23 Daily — According to techcrunch.com, Google's first satellite equipped with Tensor Processing Units (TPUs) named Suncatcher prototype went online and established communication with ground control on October 30, 2026. Launched earlier that month aboard a SpaceX Falcon 9 rocket alongside over 100 other spacecraft, this marks Google's inaugural effort to deploy deep learning chips in space.

Demonstrating AI Computing and Real-Time Responses in Orbit

The Suncatcher satellite carries four Google-made TPUs and can generate up to 1 kW of power, but due to power constraints, only one TPU can run at a time for approximately 20 minutes. Ground engineers issued commands to the onboard AI model, Google DeepMind's Gemma 4, asking it to create a space-themed haiku. The satellite successfully responded with: “Blue marble below Data streams in silent grace Starlight feeds my core.” In addition, it reported on orbital environmental conditions, demonstrating its capabilities to execute heavy inference workloads in space.

Testing TPU Performance Under Space Conditions

The satellite is currently undergoing a roughly three-week commissioning and testing campaign focused on assessing TPU performance under thermal stress and cosmic radiation. While Google previously simulated particle radiation damage on Earth, actual space deployment allows it to observe phenomena such as bit flips that can corrupt data during inference and training workloads. This real-world testing is crucial to verifying the stability of AI processing hardware in orbit.

Future Plans and Vision for Orbital Data Center Networks

Should the onboard TPUs perform well, Google plans to launch two more upgraded satellites in 2027 that will have higher power generation abilities and demonstrate inter-satellite optical links. These developments aim to create networks of satellites performing coherent distributed AI workloads in orbit. Compared to other companies like SpaceX, Starcloud, and Cowboy Space, Google’s Suncatcher project is one of the few pursuing high-speed links to support collaborative satellite computation.

Editor's take: A New Era for AI Computing in Space

Google’s deployment of deep learning chips in space represents a technological milestone and redefines possible data center architectures—transforming satellites from mere communication relays into mobile AI compute nodes orbiting the Earth. For entrepreneurs, this signals emerging opportunities in satellite computing and edge AI, encompassing software-hardware integration, space communication, and resource management. Validation methods include monitoring Google’s upcoming test results and public technical reports on radiation effects and operational stability. As the cost of space infrastructure declines, such models might spur innovative applications and services in the near future.