The Power Bottleneck Transcript and Summary

Facing a global electricity crisis, Elon Musk just filed a radical proposal to launch one million AI servers directly into orbit.

[Speaker 1]: On February 4th, the FCC accepted a filing for review that looks, at first glance, like a typo. It’s an application from SpaceX to deploy an "Orbital Data Center System." The scale of the request isn't just ambitious; it’s industrial. They aren’t asking for ten satellites, or even ten thousand. They’ve filed for authorization to deploy up to one million compute-enabled satellites. [Speaker 2]: This comes just two days after SpaceX officially acquired xAI on February 2nd. And if you put those two pieces of paper together-the acquisition deal and the FCC filing-you realize we aren't looking at a standard tech merger. We are looking at an attempt to build the largest data center in human history, and they want to put the whole thing in orbit. [Speaker 1]: But there is a reason they are looking at the stars, and it isn’t just about better internet or faster rockets. It’s because down here on Earth, the AI industry is hitting a wall. We are running out of electricity. [Speaker 2]: It’s Sunday, February 22, 2026, and you’re listening to The Angle. [Speaker 1]: If you haven't been following the power markets, that "running out of electricity" line might sound like hyperbole. It isn't. Back in 2024 and 2025, we saw grid operators in places like Ireland and Northern Virginia literally pause new data center connections. The grid was full. [Speaker 2]: Right. The International Energy Agency projects that by 2030, data centers could consume about 3% of the world's entire electricity supply. That is a massive load on aging infrastructure. So when you see SpaceX buying xAI, don't think of it as a software play. Think of it as an infrastructure escape hatch. [Speaker 1]: Exactly. The misconception here is that Musk just wants to own an LLM. But if you look at the structure of this deal, it’s designed for something much heavier. [Speaker 2]: And we should talk about that structure, because it’s specific. This was a "triangular merger." Technically, a subsidiary of SpaceX swapped stock with xAI. This matters because it walls off the liabilities. SpaceX has a very clean balance sheet; xAI, and the whole Twitter-slash-X ecosystem, has a lot more baggage. This setup protects the rocket company from the AI company's debts right before they try to go public. [Speaker 1]: Which brings us to the valuation target. They are aiming for an IPO of the combined entity later this year, and the number bankers are whispering is 1.5 trillion dollars. That would be the largest IPO in history, effectively valuing the company at 50 times its revenue. [Speaker 2]: To get that kind of number, you need a story that’s bigger than "we launch rockets." You need a story that says "we control the future of intelligence." And that is where the orbital data center comes in. [Speaker 1]: So let's walk through the vision here, because on paper, the physics of space actually solve the biggest problems we have on Earth. Problem one: power. On Earth, solar panels work maybe six, seven hours a day at peak efficiency. You have clouds, you have night, you have atmosphere filtering out the energy. [Speaker 2]: In orbit, you don't have any of that. If you put a satellite in what’s called a "Sun-Synchronous" orbit-basically riding the terminator line between day and night-it stays in perpetual sunlight. You get twenty-four hours of peak generation. The yield is roughly five times what you get from a solar panel on the ground. [Speaker 1]: And you don't need massive batteries to store power for the…

Related coverage: Technology · Popular stories