  They're trying to figure out how to scale 2 million plus tons to over per year is quite challenging.

• 2:20:08Dwarkesh Patel?

  Can I ask a question? So you said about Optimus and AI that they're gonna result in double just a growth rates within a matter of years.

• 2:20:17Speaker 00

  Oh, like the economy?

• 2:20:18Dwarkesh Patel?

  Yeah.

• 2:20:19Speaker 00

  Yes, I think that's right.

• 2:20:21Dwarkesh Patel?

  What was the point of the Doge cuts if the economy is gonna grow so much?

• 2:20:27Speaker 00

  Well, I think like wasting food are not good things to have, you know.

• 2:20:33Speaker 00

  I was actually pretty worried about, I guess, I mean, I think in the absence of AI and robotics, we're actually totally screwed because the national debt is piling up like crazy. Now, our interest payments, the interest payments to the national debt exceed the military budget, which is a trillion dollars. So if over a trillion dollars just the interest payments. You know, that was like, I was like, okay, pretty concerned about that. Maybe if I spend some time, we can slow down the bankruptcy of the United States and give us enough time for the AI and robots to, you know, help solve the national debt. Or not help solve, it's the only thing that could solve

• 2:21:17Speaker 00

  the national debt. Like, we are 1,000% gonna go bankrupt as a country and fail as a country without AI and robots. Nothing else will solve the national debt. And so we'd like to, well, we just need, we need enough time to get, build the AI and robots to not go bankrupt before then.

• 2:21:38Dwarkesh Patel?

  I guess the thing I'm curious about is when Doge starts, you have this enormous ability to enact reform and

• 2:21:47Speaker 00

  Not that enormous.

• 2:21:48Dwarkesh Patel?

  Sure, sure. But totally by your point that like, it's important that AI and robotics drive product improvements, drive GDP growth. But why not just directly go after the things you were pointing out, you know, like the tariffs on certain components, or rather it's like permitting?

• 2:22:03Speaker 00

  I'm like the president.

• 2:22:07Speaker 00

  And and very hard to cut to cut to even even to to cut things that are obvious waste and fraud, like like ridiculous waste and fraud. What I discovered that is it's extremely difficult even to cut very obvious waste and fraud from the government. Because the the government tends to operate on a on like who's complaining like if if and if you cut off payments to fraudsters, they immediately come up with the most sympathetic sounding reasons to continue the payment. They don't say, please keep the fraud going. They say, you know, it's they're like, you're killing baby pandas. And we're like, meanwhile, there's no baby pandas are dying, they're just making it up.

• 2:22:50Speaker 00

  The forces are capable of of coming up with extremely compelling sort of heart wrenching stories that are false, but nonetheless sound sympathetic. And that that's what happened. And so it's like, perhaps I should have known better. And I thought, wait, let's take a let's let's try to cut some amount of of waste and fraud from the government. Maybe there shouldn't be, you know, 20 million people marked as alive in social security who are definitely dead and over the age of 115. The oldest American is 114. So it's safe to say if somebody is 115 and marked as alive in the social security database,

• 2:23:35Speaker 00

  something is there's either a typo, like somebody should call them and say, we we seem to have your birthday wrong, or or or we need to mark you as dead. One of the two things. Very intimidating call to get. Well, so it seems like a reasonable thing. And if if like, say, their birthday is in the future, and they have, you know, a small business administration loan, and their birthday is 2165.

• 2:24:07Speaker 00

  We either gain have a typo or we have fraud. So we say we appear to have gotten the century of your birth incorrect. Or a great plot for a movie. Yes, this is, this is what I mean by ludicrous fraud. That's what I mean by ludicrous fraud. Were those people getting payments? Some were getting payments from social security. But, but, but the main fraud vector was to mark somebody as alive in social security and then use every other government payment system to basically to do fraud. Because what those other government payment systems do would do, they would simply do an are you alive check to the social security database. It's a, it's a bank shot. What would you estimate is like the total amount of fraud from this mechanism?

• 2:24:50Speaker 00

  My guess is, and other, by the way, the government accountability office has done these estimates before. I'm not the only one who's coming out of this, you know. In fact, I think they, they did, the GAO did analysis, a rough estimate of fraud during the Biden administration and calculated at roughly half a trillion dollars. So don't take my word for it. Take it a report issued during the Biden administration. How about that? From this social security mechanism? It's one of many. It's important to appreciate that the government does not, is very ineffective at stopping fraud because it's not like, like if it was a company, like stopping fraud, you've got a motivation because it's affecting the earnings of your company.

• 2:25:34Speaker 00

  But the government just, they just print more money. So it's not, like you need, you need caring and competence. And these are in short supply at the, at the federal level.

• 2:25:49Speaker 00

  Yeah, I'm sorry. I mean, when you go to the DMV, do you think, wow, this is a bastion of competence? Well, now imagine it's worse than the DMV because it's the DMV that can print money.

• 2:26:01Dwarkesh Patel?

  So was it not possible?

• 2:26:02Speaker 00

  At least the state level DMVs need to, the states more or less need to stay within their budget of legal bankrupt, but the federal government just prints full money.

• 2:26:10Dwarkesh Patel?

  Was it not possible to cut that if there's a catchy half a trillion of fraud? Why was it not possible to cut all that?

• 2:26:18Speaker 00

  Because when, as soon as you, we did, we actually, no. You really have to stand back and recalibrate your expectations for competence.

• 2:26:32Speaker 00

  Because you're operating in a world where, you know, you've got to sort of make ends meet. Like, you know, you've got to pay your bills, you've got to, you know.

• 2:26:41Dwarkesh Patel?

  Buy the microphones.

• 2:26:42Speaker 00

  Yeah, yeah, exactly. So you, if you don't have, it's not like there's a giant, largely uncaring monster bureaucracy. It's not even, and a bunch of necroistic computers that are just, that are just sending payments.

• 2:27:01Speaker 00

  Like, one of the things that those things that there was, and it sounds so simple, that probably will save, let's say a hundred billion, maybe 200 billion a year, is simply requiring that payments from the main treasury computer, which is called PAM, it's like Payment Accounts Master or something like that, there's five trillion PAMs a year, requiring that any payment that goes out have a payment appropriation code, make it mandatory, not optional, and that you have anything at all in the comment field.

• 2:27:36Speaker 00

  Because, you see, you have to recalibrate how dumb things are.

• 2:27:42Dwarkesh Patel?

  But you didn't.

• 2:27:42Speaker 00

  Payments were being sent out with no appropriation code, not checking back to any congressional appropriation, and no explanation. And this is why the Department of War, formerly Department of Defense, cannot pass an audit because the information is literally not there.

• 2:28:01Speaker 00

  Recalibrate your expectations.

• 2:28:02Dwarkesh Patel?

  I want to understand this half a trillion number because there's an IG report in 2024.

• 2:28:06Speaker 00

  How, you must like, why is it so low?

• 2:28:10Dwarkesh Patel?

  Maybe, but we found that like over seven years, the social security fraud they estimated was like 70 billions over seven years, so like 10 billion a year. So I'd be curious to see what like the other 490 billion is.

• 2:28:20Speaker 00

  Federal government expenditures are seven and a half trillion a year.

• 2:28:22Dwarkesh Patel?

  Yeah.

• 2:28:25Speaker 00

  What percentage, how competent do you think government is?

• 2:28:28Dwarkesh Patel?

  The discretionary spending there is like 15%.

• 2:28:33Speaker 00

  Yeah, but it doesn't matter. Most of the fraud is non-discretionary. It's basically fraudulent Medicare, Medicaid, social security, you know,

• 2:28:45Speaker 00

  disability, there's a zillion government payments.

• 2:28:49Dwarkesh Patel?

  Yeah.

• 2:28:50Speaker 00

  And a bunch of these payments are in fact, they're block transfers to the states. So the federal government doesn't even have the information in a lot of cases to even see, know if there's fraud. Let's consider, let's like redact your ad absurdum. The government, the government is perfect and has no fraud. What is your probability estimate of that?

• 2:29:13Speaker 00

  I mean, zero. Okay. So then would you say that fraud and waste, that the government is, has, is 90%? That also would be quite generous. But if it's only 90%, that means that there's 750 billion dollars a year of waste and fraud. And it's not 90%. It's not 90% effective.

• 2:29:36Dwarkesh Patel?

  This seems like a strange rate of first principles, the amount of fraud in the government. Just like, how much do you think there is? And then, anyways, we don't have to do a lot, but I'd be curious to like.

• 2:29:45Speaker 00

  You know a lot about fraud at Stripe. People are constantly trying to do fraud.

• 2:29:49Speaker 01

  Yeah, but as you say, it's like a little bit of a, we've really grounded down, but it's a little bit of a different problem space because we're dealing with a much more heterogeneous set of fraud vectors here than we're.

• 2:29:59Speaker 00

  Yeah, but I mean, I mean, that's right. You have high competence and you try hard.

• 2:30:06Speaker 00

  You have high competence and high caring, but still fraud is non-zero. Now, now imagine it's at a much bigger scale. There's much less competence and much less caring.

• 2:30:20Speaker 00

  You know, back PayPal, back in the day, we tried to manage fraud down to about 1% of the payment volume. And that was very difficult, took a tremendous amount of confidence and carrying to get fraud merely to 1%.

• 2:30:33Speaker 00

  Now imagine that you're an organization where there's much less carrying and much less confidence. It's going to be much more than 1%.

• 2:30:41Speaker 01

  How do you feel now, looking back on kind of politics and doing stuff there, where it feels like from the outside in, the, you know, two things have been quite impactful. One, the America PAC and two, the acquisition of, well, Twitter at the time. But also, it seems like there was a bunch of heartache. And so what's your, what's your grading of the whole experience?

• 2:31:11Speaker 00

  Well, I think, I think those things needed to be done to maximize the probability that the future is good.

• 2:31:24Speaker 00

  So, but politics generally is very tribal and it's, it's very tribal and people lose their objectivity usually with politics. Like they, they generally have trouble seeing the good on the other side or the bad on their own side. That's generally how it goes.

• 2:31:44Speaker 00

  I, I, that I guess was one of the things that surprised me the most is you often simply cannot reason with people. If they're in one tribe or the other, they, they simply believe that everything their tribe does is good and anything the other political tribe does is bad. And persuading them is, otherwise is almost impossible.

• 2:32:05Speaker 00

  So, anyway, but.

• 2:32:11Speaker 00

  I think, I think overall, those actions,

• 2:32:18Speaker 00

  acquiring Twitter,

• 2:32:21Speaker 00

  getting Trump elected, even though it makes a lot of people angry. I think those, I think those actions are good for, were good for civilization.

• 2:32:30Dwarkesh Patel?

  Yeah, well, how does it feed into the future you're excited about?

• 2:32:33Speaker 00

  Well,

• 2:32:35Speaker 00

  America needs to, America needs to be strong enough to last long enough to extend life to other planets and to try to get, I guess, AI and robotics to the point where we can ensure that the future is good. Like, on the other hand, if, if we were to descend into, say communism or some situation where the state was extremely oppressive, that, that would mean that we, we might not be able to become multi-planetary.

• 2:33:08Speaker 00

  And we might, the, the state might,

• 2:33:12Speaker 00

  you know, stamp out

• 2:33:15Speaker 00

  our progress in AI and robotics.

• 2:33:17Dwarkesh Patel?

  How do you feel about,

• 2:33:20Dwarkesh Patel?

  you know, Optimus, Grok, et cetera, are going to be leveraged by, and not just yours, any revenue maximizing company's products will be leveraged by the government over time. How does this concern manifest in what private company should be willing to give governments, what kinds of guardrails should, like, should, you know, should AI models be

• 2:33:47Dwarkesh Patel?

  made to do whatever the government that has contracted them out to do, ask them to do? Should, like, should, should Grok get to say, like, actually even the military wants to do X? No, Grok will not do that.

• 2:34:01Speaker 00

  I think probably the biggest danger of AI, well, maybe the biggest danger of fail for, for AI and robotics going wrong, wrong is, is government. Interesting. You know?

• 2:34:14Speaker 00

  I mean, the way you think, like, like, people who are opposed to corporations or worried about corporations should really worry about the most about government because government is just a corporation in the limit. It's a government, it is, it is, it is, government is just the biggest corporation with a monopoly on violence.

• 2:34:34Speaker 00

  So I always find it like a strange dichotomy where people would think corporations are bad, but the government is good. When the government is simply the biggest and worst corporation.

• 2:34:47Speaker 00

  But people have that dichotomy. They somehow think at the same time that government can be good, but corporations bad, and this is not true. Corporations are, have better morality than the government.

• 2:34:58Dwarkesh Patel?

  It is.

• 2:34:58Speaker 00

  So I actually think it's, you know, that's, that is a thing to be worried about. It's like, if, you know, should, should, if the government should not, like, the government could potentially use AI and robotics to suppress the population. Like, that is a serious concern.

• 2:35:17Dwarkesh Patel?

  As a guy building AI and robotics, how do you, how do you like, how do you prevent that?

• 2:35:22Speaker 00

  Well, I think that, like, if you have a limited government,

• 2:35:28Speaker 00

  if you limit the powers of government, which is like really what the US constitution is intended to do, is intended to limit the powers of government, then, then you're probably going to have a better outcome than if you have more government. So. Robotics will be available to all governments, right? Not about all governments. I mean,

• 2:35:49Speaker 00

  it's difficult to predict the, like I said, like, what's, what's, what's the end, end point? Or like, what is, what is many years in the future? But it's difficult to predict the, the sort of path along, along that way.

• 2:36:03Speaker 00

  Like, if civilization progresses, AI will vastly exceed the sum of all human intelligence. And there will be far more robots than humans.

• 2:36:16Speaker 00

  Along the way, what happens?

• 2:36:19Speaker 00

  It's very difficult to predict.

• 2:36:20Dwarkesh Patel?

  I mean, it seems like one thing you could do is just say,

• 2:36:24Dwarkesh Patel?

  you are not allowed to, whatever government decks, you're not allowed to use Optimus to do XYZ. Just write out like a policy. I mean, I think you treated recently that Grok should have a moral constitution. And one of those things could be that we limit what governments are allowed to do with this advanced technology.

• 2:36:40Speaker 00

  I mean, yeah, we can do what is, what, I mean,

• 2:36:47Speaker 00

  technically, I mean, if the politicians pass a law, and they can enforce that law, then it's hard to not do that law. The best thing we can have is limited government, where you have the appropriate cross checks between the executive judicial and legislative branches.

• 2:37:10Dwarkesh Patel?

  I guess the reason I'm curious about it is this like, at some point, it seems like the limits will come from you, right? Like, you've got the Optimus, you've got the space GPUs, you've got the-

• 2:37:20Speaker 00

  You think I will be the boss of the government?

• 2:37:22Dwarkesh Patel?

  Or you will get the, I mean, already it's the case with SpaceX, that for things that are crucial to the, like, the government really cares about getting certain satellites up in space, whatever, like, it needs SpaceX. It is a necessary contractor. And you are in the process of building more and more of the

• 2:37:44Dwarkesh Patel?

  the technological components of the future that will have an analogous role in different industries. And you could have this ability to like, set some policy that, you know, is suppressing classical liberalism in any way. My companies will not help in any way with that, or, you know, some policy like that.

• 2:38:05Speaker 00

  I will do my best to ensure that anything that's within my control maximizes the good outcome for humanity.

• 2:38:15Speaker 00

  I think anything else would be short sighted. Because obviously, I'm part of humanity, so. I like humans. Pro-human, pro-human.

• 2:38:29Dwarkesh Patel?

  You've mentioned that Dojo 3 will be used for space-based compute.

• 2:38:34Speaker 00

  You really read my, what I say.

• 2:38:38Dwarkesh Patel?

  I don't know if you know Twitter, but I know Elon. You have a lot of followers.

• 2:38:41Speaker 00

  Don't give away. Honestly, you've just stolen my secrets. I posted my.

• 2:38:48Dwarkesh Patel?

  How do you design a ship for space? What changes?

• 2:38:53Speaker 00

  Well, I guess you want it designed to be more radiation tolerant and run at a higher temperature. So you could, you know, roughly, if you increase the operating temperature by 20th set in degrees Kelvin, you can cut your radiator mass in half. So running at a higher temperature is helpful in space. I mean, there's various things you could do for shielding the memory. But like neural nets are going to be very resilient to bit flips. So like most of what happens for radiation is like random bit flips.

• 2:39:33Speaker 00

  But like if you've got like, you know, a multi-trailing parameter model and you get a few bit flips, it doesn't matter. It's like heuristic programs are going to be much more sensitive to bit flips than some giant parameter file. So I just designed it for unhot and. I think you pretty much do it the same way that you do things on Earth, apart from make it run hotter.

• 2:40:02Dwarkesh Patel?

  I mean, the solar arrays, most of the weight on the satellite, is there a way to make the GPUs even more power dense than what Nvidia and TPUs and et cetera are planning on doing that would, you know, be especially privileged in the space-based world?

• 2:40:17Speaker 00

  Well, I mean, the basic math is like, if you can do about a kilowatt per reticle, and then you'd need, you know, 100 million full reticle chips to do 100 gigawatts.

• 2:40:37Speaker 00

  So yeah, depending on what your yield assumptions are, you know, that tells you how many chips you need to make. But cool, you need, if you're gonna have 100 gigawatts of power, you need, you know, 100 million chips running, that are running a kilowatt sustained, but theoretical.

• 2:41:03Dwarkesh Patel?

  100

• 2:41:04Speaker 00

  Basic math.

• 2:41:05Dwarkesh Patel?

  100 million chips, it depends on, yeah, if you look at the die size of something like Black Bolt G P S or something, and how many you can get out of a wafer, you can get like on the order of dozens or less per wafer. So basically, this is a world where if we're putting that out every single year, you're producing millions, millions of wafers a month.

• 2:41:33Dwarkesh Patel?

  That's the plan with TeraFab, millions of wafers a month of advanced process notes.

• 2:41:37Speaker 00

  It's could be some number, north of a million, I think. You gotta do the memory, too.

• 2:41:41Dwarkesh Patel?

  Yeah, are you gonna make a memory fab?

• 2:41:44Speaker 00

  I think the TeraFab's gotta do memory, it's gotta do logic, memory, and packaging.

• 2:41:48Dwarkesh Patel?

  I'm very curious how somebody like gets start. This is like the most complicated thing man has ever made. And obviously, like, if anybody's up to the task, you're up to the task. Like, what do you, so you realize this is a bottleneck, and you go to your engineers and like, what is the next, like, what do you tell them to do? I want a million wafers a month in 2030. What is the next, like, what do you?

• 2:42:08Speaker 00

  That's right.

• 2:42:09Dwarkesh Patel?

  Do you like call A S M L, like, what is the?

• 2:42:10Speaker 00

  That's what I want.

• 2:42:13Dwarkesh Patel?

  What is the next step?

• 2:42:14Speaker 01

  That's so much to ask.

• 2:42:15Speaker 00

  Well, we make a little fab and see what happens. Make our mistakes at a small scale and then make a big one.

• 2:42:25Dwarkesh Patel?

  Is a little fab done?

• 2:42:27Speaker 00

  No, it's not done. Which, I mean, George, they're not gonna keep that cat in the bag.

• 2:42:33Speaker 00

  That cat's gonna come out of the bag, bro. It'll be like drones hovering over the bloody thing, you know? You'll be able to see its construction progress on X, right? You know, in real time.

• 2:42:43Speaker 00

  So, no, we, I mean, like, I don't know. We could just flounder in failure. To be clear, it's like not, success is not guaranteed, but.

• 2:42:55Speaker 00

  Since we want to try to make, you know, something like a hundred million. We need, we want a hundred gigawatts of power and a hundred, a hundred chips that can take a hundred gigawatts. And so call it, you know, but yeah, by, by 2030. So then.

• 2:43:17Speaker 00

  And we'll take as many chips as our suppliers will give us. I've said this to, I've actually said this to TSMC and Samsung and Micron. It's like, please build your more fabs faster. And we will guarantee you to buy the output of those fabs.

• 2:43:32Speaker 00

  So they're already like moving as fast as they, as they can. Like it's, it's not like, to be clear, it's not like us.

• 2:43:39Speaker 00

  You know, it's not like either, it's, it's, it's not like. It's us plus them, you know?

• 2:43:46Speaker 01

  There's an narrative that the people doing AI want a very large number of, you know, chips as quickly as possible. And then many of the input suppliers, the fabs, but also, you know, the turbine manufacturers are not ramping up production very quickly.

• 2:44:02Speaker 00

  No.

• 2:44:03Speaker 01

  Yeah. The explanation you hear is that they're dispositionally conservative. You know, they're Taiwanese or German as the story may be. And they just like, don't believe the say, like, is that really the explanation or is there something else?

• 2:44:17Speaker 00

  Well, I mean, it's reasonable to like, if somebody's been in, say, the computer memory business for

• 2:44:24Speaker 00

  thirty or forty years.

• 2:44:25Speaker 01

  And they've seen cycles.

• 2:44:27Speaker 00

  They've seen like boom and bust like 10 times.

• 2:44:29Speaker 01

  Yeah.

• 2:44:30Speaker 00

  You know, so, so like, that's a lot of layers of scar tissue, you know? So it's like, it's like during the boom times, it looks like everything is going to be great forever. And then, then, then the crash happens and then like desperately trying to avoid bankruptcy. And, and then there's another boom and another crash.

• 2:44:46Speaker 01

  Are there other, are there other ideas you think others should go pursue? That you're not for whatever reasons right now?

• 2:44:55Speaker 00

  I mean, there are a few companies that are, that are pursuing like new ways of doing chips.

• 2:45:01Speaker 00

  But there, there's just not scaling fast.

• 2:45:03Speaker 01

  I, I don't even mean within AI, I mean, just generally.

• 2:45:06Speaker 00

  I'd say like people should just should do the thing that where they find that they're highly motivated to do that thing.

• 2:45:12Dwarkesh Patel?

  Uh huh.

• 2:45:13Speaker 00

  As opposed to, you know, some idea that I suggest. But they should do the thing that they find personally interesting and motivating to do.

• 2:45:23Dwarkesh Patel?

  Uh huh.

• 2:45:28Speaker 00

  But, you know, going back to the limiting factor, he was upraised about 100 times.

• 2:45:36Speaker 00

  The, the current limiting factor that I see in the timeframe, you know, in the sort of 20, 29, 20, like in the, in the three, three to four year timeframe, it's chips.

• 2:45:53Speaker 00

  In the one year timeframe, it's, it's energy, power production, electricity. Like, it's, it's not clear to me that there's enough usable electricity to turn on all the AI chips that are being made.

• 2:46:09Speaker 00

  Towards the end of this year, I think we're going to have real trouble turning on, like the chip output will exceed the, the ability to turn chips on.

• 2:46:17Dwarkesh Patel?

  What, what's your plan to deal with that world?

• 2:46:20Speaker 00

  Well, we're trying to accelerate electricity production.

• 2:46:24Speaker 00

  I, I guess that's, that's maybe one of the reasons that XAI will be maybe the leader, hopefully the leader, is that we'll be able to turn on more chips than other people can turn on faster.

• 2:46:39Speaker 00

  Because we're, we're good at hardware. And, and, and generally, the, the innovations from the corporations that call themselves labs, the, the ideas tend to flow, like, it's, it's rare to see that there's like more than about a six month difference between, like, the ideas travel back and forth with the people. So, so I think you sort of hit the hardware wall and. And then whichever company can scale hardware the fastest will be the leader. And so I think XAI will be able to scale hardware the fastest and therefore most likely will be the leader.

• 2:47:20Speaker 01

  You you you joked or, you know, I'm self-conscious about, you know, using the the limiting factor phrase again. But I actually think there's something deep here. And if you look at a lot of things we've touched on over the course of it, maybe kind of a good note to end on. Like, if you think of a senescent, lower agency company, it would have some bottleneck and not really be doing anything about it. You know, Mark Andreessen had the line of, most people are willing to endure any amount of chronic pain to avoid acute pain. And it feels like a lot of the cases we're talking about are just leaning into the acute pain, whatever it is.

• 2:48:00Speaker 01

  It's like, OK, we gotta figure out how to, you know, work with steel or we gotta figure out how to run the chips in space. Or like, we'll take some nearer term acute pain to actually solve the bottleneck. And so that's kind of a unifying thing.

• 2:48:13Speaker 00

  I have a high pain threshold.

• 2:48:16Speaker 00

  That's helpful.

• 2:48:17Speaker 01

  Solve the bottlenecks.

• 2:48:19Speaker 00

  Yes.

• 2:48:22Speaker 00

  So.

• 2:48:25Speaker 00

  You know, one thing I can say is like, I think the future is going to be very interesting.

• 2:48:34Speaker 00

  And as I said, the Dallos up, I was literally a Dallos, I think it was on the ground for like three hours or something.

• 2:48:44Speaker 00

  It's better to be, it's better to err on the side of optimism and be wrong than err on the side of pessimism and be right for quality of life. So, you know, your your your happiness will be, you'll be happier if you if you err on the side of optimism rather than erring on the side of pessimism. And so I recommend erring on the side of optimism.

• 2:49:08Speaker 01

  Nice to that.

• 2:49:09Dwarkesh Patel?

  Cool. Yala, thanks for doing this.

• 2:49:11Speaker 01

  Thank you.

• 2:49:11Speaker 00

  All right, thanks, guys.

• 2:49:14Speaker 01

  Oh, great stamina.

• 2:49:17Dwarkesh Patel?

  Hopefully this encounters the pain in the pain tolerance.

• 2:49:21Dwarkesh Patel?

  Hey, everybody. I hope you enjoyed that episode. If you did, the most helpful thing you can do is just share it with other people who you think might enjoy it. It's also helpful if you leave a rating or a comment on whatever platform you're listening on. If you're interested in sponsoring the podcast, you can reach out at dwarkesh.com/adandise. Otherwise, I'll see you on the next one.