Elon Musk Wants AI in Space — TERAFAB Explained

Kind: captions Language: en What if the future of AI isn't on Earth at all? This week, Elon Musk announced what he calls the most ambitious chip manufacturing project in history. Terrafab. A system designed not just to power AI on Earth, but to scale intelligence into space. Yes, actual data [music] centers in orbit. At its core, Terrafab is about one thing, compute at an unimaginable scale. Musk is aiming for a terawatt of AI compute per year. A number so large, most people aren't even thinking at that level [music] yet. To put that into perspective, all current global AI chip production, just a fraction of that, even if you combine every major manufacturer, TSMC, Samsung, Micron, still not enough. So, the solution? Build an entirely new system from scratch. Because, of course, design, production, testing, iteration, all in one place. A kind of factory that improves itself. But this isn't just about better chips. Musk is connecting this to something much bigger. Billions of humanoid robots, autonomous vehicles, planet-scale AI systems, and eventually, space-based computing. And here's where it gets interesting. On Earth, energy is limited. In space, it's basically unlimited. Solar power in [music] orbit is constant. No weather, no night cycles, no interruptions. No neighbors complaining about data centers. Which leads to a bold claim. Running AI in space could soon become cheaper than running it on Earth. But getting there is [music] a different story. We're talking about launching millions of tons of hardware into orbit [music] every year. That's where SpaceX comes in. Starship is supposed to make [music] that possible. And then comes the next idea. No [music] rockets. A mass driver on the moon. A giant electromagnetic system that launches payloads into space. Because on the moon, [music] gravity is lower, there's no atmosphere, and launches become dramatically cheaper. It sounds like science fiction, and honestly, right now it mostly is. Let's be clear. Terrafab doesn't exist yet. It's a long-term vision with a lot of unsolved problems. Scaling chip production, building space infrastructure, reducing launch costs, and actually needing that much compute. Also, making chips is hard. Even industry leaders [music] struggle. And Musk is proposing to outscale all of them at once. Ambitious? Yes. Guaranteed? Not exactly. But here's the bigger picture. Even if Terrafab doesn't fully happen, it shows where things are going. AI is already hitting limits on Earth, and the next frontier might not be faster chips, but where those chips exist. So, the real question is, are we ready to move intelligence off the planet?