Google's Project Suncatcher Aims to Deploy AI Data Centers in Space

Meet Project Suncatcher, Google’s plan to put AI data centers in space
Ars Technica2

Key Points

  • Project Suncatcher envisions AI data centers on a free‑fall satellite constellation.
  • Satellites must stay within a kilometer, ideally several hundred meters apart, to maintain power links.
  • Google plans to reuse Earth‑based hardware, testing its suitability for space.
  • Radiation tests showed the latest Cloud TPU can endure nearly three times the expected exposure.
  • Prototype satellites are targeted for launch by early 2027.
  • Launch costs could fall to $200 per kilogram by the mid‑2030s, making space data centers viable.
  • Space‑based AI processing could address the noise, power, and water issues of terrestrial data centers.

Google is developing Project Suncatcher, a plan to place AI‑focused data centers on a free‑fall satellite constellation. The design calls for tightly spaced satellites—within a kilometer or even several hundred meters—to maintain power links, a formation tighter than any existing constellation but deemed feasible by Google’s models. To keep costs down, Google intends to reuse Earth‑based hardware, testing its durability by exposing its latest Cloud TPU to intense radiation. Prototype satellites could launch by early 2027, with broader deployment targeted for the mid‑2030s when launch costs may fall dramatically, offering a potential solution to the environmental and community challenges of terrestrial data centers.

Concept and Design

Project Suncatcher is Google’s proposal to situate artificial‑intelligence data centers aboard a constellation of satellites that operate in free‑fall, using no thrust. The satellites would be linked together, forming a network that processes AI workloads in orbit. Because received power diminishes with the square of the distance, Google’s plan requires the satellites to stay within a kilometer of each other, and models suggest that spacing them only several hundred meters apart would keep station‑keeping maneuvers modest.

Technical Challenges

Space‑qualified hardware traditionally costs more and offers reduced performance because it must survive extreme temperatures and radiation. Google’s approach is to reuse components designed for Earth, acknowledging that they may lack the robustness normally required for space. The company points to examples like the Snapdragon‑powered Mars Ingenuity helicopter, which demonstrates that off‑the‑shelf hardware can endure harsh environments longer than expected.

Radiation Testing of TPUs

A core requirement for the project is that Google’s Tensor Processing Units (TPUs) operate reliably for at least five years, a duration that translates to an exposure of 750 rad. To verify this, Google subjected its latest v6e Cloud TPU, known as Trillium, to a 67 MeV proton beam. The tests revealed that memory was the most vulnerable part of the chip, but the TPU withstood nearly three times the anticipated radiation level—approximately 2 krad—before any data corruption was observed.

Launch Timeline and Cost Outlook

Google intends to launch a pair of prototype satellites equipped with TPUs by early 2027. While the initial launch costs are expected to be high, the company anticipates that by the mid‑2030s launch expenses could drop to as low as $200 per kilogram. At such prices, operating space‑based data centers could become economically comparable to traditional terrestrial facilities.

Potential Impact on Terrestrial Data Centers

Conventional data centers are often criticized for being noisy, power‑hungry, and water‑intensive, leading to opposition from nearby communities. By moving AI processing to orbit, Project Suncatcher could alleviate these environmental and social concerns, though it may introduce new considerations for astronomers.

#Google#Project Suncatcher#AI#Data Centers#Satellites#TPU#Space Technology#Radiation Testing#Launch Costs#Terrestrial Data Centers
Generated with  News Factory -  Source: Ars Technica2

Also available in: