Transcription
Okay, everyone. I want you to listen very closely to this clip from Elon Musk. It has a massive amount of new information about a brand new industry that nobody is talking about. And after the clip, I want to tell you why this is such a big deal.
>> AI in space is inevitable. The way to think of AI in space is that in order to achieve any meaningful percentage of a cla 2 scale civilization where you're using even a million a millionth of the sun's energy you must have solar powered AI satellites in in deep space. What percentage of the sun's energy are you turning into useful work? Then it becomes obvious that space is overwhelmingly what matters. overwhelmingly the the sun only receives one roughly 1/2 billionth of the earth only receives roughly 1 billionth of the sun's energy. So if you want to have something that is say a million times more energy than earth could possibly produce you must go into space. This is where it's kind of handy to have a space company I guess.
>> Chips in space too.
>> Yes.
>> Easier to cool chips in space.
>> You just got to radiate. My estimate is that actually the cost of electricity like the cost effectiveness of AI in space will be overwhelmingly better than AI on the ground. So far long before you exhaust potential energy sources on on earth long before meaning like I think even perhaps in the four or five year time frame if you start doing any kind of scaling for both electricity generation and cooling you realize okay space is incredibly compelling. Like let's say you wanted to do I don't know two or 300 gawatt per year of AI energy.
>> Yeah.
>> It's very difficult to do that on earth. The US average electricity usage last time I checked was around 460 gawatt per year average usage. So something like say at 300 gawatt a year that would be like 2/3 of US electricity production per year. There's no way you're building power plants at that level. And then if you take it up to say a terowatt per year impossible. Yeah. Like you have to do that in space. There there just is there just is no way to do a terowatt uh per year on Earth. And in space you've got continuous solar. You actually don't need batteries cuz it's always sunny in space.
>> Right. Exactly.
>> And and the solar panels actually become cheaper because you don't need glass or framing and the cooling is just radiative.
When I first watched this clip, I was like, "Okay, what is he talking about?" But when I thought about a little bit more, this is genuinely one of the most important technological road maps I've heard in years. And I'm known for hyperbole, okay, when you watch this channel. But this is legitimately a massive deal that most people are going to completely miss. Here's the insight. The big thing that he's talking about here is that Earth only receives about 1 billionth of the sun's total energy output. Everyone that watches this channel knows that AI is going to be a massive deal, and AI in the long term is going to require an immense amount of compute and an immense amount of power. But the big breakthrough here is that solarp powered AI satellites are going to be the thing that's going to take AI to the next level by a factor of a 100 or potentially more.
The core insight here is that Earth only receives about 1 billionth of the sun's total energy output. Literally 1 billionth. And if you think about that for a second, that means all the solar energy that's hitting every square inch of our planet, powering all our nature, all our solar panels, literally everything, our animals, okay? That's 1 billionth of what the sun produces. the rest just radiates out into space, just completely wasted. Now, if you want to scale AI compute to the level where you're actually utilizing a meaningful percentage of the sun's energy, this is exactly what Elon's talking about here. When he mentions the Cardesev scale, you physically cannot do it on Earth. It's not about the lack of will or the lack of money. It's physics. The numbers just don't work.
And so, let me go ahead and put this in very concrete terms. Right now, the entire United States uses about 460 gawatts of electricity per year on average. That's for everything. That's for homes, businesses, data centers, factories, literally all of it. If you wanted to build a 300 gawatt AI comput facility, which is basically where we're going, that would be about 2/3 of total US electricity production. And Elon's eventually talking about scaling that to a terowatt. That's more than twice the entire US capacity. You can't build that on Earth. There's no version of reality where you're building that many power plants that quickly, especially with nimism. Not in my backyard.
And here's where it gets really interesting. The cooling problem. Right now, if you look at one of those Nvidia GB300 server racks, basically what everyone's using for their AI companies like Tesla, all these companies are deploying them, it weighs about two tons. Out of those two tons, 1.95 tons, over 95% is dedicated to cooling infrastructure. The actual compute is this tiny fraction. You're basically building massive refrigeration systems with a little bit of AI hardware tucked inside. And that's the huge unlock there. It's the space required to cool down these AI systems.
This is where I want to deep dive into something most people don't understand about space, which is that radiative cooling. And so the question becomes, how? Space is a vacuum. There's nothing in space. So how can you actually cool these satellites, these AI computers? That's the thing that makes this entire thing economically viable. Let me give an example of how cooling works here on Earth. When your computer chip generates heat, you need to move that heat away from the chip as fast as possible or it literally melts. We do that with cooling systems with fans, liquid cooling, heat sinks, massive HVAC systems. It's incredibly energy-intensive and expensive and it sometimes takes a lot of space. In space, you don't have air, you don't have water, you can't use traditional cooling. But here's the thing, you don't need to. You can use pure radiative cooling. You can just radiate it away.
Let me go ahead and explain to you how this works. Every object that has temperature emits electromagnetic radiation. That's literally what infrared is, heat radiation. There is a law that tells us that the power radiated by an object increases with the fourth power of its temperature. So if you have a chip at say 350 Kelvin is radiating energy into space continuously. Space is essentially at 3 Kelvin. The cosmic microwave background temperature, the remnants of the big bang. It's as close to absolute zero as you can get. That temperature differential 350 Kelvin to 3 Kelvin, that's a massive gradient. And because space is near perfect vacuum, there's nothing blocking that radiation. Just go straight into the infinite cold of space. You literally just need radiators, panels that can efficiently emit infrared radiation. There's no pumps, no coolant, no chillers, no complex machinery, just passive thermal radiation.
These are the kind of technologies that we cover very closely at my website, Farsza.fm. intersection of AI compute, energy infrastructure, and the manufacturing constraints that exist. If you're interested, go ahead and check it out. But the space pivot here, these solarp powered AI satellites in space, that's a solution to the constraints that Elon's describing here.
Now, if you talk about the solar power angle, this is where the economics gets super interesting. On Earth, solar panels need protective glass. They need framing. They need mountain systems. They need to withstand weather. And they only work during daylight hours, which means you need massive battery systems. In space, your solar panels can be ultra thin. Super super thin. No glass, no heavy framing, no weather proofing, just the photovoltaic material itself, just the actual solar panels. And guess what? It's always sunny in space. There's no night. There's no clouds. There's no seasons. Last time I checked, unless there's aliens, right? Your capacity factor is 100%. On Earth, even the best solar installations have capacity factors around 25 to 30% because of dayight cycles and weather. And in space, you're getting power continuously. So, you need very few batteries. The economics are just fundamentally different. Elon mentioned that the cost effectiveness of AI in space will be overwhelmingly better than on the ground. And when you add up all these factors, free cooling through radiation, continuous solar with very few batteries, lighter, cheaper solar panels, no need for massive cooling infrastructure, there might be an actual underelling of this technology. The cost advantage could be 10 times or more or greater as the technology gets better.
But the thing that really blows my mind about this is the timeline. He's saying four to 5 years, not 20, not 50, four to 5 years until space-based AI compute is cheaper than groundbased. That means SpaceX is probably already engineering these satellites right now. They've been working on the Starship rocket which is going to send a ton of mass to orbit and they need ways to figure out how to get customers for that massive rocket. These solarp powered AI satellites probably make the most sense honestly. And even knowing that Elon gets very optimistic about timelines, you probably don't make a statement like that unless you've already done the systems engineering work. You kind of figured it out a little bit.
So think about what this unlocks. If you can get a terowatt of AI compute in space, that's genuinely enough to run AGI systems at scale. Probably that's enough to process real-time data from billions of devices. That's enough to run a virtual world, full-scale climate simulations, protein folding for every drug candidate simultaneously. We're talking about AI capabilities that are orders of magnitude beyond what's physically possible with terrestrial infrastructure right here on Earth. And this is where owning a space company becomes extremely advantageous and essential. SpaceX can launch this stuff at cost. They can iterate rapidly. They control the launch cadence. Nobody else can do this. Not really. Blue Origin is years behind. China's probably working on something similar, but they don't have reusable rockets at scale yet. This is a genuine first mover advantage that could last a decade or more.
But I want to be very clear about something. This is still very hard. There are massive engineering challenges. radiation hardening for the chips, orbital mechanics, space debris, maintaining these satellites, communication latency, the actual deployment. These are not trivial problems, but they're solvable problems is engineering. And when you're Elon Musk and you've already built a rocket company that can land boosters on drone ships, in the freaking ocean, this is probably the kind of hard problem you've been preparing for, and it's going to give a pathway for others to join in as well.
And the other thing that's wild about this is the second order effects, the things that open up based on this technology. If space-based AI compute becomes the standard, then suddenly you need a lot more launch capacity. That's more revenue for Starship. That's more revenue for Blue Origin. You need global satellite communication networks to communicate with these AI satellites. That's more revenue for Starlink. You need ground stations. You need manufacturing. You need the whole new supply chain. the things that will be birthed from AGI. This is vertical integration at a scale we've never seen before. And honestly, the geopolitical implications are staggering. Whoever controls space-based AI infrastructure controls the future of AI development, period. If the US and SpaceX get this right, that's a strategic advantage that's almost impossible to compete with. If China beats us to it, that's a strategic vulnerability that is going to keep everyone in the US up at night.
I think the reason most people are going to miss this is because it sounds too futuristic. It's an Elon Musk sentence. You know, it sounds like something out of a freaking sci-fi novel. But here's what I've learned watching Elon for over a decade. When he gives a specific timeline, like four to 5 years, he's usually already figured out the engineering. He might be late by a year or two, maybe even longer. But the direction is what matters. We're not talking about colonizing Mars in some distant future. We're talking about deploying solarp powered AI satellites in the next presidential term potentially. That's how close this is. And the fact that most people haven't wrapped their heads around this yet tells me that we're still very early in understanding what's about to happen. Just like most things with artificial intelligence, this is why this matters. This is about fundamentally restructuring where and how we build the most important technology of the 21st century. And if Elon's right about the timeline, we're going to start seeing the first deployments in the second half of this decade.
I really wanted to get this out while the clip was still fresh. Let me know what you think in the comments section below. Are we really going to see AI data centers in freaking space in 5 years, or is this another Elon overpromise? I genuinely think this one's real, probably closer than people think. But I love to hear your perspective, especially with Starship's development. Thank you for watching and I'll see you on the next one. Take it easy, everybody. Bye-bye.