Today is the second class of our new April class period. The start time for our class will be 10:00am. We will begin class with a casual conversation. Our reading today is about Buzzballz. Please try to read as much as possible. Underline any words or sentences that are unfamiliar. Our listening is about space. We will finish with our grammar sentences.
Click HERE for the reading
Artificial intelligence is gobbling up electricity around the globe. Some people think the solution may be to take AI off that globe and put it into space. A few companies plan to build AI data centers in orbit around Earth. NPR's Geoff Brumfiel checks in on whether the idea really can fly.
GEOFF BRUMFIEL, BYLINE: Elon Musk likes to think in terms of equations, and here's the math problem he sees for AI. To grow, it needs microchips plus electricity.
(SOUNDBITE OF PODCAST, "CHEEKY PINT")
ELON MUSK: The output of chips is growing pretty much exponentially, but the output of electricity is flat.
BRUMFIEL: That's Musk speaking recently on the podcast "Cheeky Pint." Electricity output is flat because power plants are costly and slow to build and require lots of permits. Musk has experienced these snags firsthand trying to build a giant data center in Tennessee for his company xAI, and he thinks he's found a solution to the problem in another company he runs - SpaceX.
(SOUNDBITE OF ARCHIVED RECORDING)
UNIDENTIFIED ANNOUNCER: Three, two, one.
(SOUNDBITE OF ROCKET LAUNCHING)
UNIDENTIFIED ANNOUNCER: We have liftoff. Go, Super Heavy. Go, Starship.
BRUMFIEL: Musk recently merged xAI with SpaceX. The goal? To build AI data centers in space, where there's free and unlimited solar power.
(SOUNDBITE OF PODCAST, "CHEEKY PINT")
MUSK: You can mark my words. In 36 months, probably closer to 30 months, the most economically compelling place to put AI will be space.
BRANDON LUCIA: That's an optimistic interpretation.
BRUMFIEL: Brandon Lucia is a professor at Carnegie Mellon University who specializes in putting computers on satellites.
LUCIA: You have to think about, OK, I want to put computers inside this box, and I want to put the box in Earth's orbit.
BRUMFIEL: The napkin math looks appealing. Power is free up there. But squeezing enough power out of the solar panels is tough.
LUCIA: And you're really under quite a strict power budget, especially as you're putting more and more high-performance edge computing capability inside the box.
BRUMFIEL: To get enough juice for AI microchips, the solar panels have to be big - really big. One company called Starcloud has estimated its data center might need more than 6 square miles of solar panels. But that's not all. Rebekah Reed is a former NASA official now at Harvard University.
REBEKAH REED: In addition, because you're in the vacuum of space, all of that heat that the compute generates has to be dispelled.
BRUMFIEL: It's true that space is cold, but space is also a vacuum. So if a satellite gets hot, there's no easy way to get rid of that heat. It just builds up. The solution is radiators, which move liquids out to giant panels where the heat can be dissipated. But now, in addition to solar panels, this AI satellite will need another set of large radiators.
REED: And when you put those massive radiators together with massive solar arrays that are required in order to power and cool, you're actually talking about really large satellites or very, very large satellite constellations.
BRUMFIEL: And launching all that into space costs money - lots of money. At the moment, it can cost $500 per pound to launch a satellite into orbit. A data center weighs a lot of pounds.
(SOUNDBITE OF COMPUTERS WHIRRING)
JAMES MATHES: Right now, we're in DC7 (ph). It's definitely loud in here.
BRUMFIEL: To get a sense of just how much would need to go to space, I visited a data center in Ashburn, Virginia. James Mathes showed me around.
MATHES: Currently, we're 36 inches off the floor, walking on a raised floor. All of the power goes under the floor, and the cooling goes under the floor also.
BRUMFIEL: This facility is owned by the company DataBank - rows and rows of black computer cabinets filled with microprocessors. It's fairly run-of-the-mill, but even this average-sized facility consumes around 13 megawatts of power at any given moment. That's 130 times more than the entire International Space Station. And that's just a drop in the bucket.
I mean, how many data centers are there around here?
MATHES: I think now there's in between 160 and 170 in Ashburn alone.
BRUMFIEL: Trying to imagine moving all of this into Earth's orbit seems daunting. And there's something else. As we're walking through, we run into two other employees, Tyerra Hernandez and Sean Wolfe.
TYERRA HERNANDEZ: We're troubleshooting a circuit.
BRUMFIEL: And how often are you guys back here sort of working on stuff like this?
HERNANDEZ: Like, every day of the week, at least (laughter). Once a day, at least (laughter).
BRUMFIEL: Workers are constantly installing new servers, upgrading microchips and fixing things. Orbital data centers will need work, too.
HERNANDEZ: Yeah. We would definitely need someone there (laughter). So that would be...
BRUMFIEL: What do you think? Are you up for it?
HERNANDEZ: Not at the moment. Maybe sometime in the future, but...
SEAN WOLFE: I mean, I'll go to space if somebody else is paying for it.
BRUMFIEL: (Laughter).
And that's really the question. Putting AI into space may be technically possible, but somebody will have to pay a lot of money to launch all those microchips into the sky.