Freakonomics Radio
Freakonomics Radio

682. Should A.I. Move to Space?

July 24, 2026

AI Summary

5 min read

When AI’s Energy Appetite Outgrows the Planet

The International Energy Agency expects that roughly half of all new U.S. electricity demand through 2030 will come from data centers. Blaise Agüera y Arcas, a vice president and fellow at Google, watched this trajectory with growing unease. He started off worried, then reassured himself that efficiency improvements could deliver a factor of a thousand in energy savings. Then he did the math: a factor of a thousand, in an exponential growth landscape, buys you only about a decade. “I see no reason to believe that the demand for AI is somehow gonna saturate in the next 10 years,” he says. That realization pushed him toward a radically different solution: move the computing into space.

Why Prediction Is the Core of Intelligence

Agüera y Arcas leads Google’s Paradigms of Intelligence team, which studies intelligence as a unified phenomenon spanning brains, societies, and artificial systems. The team includes neuroscientists, philosophers, and sociologists alongside AI researchers. Their work has led to a provocative claim: the primary function of intelligence, and of life itself, is prediction.

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

What you'll learn

  • 1 (01:44) **Opening Puzzle: The Hidden Space Economy** - Steven Dubner lists diverse industries (agriculture, rideshare, banking, etc.) and reveals their common reliance on space-based infrastructure.
  • 2 (02:19) **Will Marshall: From NASA to Planet Labs** - Marshall’s journey from quantum physics at Oxford to building cheap satellites in a garage, founding the $10B company Planet.
  • 3 (05:32) **Blaise Agüera y Arcas: The Paradigms of Intelligence Team** - Levitt introduces his guest, a Google VP who runs a research team that blends AI, neuroscience, and philosophy.
  • 4 (08:29) **The Social Nature of Intelligence** - A deep dive into how disagreement among agents (both in AI and in the human brain) leads to better problem-solving.
  • 5 (12:58) **The Core Thesis: Intelligence Is Prediction** - Agüera y Arcas defends the controversial idea that prediction is the primary function of intelligence and life itself.
  • 6 (16:55) **Project Suncatcher: The Crazy Idea That Might Work** - Agüera y Arcas introduces the most ambitious engineering project his team is working on: moving AI data centers into space.
  • 7 (19:31) **The Energy Problem: Why Nuclear and Earth Solar Aren't Enough** - A comparison of terrestrial energy options for AI data centers and their limitations.

+ Full timestamped outline available in the app

Guests on this episode

Show Notes

At first it sounds ridiculous. But it might be inevitable. Guest host Steve Levitt talks to a team that hopes to push A.I. infrastructure off the planet. Part one of a two-part series.)

 

  • SOURCES:
    • Blaise Agüera y Arcas, vice president and fellow at Google, C.T.O. of technology and society.
    • Travis Beals, senior director of product management at Google, Project Suncatcher lead.
    • Will Marshall, co-founder and C.E.O. of Planet Labs.

 

 


Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Freakonomics Radio

More from this podcast

Freakonomics Radio →