1 00:00:00,040 --> 00:00:02,200 Speaker 1: Okay, I want you to close your rides for a second. Seriously, 2 00:00:02,279 --> 00:00:02,600 humor me. 3 00:00:02,759 --> 00:00:03,560 Speaker 2: Okay, I'm with you. 4 00:00:03,759 --> 00:00:06,519 Speaker 1: Imagine you are standing on the surface of the Moon. 5 00:00:07,400 --> 00:00:12,960 It's just dead silent. You're looking out over that gray, dusty, 6 00:00:13,080 --> 00:00:16,399 regless craters stretching out to the horizon. The Earth is 7 00:00:16,480 --> 00:00:19,839 just hanging there like a blue marble in this pitch 8 00:00:19,920 --> 00:00:20,519 black sky. 9 00:00:20,839 --> 00:00:24,440 Speaker 2: It's a striking image. It feels very cold, very industrial. 10 00:00:23,960 --> 00:00:27,440 Speaker 1: Right, and instead of just rocks and silence, you see 11 00:00:27,480 --> 00:00:33,320 a factory, a massive, silent industrial complex, humming with an 12 00:00:33,399 --> 00:00:35,719 energy you can almost feel through the soles of your boots, 13 00:00:36,520 --> 00:00:38,679 and stretching away from it, pointing up toward that black 14 00:00:38,719 --> 00:00:41,520 sky is a long silver track sort of like a 15 00:00:41,600 --> 00:00:44,840 high speed rail line, like a launch track exactly. And 16 00:00:44,880 --> 00:00:47,679 then suddenly, with absolutely no sound because you know, there's 17 00:00:47,679 --> 00:00:50,159 no air to carry it, right, total vacuum, no fire, 18 00:00:50,240 --> 00:00:52,719 no explosion, a satellite the size of a shipping container. 19 00:00:52,840 --> 00:00:55,920 Just it accelerates along that track. It's moving faster than 20 00:00:55,920 --> 00:00:57,520 a bullet. It hits the end of the line and 21 00:00:57,560 --> 00:01:01,759 it's just it vanishes into the heading forward of it. Wow, now, 22 00:01:01,960 --> 00:01:05,040 open your eyes. This isn't a scene from a sci 23 00:01:05,040 --> 00:01:07,760 fi novel set in the year like twenty three hundred. 24 00:01:08,280 --> 00:01:11,079 According to the roadmap we're unpacking today, this is a 25 00:01:11,120 --> 00:01:12,879 plan for the next four years. 26 00:01:13,319 --> 00:01:15,439 Speaker 2: That is the part that stops people in their tracks, 27 00:01:15,480 --> 00:01:18,319 isn't it. We're so used to hearing about future tech, 28 00:01:18,400 --> 00:01:21,159 but we always frame it as you know, decades. 29 00:01:20,760 --> 00:01:22,159 Speaker 1: Away, right, it's always someday. 30 00:01:22,280 --> 00:01:25,359 Speaker 2: We think of flying cars and moon bases as things 31 00:01:25,359 --> 00:01:28,239 our great grandchildren might see. But the roadmap we're looking 32 00:01:28,280 --> 00:01:31,239 at today suggests humanity is staring down the barrel of 33 00:01:31,239 --> 00:01:34,760 a transition that is so radical and so fast that 34 00:01:34,840 --> 00:01:37,079 we barely have the vocabulary to describe it. 35 00:01:37,680 --> 00:01:40,760 Speaker 1: Welcome to thrilling threads. Today we are pulling on a 36 00:01:40,799 --> 00:01:43,879 thread that starts with a single post on X by 37 00:01:43,920 --> 00:01:47,840 Elon Musk, analyzed by the sharp mind of researcher named Farzad, 38 00:01:48,200 --> 00:01:51,239 and ends with well, it ends with the civilization level shift. 39 00:01:51,359 --> 00:01:53,959 We're talking about the space based AI roadmap. 40 00:01:54,079 --> 00:01:55,959 Speaker 2: And just to set the stakes right at the top, 41 00:01:56,000 --> 00:01:59,079 because I think people hear space AI and they immediately think, oh, cool, 42 00:01:59,120 --> 00:02:01,079 faster internet from my phone or something like that. 43 00:02:01,159 --> 00:02:02,599 Speaker 1: Yeah, better Netflix stream on. 44 00:02:02,560 --> 00:02:05,359 Speaker 2: A plane exactly. But this is not that. This isn't 45 00:02:05,359 --> 00:02:09,719 about better reception. This is about a fundamental physics problem. 46 00:02:10,199 --> 00:02:13,479 The entire premise here is that artificial intelligence is about 47 00:02:13,520 --> 00:02:16,800 to leave the planet because Earth Earth literally cannot handle 48 00:02:16,840 --> 00:02:17,319 it anymore. 49 00:02:17,439 --> 00:02:18,159 Speaker 1: We've hit the ceiling. 50 00:02:18,479 --> 00:02:20,680 Speaker 2: We have absolutely hit the energy wall, We've hit the 51 00:02:20,680 --> 00:02:24,280 thermal wall. And the solution proposed in these documents involves 52 00:02:24,400 --> 00:02:27,840 first putting massive data centers in Sun synchronous orbits, and 53 00:02:27,960 --> 00:02:32,080 then well then using electromagnetic railguns on the Moon to 54 00:02:32,080 --> 00:02:32,680 put them there. 55 00:02:32,719 --> 00:02:35,599 Speaker 1: Just like you described, I still can't get over the railguns. 56 00:02:35,879 --> 00:02:37,479 I mean, we're going to get there, I promise, But 57 00:02:37,520 --> 00:02:40,120 before we get to the Moon. 58 00:02:39,280 --> 00:02:41,439 Speaker 2: There aren't weapons, to be clear, right right. 59 00:02:41,319 --> 00:02:44,560 Speaker 1: There, they're delivery trucks, but it sound like weapons. We 60 00:02:44,639 --> 00:02:46,879 have to eat our vegetables. First. We have to understand 61 00:02:46,919 --> 00:02:51,199 the why why would anyone go through the immense pain 62 00:02:51,319 --> 00:02:53,800 and cost of building a factory on the Moon. It 63 00:02:53,879 --> 00:02:56,719 just seems so exorbitantly expensive and. 64 00:02:56,759 --> 00:03:00,560 Speaker 2: Risky because we're running out of juice. It is basically 65 00:03:00,759 --> 00:03:02,960 that simple. If you look at the last couple of 66 00:03:03,000 --> 00:03:06,919 years of AI news, what has the headline always been, 67 00:03:07,039 --> 00:03:08,120 It's been about chips. 68 00:03:08,479 --> 00:03:10,919 Speaker 1: Oh yeah, do we have enough? H one hundreds. Is 69 00:03:10,960 --> 00:03:15,479 TSMC printing enough silicon is invidious stock going up again? 70 00:03:15,639 --> 00:03:18,719 Speaker 2: Exactly. It's all been about supply chain constraints on the 71 00:03:18,759 --> 00:03:21,240 silicon itself, the actual physical hardware. 72 00:03:21,319 --> 00:03:24,840 Speaker 1: Right, It's been a total supply chain panic. Everyone is 73 00:03:24,879 --> 00:03:28,039 hoarding GPUs like they were toilet paperback in twenty twenty. 74 00:03:28,120 --> 00:03:31,520 Speaker 2: You're not wrong, but that's yeah, that's kind of yesterday's problem, 75 00:03:31,719 --> 00:03:34,840 or at least it's a problem that's being solved. Factories 76 00:03:34,879 --> 00:03:38,319 are coming online, production is ramping up. The new problem, 77 00:03:38,439 --> 00:03:41,639 the hard problem that keeps utility CEOs awake at night 78 00:03:42,039 --> 00:03:45,719 is electricity. The grid cannot support the silicon we are 79 00:03:45,719 --> 00:03:46,599 now able to print. 80 00:03:46,719 --> 00:03:49,599 Speaker 1: That's the real bottleneck, not the chips, but the power 81 00:03:49,639 --> 00:03:50,280 to run them. 82 00:03:50,360 --> 00:03:50,919 Speaker 2: Precisely. 83 00:03:51,159 --> 00:03:53,840 Speaker 1: Varzod's analysis broke down some numbers on the US power 84 00:03:53,840 --> 00:03:57,240 grid that I found genuinely staggering. I think most people, 85 00:03:57,240 --> 00:03:59,319 myself included. You just you plug your phone in and 86 00:03:59,400 --> 00:04:02,199 you just assume the wall has infinite power. You don't 87 00:04:02,199 --> 00:04:03,479 really think about where it comes from. 88 00:04:03,520 --> 00:04:05,759 Speaker 2: And it's not infinite, not even close. So let's look 89 00:04:05,759 --> 00:04:09,360 at the reality. The United States in total generates approximately 90 00:04:09,520 --> 00:04:11,520 four hundred and sixty gigawatts of power. 91 00:04:11,599 --> 00:04:12,560 Speaker 1: Four hundred and sixtys. 92 00:04:12,599 --> 00:04:15,400 Speaker 2: That is the whole pie. That number accounts for every 93 00:04:15,400 --> 00:04:18,319 single house in the suburbs, every steel mill in the 94 00:04:18,399 --> 00:04:20,319 Rust bell, every street light in New York. 95 00:04:20,199 --> 00:04:22,680 Speaker 1: City, every hospital, every EV charger. 96 00:04:22,399 --> 00:04:25,439 Speaker 2: Every single toaster in the country. Everything, four hundred and 97 00:04:25,480 --> 00:04:26,279 sixty gigawatts. 98 00:04:26,319 --> 00:04:28,639 Speaker 1: Okay, so four hundred and sixty is the cap. That's 99 00:04:28,680 --> 00:04:30,680 everything we've managed to build over the last one hundred 100 00:04:30,720 --> 00:04:31,120 years or so. 101 00:04:31,319 --> 00:04:33,800 Speaker 2: That's our baseline. Now, look at the demand that's coming 102 00:04:33,800 --> 00:04:36,120 down the pipeline just from the hyperscalers. You know, the 103 00:04:36,160 --> 00:04:37,759 big tech giants. 104 00:04:37,399 --> 00:04:40,480 Speaker 1: Google, Microsoft, Open AI, Amazon, the. 105 00:04:40,399 --> 00:04:43,360 Speaker 2: Big players, right, they all have these road maps to 106 00:04:43,439 --> 00:04:46,639 scale up their AI infrastructure ultimately to try and reach 107 00:04:46,720 --> 00:04:50,240 AGI artificial general intelligence. To do that, they want to 108 00:04:50,240 --> 00:04:52,040 add another one hundred gigawatts of demand. 109 00:04:52,120 --> 00:04:53,800 Speaker 1: Wait, wait, one hundred. 110 00:04:53,639 --> 00:04:56,279 Speaker 2: One hundred gigawatts just for data centers. 111 00:04:56,439 --> 00:04:59,399 Speaker 1: So hang on, you're saying they want to add demand 112 00:04:59,480 --> 00:05:02,160 that is on almost a quarter of the entire existing grid. 113 00:05:02,360 --> 00:05:06,120 Speaker 2: Correct, narrowly twenty five percent more demand just for AI. 114 00:05:06,560 --> 00:05:08,800 And here's the friction point, the part where the plan 115 00:05:08,920 --> 00:05:12,000 hits a wall. You cannot just snap your fingers and 116 00:05:12,000 --> 00:05:14,480 build twenty five percent more power plants in two years. 117 00:05:14,959 --> 00:05:17,839 It is physically and bureaucratically impossible. 118 00:05:18,120 --> 00:05:20,759 Speaker 1: I mean, it takes a decade just to get the 119 00:05:20,800 --> 00:05:23,399 permits for a new nuclear plant, right, let alone pour 120 00:05:23,439 --> 00:05:25,439 the concrete and actually build the thing exactly. 121 00:05:25,480 --> 00:05:28,399 Speaker 2: Look at the votical nuclear plant in Georgia. That took 122 00:05:28,680 --> 00:05:32,079 way way longer and costs billions more than expected. And 123 00:05:32,120 --> 00:05:33,399 that's just one plant. 124 00:05:33,439 --> 00:05:36,360 Speaker 1: And it's not just nuclear. Wind farms face all that 125 00:05:36,439 --> 00:05:39,800 local opposition. People don't want them ruining the view the 126 00:05:39,839 --> 00:05:41,199 whole NMB thing. 127 00:05:40,800 --> 00:05:44,199 Speaker 2: Sure not in my backyard. And solar, which is great, 128 00:05:44,439 --> 00:05:48,199 needs massive battery storage to be what's called baseload reliable, 129 00:05:48,319 --> 00:05:50,399 meaning it works when the sun isn't shining, and that 130 00:05:50,519 --> 00:05:51,759 adds huge costs. 131 00:05:51,839 --> 00:05:55,240 Speaker 1: And what about the wires themselves, the transmission. 132 00:05:54,639 --> 00:05:59,079 Speaker 2: Lines another bottleneck. Upgrading the actual copper wires to carry 133 00:05:59,120 --> 00:06:01,639 that much more power or can take years and years 134 00:06:01,680 --> 00:06:05,279 of planning and construction. So you have this immovable object 135 00:06:05,319 --> 00:06:10,000 of infrastructure reality meeting this unstoppable force of AI demand. 136 00:06:10,160 --> 00:06:12,079 Speaker 1: And this isn't happening in a vacuum. It's not like 137 00:06:12,120 --> 00:06:14,759 that AI demand is the only thing asking for more power, 138 00:06:14,879 --> 00:06:18,360 not at all. It's competing with civilization's most basic needs. 139 00:06:18,720 --> 00:06:20,600 I mean, if you plug in a new data center 140 00:06:20,639 --> 00:06:22,839 in northern Virginia that sucks up as much power as 141 00:06:22,839 --> 00:06:25,560 a small city, and it's the middle of January and 142 00:06:25,600 --> 00:06:29,519 a polar vortex hits, people need to heat their homes, and. 143 00:06:29,399 --> 00:06:31,920 Speaker 2: The grid operator have to make a choice. It's a 144 00:06:32,000 --> 00:06:35,040 zero sum game. Do we cut power to the data 145 00:06:35,040 --> 00:06:37,680 center that's training the next chatbot, or do we let 146 00:06:37,720 --> 00:06:39,759 the grandmother and the suburbs freeze? 147 00:06:39,959 --> 00:06:42,720 Speaker 1: In a democracy, the data center loses that fight every 148 00:06:42,759 --> 00:06:43,959 single time, and it. 149 00:06:43,920 --> 00:06:47,160 Speaker 2: Should and rightly so. So the AI companies are looking 150 00:06:47,160 --> 00:06:50,240 at this map, at this reality, and they're realizing we 151 00:06:50,319 --> 00:06:53,120 can't scale here. It's not a question of money anymore. 152 00:06:53,120 --> 00:06:54,879 It's a question of physics and politics. 153 00:06:54,959 --> 00:06:57,800 Speaker 1: We are physically unable to build power plans fast enough 154 00:06:57,800 --> 00:06:59,959 to keep up with the exponential growth of these AI 155 00:07:00,120 --> 00:07:00,759 training runs. 156 00:07:00,920 --> 00:07:04,000 Speaker 2: But there is another bottleneck, and this one is arguably 157 00:07:04,040 --> 00:07:08,000 even worse than the power generation. This one is pure thermodynamics. 158 00:07:08,240 --> 00:07:08,639 Speaker 1: The heat. 159 00:07:09,000 --> 00:07:12,680 Speaker 2: The heat thermodynamics is a cruel mistress when you run 160 00:07:12,720 --> 00:07:15,600 electricity through a computer chip to do math, which is 161 00:07:15,639 --> 00:07:19,639 all AI really is just massive, massive matrix multiplication. That 162 00:07:19,759 --> 00:07:21,319 electricity encounters. 163 00:07:20,920 --> 00:07:25,720 Speaker 1: Resistance and resistance creates heat. Basic physics, basic physics. 164 00:07:25,759 --> 00:07:28,439 Speaker 2: And if you don't remove that heat, and I mean immediately, 165 00:07:28,600 --> 00:07:31,040 the chip melts, It literally destroys itself. 166 00:07:31,079 --> 00:07:34,319 Speaker 1: There was a stat from Jensen Hwan, the CEO of Nvidia, 167 00:07:34,360 --> 00:07:36,759 that was mentioned in our source material, and this one, 168 00:07:37,120 --> 00:07:38,959 I mean I had to read it three times. It 169 00:07:39,000 --> 00:07:41,240 made me laugh out of sheer disbelief. It sounds like 170 00:07:41,279 --> 00:07:41,680 a joke. 171 00:07:41,959 --> 00:07:43,759 Speaker 2: It is a wild statistic, I know, the one you're 172 00:07:43,759 --> 00:07:46,199 talking about. He said that for every server rac that 173 00:07:46,240 --> 00:07:49,600 weighs two tons, So think of a big, heavy metal 174 00:07:49,600 --> 00:07:51,519 cabinet packed with expensive chips. 175 00:07:51,600 --> 00:07:54,240 Speaker 1: Okay, two tons like a car roughly. 176 00:07:54,399 --> 00:07:56,959 Speaker 2: Yeah, So for that two tons server rack, he said, 177 00:07:56,959 --> 00:07:59,199 about one point nine to five tons of that weight 178 00:07:59,360 --> 00:08:00,680 is just cool equipment. 179 00:08:00,720 --> 00:08:01,959 Speaker 1: That's ninety seven percent. 180 00:08:02,120 --> 00:08:03,240 Speaker 2: Ninety seven percent. 181 00:08:03,360 --> 00:08:06,639 Speaker 1: So you're telling me that only three percent of this giant, heavy, 182 00:08:06,759 --> 00:08:10,800 expensive machine is actually doing the thinking. The other ninety 183 00:08:10,800 --> 00:08:12,920 seven percent is just there to keep the three percent 184 00:08:12,959 --> 00:08:14,000 from catching on fire. 185 00:08:14,480 --> 00:08:18,560 Speaker 2: That is the absurd reality of high performance computing on Earth. 186 00:08:19,000 --> 00:08:21,920 Think about the inefficiency of that. You're building these massive 187 00:08:21,959 --> 00:08:25,439 concrete buildings. You're filling them with water pipes, huge air 188 00:08:25,480 --> 00:08:30,639 conditioning units, chillers, fans, heat exchangers, just to run a 189 00:08:30,639 --> 00:08:31,759 little slice of silicon. 190 00:08:31,839 --> 00:08:33,840 Speaker 1: It's like buying a semi truck just to transport a 191 00:08:33,879 --> 00:08:34,679 single envelope. 192 00:08:34,799 --> 00:08:37,679 Speaker 2: That is a perfect analogy. And on Earth we are 193 00:08:37,759 --> 00:08:41,799 constantly fighting the environment. We have ambient heat. If it's 194 00:08:41,879 --> 00:08:44,840 ninety five degrees outside in Texas, your AC has to 195 00:08:44,879 --> 00:08:46,960 work that much harder just to keep the chips that 196 00:08:47,000 --> 00:08:48,080 they're operating temperature. 197 00:08:48,159 --> 00:08:50,440 Speaker 1: You have a humidity, you have dust, all of it. 198 00:08:50,440 --> 00:08:53,679 Speaker 2: It's a constant battle. So we've hit this hard physical ceiling. 199 00:08:54,159 --> 00:08:56,679 We're running out of power to feed the AI, and 200 00:08:56,759 --> 00:08:59,320 we're running out of water and space and equipment to 201 00:08:59,320 --> 00:08:59,960 cool the AI. 202 00:09:00,399 --> 00:09:02,279 Speaker 1: So the premise of this whole thread today is that 203 00:09:02,639 --> 00:09:05,480 Muskiness team have looked at this game board, realized all 204 00:09:05,519 --> 00:09:07,919 the pieces are completely locked, and they've just said, Okay, 205 00:09:08,399 --> 00:09:10,399 we can't win this game on Earth, so let's change 206 00:09:10,399 --> 00:09:10,759 the board. 207 00:09:10,919 --> 00:09:13,759 Speaker 2: And that's where we enter phase one. The orbital solution. 208 00:09:14,240 --> 00:09:17,440 Speaker 1: The timeline for this and the roadmap is it's just 209 00:09:17,480 --> 00:09:22,279 so aggressive. The source suggests this is targeted for within 210 00:09:22,360 --> 00:09:25,759 three years. Three years. My shoes are older than three years. 211 00:09:26,679 --> 00:09:28,159 What exactly is phase one? 212 00:09:28,399 --> 00:09:32,240 Speaker 2: The concept is pretty straightforward at a high level. You 213 00:09:32,519 --> 00:09:38,279 launch the AI, compute the actual servers into space, specifically 214 00:09:38,360 --> 00:09:40,639 into what are called sun synchronous orbits. 215 00:09:40,679 --> 00:09:42,799 Speaker 1: Okay, break that down for us. I know orbit means 216 00:09:42,840 --> 00:09:45,600 it's going around the Earth, But what makes it sun synchronous? 217 00:09:45,679 --> 00:09:47,639 Is there just a fancy term for being really high up? 218 00:09:47,879 --> 00:09:50,080 Speaker 2: No, it's not about altitude so much as it is 219 00:09:50,120 --> 00:09:53,200 about the angle and the timing. See, a normal satellite 220 00:09:53,240 --> 00:09:56,039 like the International Space Station goes around the Earth and 221 00:09:56,039 --> 00:09:58,679 spends about half its time in direct sunlight and the 222 00:09:58,720 --> 00:10:00,799 other half in the Earth's night time. 223 00:10:01,000 --> 00:10:02,360 Speaker 1: And that's bad for solar power. 224 00:10:02,639 --> 00:10:05,440 Speaker 2: It's terrible for solar power because it means you need huge, 225 00:10:05,559 --> 00:10:07,240 heavy batteries to survive the night. 226 00:10:07,480 --> 00:10:10,320 Speaker 1: Batteries are heavy and expensive, and heavy things cost a 227 00:10:10,320 --> 00:10:11,720 lot of money to launch into space. 228 00:10:12,279 --> 00:10:15,120 Speaker 2: You've got it, that's the core problem. But if you 229 00:10:15,200 --> 00:10:18,440 launch into a very specific inclination, it's almost a pole 230 00:10:18,480 --> 00:10:22,559 to pole orbit. You can effectively ride the terminator line. 231 00:10:22,559 --> 00:10:25,080 Speaker 1: The terminator line, yeah, the line between day and night 232 00:10:25,120 --> 00:10:25,519 on Earth. 233 00:10:25,559 --> 00:10:26,159 Speaker 2: That's the one. 234 00:10:26,200 --> 00:10:29,639 Speaker 1: So you're basically surfing the sunset. Yeah, forever. 235 00:10:29,799 --> 00:10:32,399 Speaker 2: That is a fantastic way to put it. Yes, the 236 00:10:32,440 --> 00:10:35,519 satellite never goes into the shadow because the Earth is 237 00:10:35,639 --> 00:10:38,279 rotating underneath it, and the satellite is orbiting at a 238 00:10:38,360 --> 00:10:41,519 rate that perfectly matches that rotation relative to the sun. 239 00:10:41,919 --> 00:10:44,240 It is bathed in sunlight twenty four hours a day, 240 00:10:44,279 --> 00:10:45,279 seven days a week. 241 00:10:45,159 --> 00:10:48,120 Speaker 1: No clouds, no night, no rain, no weather. 242 00:10:47,919 --> 00:10:51,120 Speaker 2: At all, and crucially no atmosphere to filter the light. 243 00:10:51,519 --> 00:10:54,039 The Sun pumps out more energy in a single second 244 00:10:54,120 --> 00:10:57,840 than humanity has used in its entire history. On Earth, 245 00:10:57,919 --> 00:11:00,559 we catch a tiny, featured fraction of that after it 246 00:11:00,600 --> 00:11:02,759 pushes through the ozone layer of the clouds and all 247 00:11:02,799 --> 00:11:05,879 the air. But in space, in space it's an energy buffet. 248 00:11:05,960 --> 00:11:09,799 It is for all practical purposes infinite clean power. 249 00:11:10,159 --> 00:11:13,399 Speaker 1: So that just it completely solves the power problem, the 250 00:11:13,519 --> 00:11:17,600 hundred gigawatt problem. Solar panels in space are vastly more efficient, 251 00:11:17,639 --> 00:11:20,159 and they run constantly. You don't need the batteries. But 252 00:11:20,720 --> 00:11:23,759 what about the cooling Ah, Yes, you said earlier that 253 00:11:23,840 --> 00:11:26,000 ninety seven percent of a server rack on Earth is 254 00:11:26,120 --> 00:11:28,960 cooling gear because you have to use water and air 255 00:11:29,000 --> 00:11:32,440 to physically move the heat away. Space is a vacuum, 256 00:11:32,720 --> 00:11:35,440 doesn't that make cooling harder? I remember from a physics 257 00:11:35,440 --> 00:11:37,879 class that a vacuum is a great insulator, like a 258 00:11:37,879 --> 00:11:38,639 thermos flask. 259 00:11:38,840 --> 00:11:39,320 Speaker 2: That's right. 260 00:11:39,519 --> 00:11:41,679 Speaker 1: If I put hot coffee in a vacuum flask, it 261 00:11:41,720 --> 00:11:45,679 stays hot for hours. Won't the chips just melt even faster? 262 00:11:45,919 --> 00:11:48,519 Speaker 2: You're spot on. And this is where it gets really counterintuitive, 263 00:11:48,559 --> 00:11:52,200 but also incredibly elegant. A vacuum is a perfect insulator 264 00:11:52,279 --> 00:11:55,120 against conduction or convection. You can't use a fan in 265 00:11:55,200 --> 00:11:57,440 space because there is no air to blow, okay, but 266 00:11:57,679 --> 00:12:01,200 space allows for something called radiative cool and it's incredibly 267 00:12:01,240 --> 00:12:02,799 effective radiative cooling. 268 00:12:02,919 --> 00:12:05,279 Speaker 1: Unpack that for the non physicists among us. 269 00:12:05,679 --> 00:12:08,879 Speaker 2: All objects that are hot emit infrared radiation. They glow 270 00:12:08,919 --> 00:12:10,759 with heat even if our eyes can't see it. It's 271 00:12:10,799 --> 00:12:12,480 how those thermal cameras work, right. 272 00:12:12,559 --> 00:12:12,879 Speaker 1: Okay. 273 00:12:12,960 --> 00:12:16,840 Speaker 2: On Earth, radiating heat is very slow and inefficient because 274 00:12:17,000 --> 00:12:20,480 the air around the object is also warm and radiates back. 275 00:12:20,799 --> 00:12:23,120 So we force air or water over the chip to 276 00:12:23,120 --> 00:12:26,440 strip the heat away physically. But in space you have 277 00:12:26,519 --> 00:12:30,159 the ultimate heat sink, the deep dark void. 278 00:12:30,320 --> 00:12:30,919 Speaker 1: How cold do that? 279 00:12:31,360 --> 00:12:35,639 Speaker 2: It's roughly three kelvin, which is minus two hundred and 280 00:12:35,679 --> 00:12:38,759 seventy degrees celsius minus four fifty five fahrenheit. 281 00:12:38,960 --> 00:12:41,759 Speaker 1: That is cold. That is unfathomably cold. 282 00:12:41,879 --> 00:12:45,200 Speaker 2: It is effectively absolute zero. So if you attach your 283 00:12:45,399 --> 00:12:49,000 hot computer chips to large radiator panels, just big flat 284 00:12:49,080 --> 00:12:51,679 sheets of metal, and point those panels away from the 285 00:12:51,720 --> 00:12:54,600 sun and toward deep space, the heat just leaves. It 286 00:12:54,720 --> 00:12:56,799 radiates away as light energy into the nothingness. 287 00:12:56,840 --> 00:12:59,159 Speaker 1: You don't need the pumps. No pumps, no compressors, no 288 00:12:59,279 --> 00:13:00,600 water loops, no fans. 289 00:13:00,639 --> 00:13:01,320 Speaker 2: I think zero. 290 00:13:01,720 --> 00:13:03,840 Speaker 1: So that one point nine to five tons of cooling 291 00:13:03,879 --> 00:13:05,120 equipment we talked about. 292 00:13:04,960 --> 00:13:07,120 Speaker 2: Gone, You delete it. You just you don't build it. 293 00:13:07,200 --> 00:13:09,600 You replace it with a passive metal sheet. You can 294 00:13:09,720 --> 00:13:12,399 strip the hardware down to just the essential components, the 295 00:13:12,440 --> 00:13:14,480 compute chips, the solar panels, and. 296 00:13:14,440 --> 00:13:17,120 Speaker 1: The radiator, which has to drastically reduce the cost and 297 00:13:17,159 --> 00:13:19,799 the complexity of the hardware itself massively, and. 298 00:13:19,759 --> 00:13:22,360 Speaker 2: There's zero maintenance. There are no moving parts for the 299 00:13:22,360 --> 00:13:26,039 cooling system that can break down. It's just elegant engineering. 300 00:13:26,159 --> 00:13:28,759 Speaker 1: Okay, that sounds like a dream for an engineer. Yeah, 301 00:13:28,799 --> 00:13:31,440 but getting stuff to stace is expensive. Or at least 302 00:13:31,440 --> 00:13:33,080 it always used to be. I mean, if I want 303 00:13:33,080 --> 00:13:35,279 to ship a ton of gold to orbit, it used 304 00:13:35,320 --> 00:13:37,159 to cost way more than the gold was worth. 305 00:13:37,399 --> 00:13:40,720 Speaker 2: This is the lynchpin. This is the piece where without it, 306 00:13:40,919 --> 00:13:45,200 this entire plan is a hallucination. It's a PowerPoint presentation 307 00:13:45,320 --> 00:13:49,159 that goes nowhere. And that piece is the Starship rocket. 308 00:13:49,720 --> 00:13:52,679 With Starship, this becomes a logistics problem and not a 309 00:13:52,679 --> 00:13:53,559 physics problem. 310 00:13:53,600 --> 00:13:57,120 Speaker 1: The numbers mentioned in Farzod's analysis were about mega tons 311 00:13:57,679 --> 00:13:59,039 a mega ton of satellites. 312 00:13:59,360 --> 00:14:02,480 Speaker 2: The goal, as stated in the roadmap, is to launch 313 00:14:02,600 --> 00:14:06,320 one mega ton of satellites per year a million kilograms. 314 00:14:06,399 --> 00:14:08,279 Speaker 1: Yeah, a thousand metric. 315 00:14:08,039 --> 00:14:10,919 Speaker 2: Tons, which sounds immense, and historically it is immense. But 316 00:14:10,960 --> 00:14:12,440 you have to look at the new tool we have. 317 00:14:12,519 --> 00:14:14,360 You have to look at the lift capacity of. 318 00:14:14,320 --> 00:14:16,399 Speaker 1: Starship, which is designed for what exactly. 319 00:14:16,759 --> 00:14:19,320 Speaker 2: It's designed to carry one hundred to one hundred and 320 00:14:19,360 --> 00:14:22,960 fifty tons per flight to low Earth orbit. So do 321 00:14:23,080 --> 00:14:26,559 the map to hit that megaton? That thousand tons heart 322 00:14:26,559 --> 00:14:29,799 you're talking about roughly what seven to ten launches per 323 00:14:29,879 --> 00:14:30,240 year are. 324 00:14:30,159 --> 00:14:33,480 Speaker 1: Think seven to ten launches. That's that's nothing. 325 00:14:33,600 --> 00:14:35,240 Speaker 2: It sounds like nothing, right, I. 326 00:14:35,200 --> 00:14:38,440 Speaker 1: Mean SpaceX is currently launching their Falcon nine rockets almost 327 00:14:38,480 --> 00:14:41,080 every other day. They did something like ninety launches last year. 328 00:14:41,279 --> 00:14:43,639 Seven launches is a slow Tuesday for them. 329 00:14:43,679 --> 00:14:47,240 Speaker 2: Precisely, that is why this is suddenly feasible. We aren't 330 00:14:47,279 --> 00:14:50,399 talking about needing one thousand single use rockets. We are 331 00:14:50,440 --> 00:14:53,399 talking about a handful of launches from a vehicle Starship, 332 00:14:53,679 --> 00:14:56,480 that is designed from the ground up to be rapidly reusable. 333 00:14:56,720 --> 00:14:58,960 Land it, refill it, and launch it again, maybe even 334 00:14:58,960 --> 00:14:59,519 the same day. 335 00:15:00,080 --> 00:15:03,240 Speaker 1: And we are already seeing Starship flight tests. We've seen 336 00:15:03,240 --> 00:15:05,600 them land, We've seen them catch the booster with the 337 00:15:05,600 --> 00:15:08,919 giant chopsticks at the launch tower. We know the hardware exists. 338 00:15:09,159 --> 00:15:10,399 This is not a paper rocket. 339 00:15:10,480 --> 00:15:11,840 Speaker 2: It's not imaginary tech now. 340 00:15:12,000 --> 00:15:15,679 Speaker 1: So from a pure execution standpoint, putting massive amounts of 341 00:15:15,679 --> 00:15:20,399 compute into orbit is surprisingly doable. It sounds completely crazy 342 00:15:20,440 --> 00:15:22,480 on the face of it, but when you break it down, 343 00:15:22,919 --> 00:15:24,159 the MAK actually works. 344 00:15:24,440 --> 00:15:27,480 Speaker 2: It's the principle of economies of scale being applied to 345 00:15:27,600 --> 00:15:30,080 orbit for the first time. When the cost to launch 346 00:15:30,120 --> 00:15:32,720 a kilogram of mass drops by a factor of one hundred, 347 00:15:33,279 --> 00:15:36,799 the entire business case for space manufacturing and infrastructure just 348 00:15:37,000 --> 00:15:39,279 great round it flips on its head. 349 00:15:39,360 --> 00:15:42,159 Speaker 1: Okay. So we have these giant server farms floating in space. 350 00:15:42,200 --> 00:15:44,240 They're powered two hundred and forty seven by the sun. 351 00:15:44,240 --> 00:15:47,600 They're passively cool by the void. But they need to 352 00:15:47,639 --> 00:15:50,320 talk to us on the ground. A supercomputer in space 353 00:15:50,440 --> 00:15:52,919 is totally useless if I can't ask it to write 354 00:15:52,960 --> 00:15:56,039 my emails or solve my protein folding simulation. 355 00:15:56,320 --> 00:15:56,480 Speaker 2: Right. 356 00:15:56,720 --> 00:15:59,759 Speaker 1: If the Internet connection is as slow as bad hotel WiFi, 357 00:16:00,240 --> 00:16:01,879 doesn't matter how smart the AI is. 358 00:16:02,039 --> 00:16:04,799 Speaker 2: Connectivity is the final piece of this Phase one puzzle, 359 00:16:04,840 --> 00:16:07,080 and the beautiful thing is the infrastructure for it is 360 00:16:07,120 --> 00:16:08,159 practically already there. 361 00:16:08,200 --> 00:16:11,279 Speaker 1: It's called starlink, right, the thousands of Internet satellites that 362 00:16:11,320 --> 00:16:13,600 SpaceX has already put up there exactly. 363 00:16:13,879 --> 00:16:16,360 Speaker 2: The Starlink satellites don't just talk to the ground. They 364 00:16:16,360 --> 00:16:18,720 talk to each other using optical laser links. It's a 365 00:16:18,799 --> 00:16:21,240 mesh network in the sky. So the plan is to 366 00:16:21,360 --> 00:16:25,600 integrate these new AI compute satellites directly into that existing 367 00:16:25,799 --> 00:16:26,679 laser network. 368 00:16:26,960 --> 00:16:28,639 Speaker 1: So walk me through it. If I'm on my phone 369 00:16:28,679 --> 00:16:32,799 in New York and I ask a really complex AI query, Okay. 370 00:16:32,679 --> 00:16:36,360 Speaker 2: So your request goes up to the nearest starlink satellite 371 00:16:36,399 --> 00:16:40,399 above you. That satellite then laser beams your query over 372 00:16:40,480 --> 00:16:43,679 to an AI compute cluster that's also in orbit, probably 373 00:16:43,759 --> 00:16:46,799 in that Sunsein corporate nearby. The AI does the thinking, 374 00:16:47,000 --> 00:16:49,720 that's the energy intensive part, and then it beams the 375 00:16:49,759 --> 00:16:52,480 answer back down to you through the starlink network. 376 00:16:52,639 --> 00:16:55,039 Speaker 1: And because this is all happening in lowerth orbit, we're 377 00:16:55,039 --> 00:16:57,799 only talking about a distance of a few hundred kilometers. 378 00:16:57,440 --> 00:17:01,080 Speaker 2: The latency, the delay is negligible. It's actually comparable to 379 00:17:01,159 --> 00:17:03,960 using fiber optic cables on Earth. And here's a fun fact. 380 00:17:04,519 --> 00:17:07,559 Light travels roughly forty percent faster in the vacuum of 381 00:17:07,599 --> 00:17:10,079 space than it does through glass fiber cables. 382 00:17:10,119 --> 00:17:11,880 Speaker 1: Wait, really, the glass slows it down. 383 00:17:12,119 --> 00:17:18,359 Speaker 2: Yes, the refractive index of glass slows light down. Vacuum doesn't. 384 00:17:19,240 --> 00:17:22,559 So over very long distances, say from New York to Tokyo, 385 00:17:23,079 --> 00:17:26,400 a space based internet connection can actually be faster than 386 00:17:26,440 --> 00:17:27,480 a ground based one. 387 00:17:27,519 --> 00:17:30,359 Speaker 1: That's amazing. So you're not sending the raw data back 388 00:17:30,400 --> 00:17:32,559 and forth, like you're not trying to upload the entire 389 00:17:32,640 --> 00:17:34,599 Library of Congress to space and back. 390 00:17:34,759 --> 00:17:36,599 Speaker 2: No, that would be slow and expensive. 391 00:17:36,680 --> 00:17:38,559 Speaker 1: You're just sending the question up and getting the answer 392 00:17:38,640 --> 00:17:39,519 down exactly. 393 00:17:39,559 --> 00:17:43,160 Speaker 2: It's a highly efficient architecture. You keep all the heavy lifting, 394 00:17:43,359 --> 00:17:46,519 all the energy consumption in orbit where the energy is 395 00:17:46,559 --> 00:17:49,480 free and abundant, and you just beam the results down 396 00:17:49,519 --> 00:17:50,079 to the user. 397 00:17:50,240 --> 00:17:53,640 Speaker 1: So that's Phase one. And honestly, if the roadmap just 398 00:17:53,680 --> 00:17:57,119 stopped right there, I would still be incredibly impressed. Data 399 00:17:57,119 --> 00:18:00,079 centers in the sky. That's a cool Wired magazine headline. 400 00:18:00,160 --> 00:18:01,640 It solves the earth bound bottleneck. 401 00:18:01,759 --> 00:18:03,359 Speaker 2: It's a huge step forward on its own. 402 00:18:03,519 --> 00:18:06,000 Speaker 1: But the roadmap does not stop there. It goes to 403 00:18:06,000 --> 00:18:08,240 phase two. And this is where I really need you 404 00:18:08,279 --> 00:18:10,640 to walk me through the physics, because this part sounds 405 00:18:10,680 --> 00:18:11,759 like complete madness. 406 00:18:11,799 --> 00:18:15,279 Speaker 2: Phase two Moon factories and mass drivers. 407 00:18:15,200 --> 00:18:18,559 Speaker 1: And the timeline is targeted to begin within four years. 408 00:18:18,839 --> 00:18:21,200 Four years. Now. I have to play the skeptic here 409 00:18:21,240 --> 00:18:23,640 for a minute. NASA had been trying to get people 410 00:18:23,680 --> 00:18:25,519 back to the Moon with the art of his program 411 00:18:25,599 --> 00:18:30,440 for a decade. Timelines in aerospace are notorious for slipping. 412 00:18:30,440 --> 00:18:33,200 How can anyone seriously talk about building factories on the 413 00:18:33,200 --> 00:18:34,119 Moon in four years? 414 00:18:34,359 --> 00:18:38,319 Speaker 2: It is an incredibly unbelievably aggressive timeline. Let's just call 415 00:18:38,359 --> 00:18:39,160 it elon time. 416 00:18:39,559 --> 00:18:42,160 Speaker 1: We all know the drill elon time. 417 00:18:42,559 --> 00:18:45,480 Speaker 2: So maybe four years means eight, maybe means ten. But 418 00:18:45,519 --> 00:18:47,720 the key thing is the logic of the plan holds 419 00:18:47,799 --> 00:18:50,759 up regardless of the calendar date. The question is why 420 00:18:50,880 --> 00:18:51,880 build on the Moon. 421 00:18:51,920 --> 00:18:54,880 Speaker 1: And I assume the answer is because hauling everything up 422 00:18:54,920 --> 00:18:57,400 from Earth is still a massive pain, even with a 423 00:18:57,480 --> 00:18:59,400 rocket as big as starship Exactly. 424 00:19:00,039 --> 00:19:02,519 Speaker 2: Earth is what we call a deep gravity well, think 425 00:19:02,519 --> 00:19:05,079 of it like a deep pit. To get anything out 426 00:19:05,119 --> 00:19:07,279 of that pit and into orbit, you have to achieve 427 00:19:07,319 --> 00:19:10,000 an escape velocity of about eleven kilometers per. 428 00:19:09,920 --> 00:19:12,119 Speaker 1: Second, which is what mock thirty three. 429 00:19:12,279 --> 00:19:16,880 Speaker 2: Something like that. It's incredibly fast, and it requires massive 430 00:19:16,960 --> 00:19:20,240 chemical rockets, huge fuel tanks, and a ton of money. 431 00:19:20,680 --> 00:19:23,440 You are fighting the planet's gravity every single inch of 432 00:19:23,440 --> 00:19:23,799 the way. 433 00:19:24,160 --> 00:19:26,480 Speaker 1: The old saying is you have to burn ninety percent 434 00:19:26,519 --> 00:19:29,119 of your fuel just to lift the other ten percent 435 00:19:29,160 --> 00:19:29,680 of your fuel. 436 00:19:29,720 --> 00:19:31,839 Speaker 2: That's the tyranny of the rocket equation in a nutshell. 437 00:19:32,039 --> 00:19:34,680 But the Moon, the Moon is different. It's tiny. It 438 00:19:34,720 --> 00:19:38,279 has only one sixth of Earth's gravity. The escape velocity 439 00:19:38,279 --> 00:19:40,480 from the lunar surface is only about two point four 440 00:19:40,559 --> 00:19:41,640 kilometers per second. 441 00:19:41,720 --> 00:19:43,599 Speaker 1: It's a much shallower pit to climb out of. 442 00:19:43,759 --> 00:19:47,039 Speaker 2: It's barely a ditch. And that fundamental difference in physics 443 00:19:47,119 --> 00:19:50,359 is what enables the next step, the rail gun. 444 00:19:50,319 --> 00:19:52,759 Speaker 1: Which is technically called a mass driver, right. 445 00:19:52,720 --> 00:19:55,400 Speaker 2: Right, And the concept is actually quite simple. You build 446 00:19:55,440 --> 00:19:57,680 a long track on the lunar surface, maybe a few 447 00:19:57,759 --> 00:20:01,759 kilometers long. You line that track with powerful electromagnets basically 448 00:20:01,799 --> 00:20:05,559 linear motors. Okay, You place your payload, say a satellite 449 00:20:05,599 --> 00:20:07,960 casing made of aluminum that you mind and process right 450 00:20:07,960 --> 00:20:10,640 there on the Moon, onto a magnetic sled, and then 451 00:20:10,640 --> 00:20:13,599 you just you pulse the magnets in sequence like a. 452 00:20:13,599 --> 00:20:15,440 Speaker 1: Bullet train, but way, way faster. 453 00:20:15,599 --> 00:20:18,519 Speaker 2: Much much faster. You accelerate the object along the track 454 00:20:18,599 --> 00:20:21,039 until it hits that lunar uscape velocity of two point 455 00:20:21,039 --> 00:20:24,880 four kilometers per second. And here is the absolutely crucial part, 456 00:20:25,359 --> 00:20:28,400 because there is essentially no atmosphere on the Moon. There 457 00:20:28,440 --> 00:20:30,039 is no air resistance. 458 00:20:29,599 --> 00:20:31,759 Speaker 1: No drag, nothing to slow it down. 459 00:20:31,880 --> 00:20:34,559 Speaker 2: Nothing. On Earth, if you try to launch something at 460 00:20:34,599 --> 00:20:36,440 mock seven at sea level, it would just burn up 461 00:20:36,519 --> 00:20:39,319 instantly like a meteor. The air is too thick. But 462 00:20:39,480 --> 00:20:42,519 on the Moon you can accelerate it to incredible speeds 463 00:20:42,599 --> 00:20:43,680 right there on the ground and. 464 00:20:43,680 --> 00:20:45,759 Speaker 1: At the end of the track. What happens. 465 00:20:45,880 --> 00:20:49,880 Speaker 2: It just flies off, It coasts into orbit. No rocket engine, 466 00:20:50,319 --> 00:20:55,039 no burning fuel, no explosion, just pure kinetic energy that 467 00:20:55,079 --> 00:20:56,519 you derive from electricity. 468 00:20:56,799 --> 00:20:58,440 Speaker 1: And where does that electricity come from? 469 00:20:58,480 --> 00:20:59,799 Speaker 2: From solar panels on the Moon. 470 00:21:00,039 --> 00:21:03,000 Speaker 1: So let's trace this loop out because it's just wild. 471 00:21:03,359 --> 00:21:06,599 You mind materials on the moon lunar regolith, which has 472 00:21:06,720 --> 00:21:10,279 metals and silicon. You use solar power to run a 473 00:21:10,279 --> 00:21:12,880 factory on the Moon that builds the bodies of the satellites. 474 00:21:13,400 --> 00:21:15,440 Then you use more solar power to run a giant 475 00:21:15,440 --> 00:21:17,720 magna gun that shoots those satellites into space. 476 00:21:18,079 --> 00:21:21,920 Speaker 2: It becomes a self sustaining industrial loop. You stop depending 477 00:21:21,960 --> 00:21:24,559 on Earth for the bulk materials. The Earth becomes a 478 00:21:24,599 --> 00:21:27,559 place for living and space becomes the place for industry. 479 00:21:28,119 --> 00:21:31,799 You fundamentally decouple our economic growth from the finite resources 480 00:21:31,839 --> 00:21:32,680 of our home planet. 481 00:21:32,759 --> 00:21:34,880 Speaker 1: And this is what allows you to scale to numbers 482 00:21:34,960 --> 00:21:39,119 that are frankly incomprehensible by today's standards. The roadmap we're 483 00:21:39,119 --> 00:21:41,759 looking at talks about scaling to one hundred tarot whites 484 00:21:41,799 --> 00:21:43,119 of AI compute per year. 485 00:21:43,240 --> 00:21:46,559 Speaker 2: That numbers it's really hard to visualize. It's almost meaningless 486 00:21:46,559 --> 00:21:47,440 without context. 487 00:21:47,519 --> 00:21:49,640 Speaker 1: Yeah, I need a comparison for one hundred tarot watts. 488 00:21:49,680 --> 00:21:50,359 Give me an anchor. 489 00:21:50,440 --> 00:21:54,440 Speaker 2: Okay, remember our starting number, the entire United States power 490 00:21:54,480 --> 00:21:57,480 grid everything we have is four hundred and sixty gigawatts. 491 00:21:57,559 --> 00:22:00,000 That is zero point four to six terrawats. 492 00:22:00,160 --> 00:22:02,319 Speaker 1: Okay, point four to six. So one hundred taro watts 493 00:22:02,440 --> 00:22:03,880 is my god. 494 00:22:04,039 --> 00:22:06,400 Speaker 2: It is roughly two hundred times more energy than the 495 00:22:06,519 --> 00:22:09,720 entire United States generates today. For all purposes. We are 496 00:22:09,759 --> 00:22:12,039 talking about a level of computing power that is orders 497 00:22:12,039 --> 00:22:15,359 of magnitude beyond anything that exists or could ever exist 498 00:22:15,440 --> 00:22:18,240 on Earth today. It is a capability level that is 499 00:22:18,240 --> 00:22:20,920 functionally infinite compared to our current constraints. 500 00:22:21,160 --> 00:22:23,839 Speaker 1: If you tried to build one hundred tarowatts of data 501 00:22:23,839 --> 00:22:26,759 thenters on Earth, you would literally boil the oceans. 502 00:22:26,880 --> 00:22:31,519 Speaker 2: You would absolutely alter the climate in catastrophic ways. You 503 00:22:31,559 --> 00:22:34,039 would run out of land, out of water. It is 504 00:22:34,079 --> 00:22:36,759 physically impossible to do one hundred tarrawats of compute on 505 00:22:36,799 --> 00:22:40,440 the surface of this planet. But in space, the Sun 506 00:22:40,519 --> 00:22:43,319 is right there. It doesn't care. It's already burning. You're 507 00:22:43,359 --> 00:22:45,319 just catching what it's already putting out. 508 00:22:45,440 --> 00:22:48,079 Speaker 1: This feels like the moment in the conversation where we 509 00:22:48,160 --> 00:22:50,920 have to zoom out, because when you start throwing around 510 00:22:51,000 --> 00:22:53,839 numbers like two hundred times the US power grid and 511 00:22:54,160 --> 00:22:57,799 moon factories, we are talking about tech companies anymore. We're 512 00:22:57,839 --> 00:23:00,000 not talking about quarterly earnings reports. 513 00:23:00,119 --> 00:23:00,759 Speaker 2: It is bigger. 514 00:23:01,160 --> 00:23:04,680 Speaker 1: We are talking about the fundamental evolution of the human specie. 515 00:23:04,720 --> 00:23:06,759 Speaker 2: You're talking about the Kardashev scale. 516 00:23:06,799 --> 00:23:09,640 Speaker 1: I love the Kardashchev scale. It's one of those concepts 517 00:23:09,680 --> 00:23:11,799 that once you hear it, you start seeing it everywhere. 518 00:23:11,920 --> 00:23:13,920 Could you break it down for any listener who might 519 00:23:13,920 --> 00:23:15,079 not have come across it before? 520 00:23:15,319 --> 00:23:18,880 Speaker 2: Sure. It was proposed by a Soviet astronomer, Nikolai Kardashev, 521 00:23:19,000 --> 00:23:21,880 back in the nineteen sixties. He was trying to figure 522 00:23:21,880 --> 00:23:25,839 out a way to classify hypothetical alien civilizations if we 523 00:23:25,880 --> 00:23:28,920 ever found them. He decided the most objective metric would 524 00:23:28,960 --> 00:23:32,240 be to measure a civilization's level of advancement solely by 525 00:23:32,279 --> 00:23:34,559 the amount of energy it can harness and use. 526 00:23:34,839 --> 00:23:37,559 Speaker 1: Okay, so where are we right now on that scale? 527 00:23:37,599 --> 00:23:41,240 Speaker 2: We are effectively a type zero civilization, or maybe a 528 00:23:41,240 --> 00:23:43,000 type point seven if you want to be generous on 529 00:23:43,039 --> 00:23:45,680 a good day, a type one civilization is one that 530 00:23:45,720 --> 00:23:48,039 can harness all of the energy that falls on its 531 00:23:48,079 --> 00:23:49,240 home planet from its star. 532 00:23:49,400 --> 00:23:50,799 Speaker 1: And we aren't even there yet. 533 00:23:50,559 --> 00:23:54,640 Speaker 2: Not even close. We still mostly burn dead plants and dinosaurs, 534 00:23:54,640 --> 00:23:58,240 fossil fuels which are just stored ancient sunlight. We're scraping 535 00:23:58,279 --> 00:23:59,200 the bottom of the barrel. 536 00:23:59,599 --> 00:23:59,960 Speaker 1: Two. 537 00:24:00,440 --> 00:24:03,839 Speaker 2: A Type two civilization harnesses the total energy output of 538 00:24:03,880 --> 00:24:04,839 its host star. 539 00:24:05,160 --> 00:24:08,599 Speaker 1: That's the classic dison sphere concept, right, building a giant 540 00:24:08,640 --> 00:24:11,720 structure that completely encloses the Sun to catch every single 541 00:24:11,799 --> 00:24:13,319 drop of energy exactly. 542 00:24:13,400 --> 00:24:15,920 Speaker 2: Now. To be clear, we are not building a dice 543 00:24:15,960 --> 00:24:20,400 and sphere next week. But this roadmap, this space based 544 00:24:20,480 --> 00:24:26,119 AI plan, this is the first concrete, industrial engineered step 545 00:24:26,200 --> 00:24:29,079 toward becoming a Type two civilization. 546 00:24:28,839 --> 00:24:32,319 Speaker 1: Because we're stopping our reliance on the filtered secondhand energy 547 00:24:32,319 --> 00:24:34,680 that hits the Earth's surface, and we're going directly to 548 00:24:34,759 --> 00:24:35,200 the source. 549 00:24:35,519 --> 00:24:37,920 Speaker 2: We are going straight to the tap. The Sun is 550 00:24:37,920 --> 00:24:40,599 a massive fusion reactor that burns twenty four to seven 551 00:24:40,960 --> 00:24:44,200 with or without us. By putting our most energy hungry 552 00:24:44,279 --> 00:24:48,720 infrastructure or intelligence infrastructure into space, we are bypassing the 553 00:24:48,720 --> 00:24:51,559 filter of the atmosphere. We're harvesting the star directly. 554 00:24:51,920 --> 00:24:54,160 Speaker 1: It honestly gives me chills. I look at my toddler 555 00:24:54,200 --> 00:24:56,720 who's napping in the next room, and I think she's 556 00:24:56,759 --> 00:24:58,119 going to grow up in a world where this is 557 00:24:58,160 --> 00:24:59,559 just normal. She might look up at the moon and 558 00:24:59,559 --> 00:25:01,200 see it time glint of light and just know that's 559 00:25:01,240 --> 00:25:04,920 a factory. Oh yeah, the moon factories. They've always been there, right. 560 00:25:05,960 --> 00:25:10,640 Speaker 2: It completely changes the definition of scarcity. Right now, intelligence 561 00:25:10,720 --> 00:25:13,599 is scarce because compute is scarce, and compute is scarce 562 00:25:13,640 --> 00:25:17,440 because energy is scarce. But if you unlock one hundred 563 00:25:17,559 --> 00:25:22,799 tarowatts of solar powered compute in orbit, intelligence becomes abundant. 564 00:25:23,240 --> 00:25:24,400 It becomes like air. 565 00:25:24,559 --> 00:25:25,359 Speaker 1: You can waste it. 566 00:25:25,559 --> 00:25:28,279 Speaker 2: You can waste it. You can have an AI design 567 00:25:28,359 --> 00:25:31,440 one thousand different variations of a coffee cup just for fun, 568 00:25:31,839 --> 00:25:34,599 because the marginal energy cost is effectively zero. 569 00:25:35,119 --> 00:25:37,240 Speaker 1: But let's bring this back down to Earth for a second, 570 00:25:37,279 --> 00:25:39,799 because while we are dreaming of star harvesting and Type 571 00:25:39,799 --> 00:25:44,400 two civilizations, there are very real economic and geopolitical games 572 00:25:44,400 --> 00:25:47,039 being played right here, right now. There are gonna be 573 00:25:47,079 --> 00:25:50,279 winners and losers in this transition. This isn't just kumbaya. 574 00:25:50,359 --> 00:25:52,559 We all go to space together. This is big business. 575 00:25:52,559 --> 00:25:54,599 Speaker 2: Oh, absolutely, this is going to disrupt everything. 576 00:25:54,680 --> 00:25:59,039 Speaker 1: Okay, let's talk cost. Farz Odd's analysis suggests that counterintuitively, 577 00:25:59,440 --> 00:26:02,559 space based AI eventually becomes the cheapest option. 578 00:26:02,839 --> 00:26:05,079 Speaker 2: It has to be for it to work at this scale. 579 00:26:05,240 --> 00:26:07,559 Think about all the costs you incur building a data 580 00:26:07,559 --> 00:26:09,720 center on Earth. First you have to buy the land, 581 00:26:09,799 --> 00:26:12,799 which is expensive, especially near population centers. 582 00:26:13,079 --> 00:26:15,400 Speaker 1: Then you have to deal with an NV laws not 583 00:26:15,559 --> 00:26:18,720 in my backyard. Yeah, no one wants a power hungry 584 00:26:18,759 --> 00:26:20,400 data center built next to their neighborhood. 585 00:26:20,480 --> 00:26:24,079 Speaker 2: Then you have environmental regulations, You have to hire security guards, 586 00:26:24,240 --> 00:26:26,839 you have to pay for electricity at peak commercial rates 587 00:26:26,839 --> 00:26:29,559 which are always going up. Okay, and in space, no 588 00:26:29,680 --> 00:26:32,480 land cost, no neighbors to complain about the noise or 589 00:26:32,559 --> 00:26:36,160 the view, no environmental regulations in orbit at least not yet, 590 00:26:36,359 --> 00:26:38,720 no staff to pay. It's all going to be automated 591 00:26:38,799 --> 00:26:41,720 robots and the fuel the energy is free. 592 00:26:42,359 --> 00:26:45,000 Speaker 1: So once you pay that initial capital cost of the launch, 593 00:26:45,039 --> 00:26:49,519 which thanks to Starship and eventually Moon manufacturing is dropping 594 00:26:49,519 --> 00:26:52,839 towards zero, the operating margin is just incredible. 595 00:26:52,880 --> 00:26:54,000 Speaker 2: It's almost all profit. 596 00:26:54,240 --> 00:26:57,000 Speaker 1: So if you are at Google or Microsoft or Amazon 597 00:26:57,559 --> 00:26:59,799 and you are currently pouring billions and billions of dollars 598 00:26:59,839 --> 00:27:04,359 in building these concrete bunkers in Iowa, Virginia to Housier servers. 599 00:27:05,119 --> 00:27:07,400 Do you suddenly own a bunch of stranded assets? 600 00:27:07,720 --> 00:27:10,680 Speaker 2: That is the multi trillion dollar question. If Musk and 601 00:27:10,720 --> 00:27:14,039 SpaceX pull this off, terrestrial data centers might start to 602 00:27:14,039 --> 00:27:17,640 look like landline telephones. They still work, but they're obsolete. 603 00:27:17,680 --> 00:27:20,519 They're just too slow and too expensive to run compared 604 00:27:20,559 --> 00:27:22,640 to the orbital alternative. 605 00:27:22,960 --> 00:27:25,039 Speaker 1: Does this level the playing field? Though? I could see 606 00:27:25,039 --> 00:27:27,640 an argument where if a small startup can rent some 607 00:27:27,799 --> 00:27:31,400 cheap space compute, maybe they can finally compete with giants 608 00:27:31,759 --> 00:27:34,519 or are or does it just make the one company 609 00:27:34,519 --> 00:27:37,440 that owns the rockets the absolute, undisputed king of the 610 00:27:37,559 --> 00:27:38,359 entire economy. 611 00:27:38,400 --> 00:27:41,240 Speaker 2: It's the classic railroad tycoon scenario, isn't it. If you 612 00:27:41,279 --> 00:27:44,519 own the tracks, you own the economy. In the eighteen hundreds, 613 00:27:44,559 --> 00:27:47,400 it wasn't the farmers who got fabulously wealthy. It was 614 00:27:47,519 --> 00:27:49,920 Vanderbilt who owned the rail lines that carried the farm 615 00:27:49,960 --> 00:27:50,640 goods to market. 616 00:27:50,720 --> 00:27:53,640 Speaker 1: Precisely and right now, Scarship is the only vehicle capable 617 00:27:53,680 --> 00:27:55,880 of this kind of lift. There is no second place. 618 00:27:56,240 --> 00:27:58,960 China's working on things. Blue Origin is working on New Glen, 619 00:27:59,400 --> 00:28:03,440 But in terms of a flight proven, reusable super heavy lifter, 620 00:28:03,720 --> 00:28:04,720 it's just SpaceX. 621 00:28:05,039 --> 00:28:08,519 Speaker 2: And this brings up the massive national security angle. If 622 00:28:08,559 --> 00:28:12,319 the United States, assuming SpaceX remains a US aligned company, 623 00:28:12,759 --> 00:28:17,279 has a functional monopoly on orbital supercomputing, that is a 624 00:28:17,319 --> 00:28:19,240 trump card unlike any other in history. 625 00:28:19,799 --> 00:28:23,759 Speaker 1: It is the ultimate strategic high ground. Whoever controls the 626 00:28:23,880 --> 00:28:27,480 orbital compute layer, they control the future global economy. And 627 00:28:27,559 --> 00:28:30,519 it's not just about running chatch EPT faster. It's about 628 00:28:30,559 --> 00:28:34,759 running simulations for material science, for drug discovery, for advanced 629 00:28:34,759 --> 00:28:38,559 weather modeling, for cryptographic cracking. It becomes the brain of 630 00:28:38,599 --> 00:28:39,079 the world. 631 00:28:39,480 --> 00:28:42,480 Speaker 2: Imagine a future scenario where the US has one hundred 632 00:28:42,559 --> 00:28:45,839 terawatts of AI compute available to simulate a wargame or 633 00:28:45,880 --> 00:28:49,000 a national cyber defense, and a rival nation has a 634 00:28:49,039 --> 00:28:51,640 few coal powered data centers on the ground. It's not 635 00:28:51,720 --> 00:28:53,880 a fair fight. It's like bringing a calculator to a 636 00:28:53,920 --> 00:28:54,759 supercomputer fight. 637 00:28:54,839 --> 00:28:57,359 Speaker 1: And there's a Tesla connection here too, right The source 638 00:28:57,359 --> 00:28:59,960 material mentions something about the rebellionaire. 639 00:28:59,480 --> 00:29:03,680 Speaker 2: Investment these right Well, without giving any financial advice, we're 640 00:29:03,680 --> 00:29:06,440 just analyzing the source here. The logic is all about 641 00:29:06,559 --> 00:29:10,359 vertical integration. Tesla isn't just a car company. It's an 642 00:29:10,440 --> 00:29:15,039 AI and robotics company. They're building optimists, the humanoid robot 643 00:29:15,400 --> 00:29:17,480 and FSD they're full self. 644 00:29:17,319 --> 00:29:20,640 Speaker 1: Driving software, both of which need just obscene amounts of 645 00:29:20,680 --> 00:29:23,200 training data and processing power exactly. 646 00:29:23,319 --> 00:29:27,799 Speaker 2: So, if Tesla can utilize SpaceX's orbital compute network to 647 00:29:27,880 --> 00:29:32,000 train their robots and their cars, their capability scales almost 648 00:29:32,039 --> 00:29:34,960 infinitely compared to a competitor who is stuck on a 649 00:29:34,960 --> 00:29:37,039 waiting list for a grid connection in Texas. 650 00:29:37,119 --> 00:29:39,119 Speaker 1: It's a synergy that spans from the surface of the 651 00:29:39,160 --> 00:29:41,759 Earth to lower orbit, all the way to the surf 652 00:29:41,839 --> 00:29:42,400 hence of the Moon. 653 00:29:42,519 --> 00:29:44,839 Speaker 2: It's the definition of long term thinking, the idea of 654 00:29:44,880 --> 00:29:47,599 holding a stock through years of volatility because you believe 655 00:29:47,640 --> 00:29:50,440 the underlying infrastructure play is just that massive. 656 00:29:50,559 --> 00:29:52,880 Speaker 1: It's not a bet on car sales next quarter. It's 657 00:29:52,880 --> 00:29:53,680 a bet on the physics. 658 00:29:53,759 --> 00:29:56,359 Speaker 2: The physics say space is better for AI, so the 659 00:29:56,359 --> 00:30:00,039 theory goes, the economy will eventually follow the physics. I 660 00:30:00,200 --> 00:30:03,359 take time, but money always always finds the path of 661 00:30:03,440 --> 00:30:04,240 least resistance. 662 00:30:04,519 --> 00:30:07,599 Speaker 1: Okay, let's do some speculating. We've covered the roadmap we've 663 00:30:07,640 --> 00:30:10,640 covered the physics, the economics, what does the world actually 664 00:30:10,680 --> 00:30:14,680 look like if this happens, let's say five or six 665 00:30:14,759 --> 00:30:17,279 years from now. The timeline slips a bit, as it 666 00:30:17,279 --> 00:30:20,279 always does, but the system is online and starting to scale. 667 00:30:20,440 --> 00:30:22,799 Speaker 2: Well. The acceleration of AI itself is the first thing 668 00:30:22,799 --> 00:30:26,000 you'll feel. I think if we truly remove the power 669 00:30:26,039 --> 00:30:30,160 and cooling constraints, AI improves much faster than even the 670 00:30:30,319 --> 00:30:32,839 current scaling laws predict So we. 671 00:30:32,799 --> 00:30:35,279 Speaker 1: Aren't just talking about a slightly better seria on our phones. 672 00:30:35,480 --> 00:30:38,359 Speaker 2: No, No, we are talking about medical breakthroughs happening at 673 00:30:38,440 --> 00:30:42,000 warp speed. Imagine an AI running in the cold vacuum 674 00:30:42,000 --> 00:30:46,920 of space with functionally unlimited solar power, simulating millions of 675 00:30:46,920 --> 00:30:50,880 potential cancer treatments in parallel twenty forty seven without stopping. 676 00:30:51,119 --> 00:30:53,319 It accelerates the process of science itself. 677 00:30:53,480 --> 00:30:57,119 Speaker 1: Self driving cars become perfect, or at least dramatically better. 678 00:30:57,240 --> 00:31:00,160 Speaker 2: Very likely, the brain of the car network gets smartrder, 679 00:31:00,240 --> 00:31:02,960 and smarter because the training models running in orbit can 680 00:31:03,000 --> 00:31:06,079 be exponentially larger and more complex than anything we can 681 00:31:06,160 --> 00:31:06,960 run on the ground. 682 00:31:07,240 --> 00:31:09,480 Speaker 1: And then there's the elon time factor. We have to 683 00:31:09,519 --> 00:31:12,240 address the elephant in the room. Again. We've been hearing 684 00:31:12,240 --> 00:31:15,359 about robotaxes next year for what almost ten years. 685 00:31:15,240 --> 00:31:17,799 Speaker 2: Now, laughs. That is a polite way to put it. Yes, 686 00:31:18,400 --> 00:31:21,720 he said three years for phase one and four years 687 00:31:21,799 --> 00:31:22,640 for phase. 688 00:31:22,359 --> 00:31:25,000 Speaker 1: Two, which in the real world usually means. 689 00:31:24,759 --> 00:31:26,960 Speaker 2: Maybe it takes six years for phase one. Maybe it 690 00:31:26,960 --> 00:31:30,000 takes ten years to get the Moon factory running. But 691 00:31:30,079 --> 00:31:32,559 here's the catch, and this is the important part. Even 692 00:31:32,599 --> 00:31:37,559 if the timeline slips, the physics remain true. The fundamental 693 00:31:37,599 --> 00:31:41,480 economics of solar energy in space plus passive vacuum cooling, 694 00:31:41,799 --> 00:31:45,240 plus the low gravity of the Moon, those facts don't 695 00:31:45,319 --> 00:31:47,319 change just because of project schedule slips. 696 00:31:47,400 --> 00:31:49,759 Speaker 1: It's inevitable. It might be late, but it's inevitable. 697 00:31:49,839 --> 00:31:52,279 Speaker 2: That's the argument. Once you realize that it is physically 698 00:31:52,319 --> 00:31:55,319 cheaper and more efficient to do this in space, capitalism 699 00:31:55,400 --> 00:31:57,920 itself will drive it there, whether it takes three years 700 00:31:58,000 --> 00:32:01,799 or thirteen years. Gravity and the sunlight aren't changing their minds. 701 00:32:01,960 --> 00:32:04,119 Speaker 1: I have this mental image that I can't shake. Now. 702 00:32:04,480 --> 00:32:06,920 I'm walking my dog at night years from now, A 703 00:32:07,000 --> 00:32:09,240 look up at the Moon and it looks exactly the 704 00:32:09,279 --> 00:32:12,640 same as it always has, that glowing white orb But 705 00:32:12,720 --> 00:32:17,240 I'll know intellectually that up there, silent robots are quietly 706 00:32:17,279 --> 00:32:23,079 building other machines, that invisible railguns are silently firing canisters 707 00:32:23,079 --> 00:32:26,519 of intelligence into the void, and that whole invisible web 708 00:32:26,559 --> 00:32:27,880 is running the phone in my pocket. 709 00:32:27,960 --> 00:32:31,640 Speaker 2: It's the silicon sky. We are literally wrapping the planet 710 00:32:31,680 --> 00:32:33,720 in a new layer of intelligence, which. 711 00:32:33,519 --> 00:32:36,000 Speaker 1: Brings us, I think, to the end of this particular thread. 712 00:32:36,240 --> 00:32:38,759 We started with a tweet about running out of power sockets, 713 00:32:38,839 --> 00:32:41,599 and we somehow ended up with the blueprint for a 714 00:32:41,759 --> 00:32:43,039 Type two civilization. 715 00:32:43,319 --> 00:32:45,640 Speaker 2: It's a journey from a mindset of scarcity to one 716 00:32:45,680 --> 00:32:48,319 of abundance, from the physical limits of Earth to the 717 00:32:48,359 --> 00:32:50,200 vast potential of the Solar system. 718 00:32:50,279 --> 00:32:52,759 Speaker 1: So here's the final thought for you listening right now, 719 00:32:52,920 --> 00:32:54,480 and I want you to really sit with this for 720 00:32:54,519 --> 00:32:58,680 a moment. This roadmap we've discussed paints a picture of 721 00:32:58,720 --> 00:33:02,440 a world where are our intelligence infrastructure. The very brain 722 00:33:02,480 --> 00:33:06,400 of our society floats above us. It's completely detached from 723 00:33:06,400 --> 00:33:09,960 Earth's limitations. It is immune to our weather, our power grids, 724 00:33:09,960 --> 00:33:11,359 our politics, our borders. 725 00:33:11,400 --> 00:33:14,720 Speaker 2: It is arguably the biggest leap and fundamental infrastructure since 726 00:33:14,759 --> 00:33:16,480 the invention of the electric grid itself. 727 00:33:16,680 --> 00:33:20,200 Speaker 1: So my question is this. Does that make you feel liberated? 728 00:33:20,319 --> 00:33:23,599 Does the idea that humanity is finally leveling up, securing 729 00:33:23,599 --> 00:33:27,319 our energy future and truly reaching for the stars does 730 00:33:27,359 --> 00:33:30,319 that fill you with hope? Or does the idea of 731 00:33:30,359 --> 00:33:33,960 a silicon sky controlled by a handful of entities totally 732 00:33:34,000 --> 00:33:36,400 and physically out of reach of the average person does 733 00:33:36,440 --> 00:33:37,160 that terrify you? 734 00:33:37,279 --> 00:33:39,799 Speaker 2: It is the ultimate high ground in every sense of 735 00:33:39,799 --> 00:33:40,160 the word. 736 00:33:40,519 --> 00:33:42,720 Speaker 1: It really is. We want to know where you stand 737 00:33:42,759 --> 00:33:46,279 on this. Are you ready for the moon factories or 738 00:33:46,319 --> 00:33:48,440 do you want to keep the servers down here on 739 00:33:48,480 --> 00:33:51,240 the ground where we can at least see them. Leave 740 00:33:51,240 --> 00:33:53,240 a comment below let us know if you think this 741 00:33:53,319 --> 00:33:56,799 is the great savior of AI or maybe the beginning 742 00:33:56,839 --> 00:33:58,640 of a classic sci fi dystopia. 743 00:33:59,160 --> 00:34:02,279 Speaker 2: I'm cautiously optimistic, but I'll definitely be watching the moon 744 00:34:02,279 --> 00:34:03,680 a little more closely from now on. 745 00:34:03,839 --> 00:34:06,920 Speaker 1: Me too. Thanks for joining us on thrilling Threads. We'll 746 00:34:06,920 --> 00:34:07,680 see you the next one.