Google’s inaugural data center satellite, launched earlier this month via a SpaceX Falcon 9 rocket, has successfully established communication with ground control and begun operations. The prototype satellite, part of Google's Suncatcher project, carries four Tensor Processing Units (TPUs) capable of running artificial intelligence workloads in orbit, generating up to 1 kilowatt of power. Due to the high energy consumption, only one TPU can be activated at a time for about 20 minutes, which presents challenges for space-to-ground data transmission.

After a three-week commissioning phase to verify all systems, engineers ran the first test AI task on the satellite using Google DeepMind’s Gemma 4 model. The AI delivered a haiku about space, showcasing its functionality with writing creative poetry and generating space-themed limericks. Beyond these playful tests, the satellite also monitored its orbital environment, confirming it is operating normally. The mission aims to evaluate how the TPUs withstand space conditions like thermal stress and radiation, with the latter known to cause data errors in electronics.

While Google has tested TPU resilience in Earth-based particle accelerators, this mission provides real-world data on radiation’s effects on AI hardware in orbit. The results will influence the design of two additional advanced AI satellites Google plans to launch in 2027, which will have greater power capacity and demonstrate laser-based optical communication links for satellite-to-satellite networking. This effort aligns with Google’s vision of creating high-speed, distributed space data center networks capable of performing large-scale, coherent AI workloads.

Google’s Suncatcher initiative operates in a competitive but complementary landscape alongside SpaceX, in which Alphabet holds a significant stake. Both rely on the future development of SpaceX’s Starship heavy-lift rocket to make space data centers viable at commercial scale. Industry analysis suggests Starship will need roughly 1,800 flights to achieve cost efficiencies comparable to Earth-based data centers. Meanwhile, Alphabet faces near-term revenue pressures from integrating AI into its business, underscoring the long-term nature of these ambitious space computing projects.