Story perspectives

Google Tests Gemini AI on Prototype Suncatcher Satellite

By Drooid · Oct 1, 2026, 10:52 PM UTC · How we work

Summary

Story summary

Google launched a prototype satellite for Project Suncatcher on SpaceX’s Transporter-18 rideshare on 1 October 2026.
The satellite carries four Tensor Processing Units that will run the Gemini AI model for 15 minutes per session.
Heat-dissipation tests require chips to power down every 20 minutes, prompting trials of radiators and heat-pipes.
Solar panels in sun-synchronous orbit are claimed to generate up to eight times more power than ground installations.

History

Historical Context

In 1999, NASA launched the Tracking and Data Relay Satellite-5 (TDRS-5), the first communications satellite to employ active thermal-radiator panels for heat management. Lockheed Martin built TDRS-5, equipping it with 2.5 kW of radiator capacity that reduced onboard processor temperatures by 15 % during peak operations. Google’s 2026 Project Suncatcher prototype mirrors this approach, mounting radiators and heat-pipes to sustain Tensor Processing Units running the Gemini AI model in sun-synchronous orbit. The precedent shows that thermal-control innovations can enable high-power AI workloads in space, prompting SpaceX’s Transporter-18 rideshare missions and broader commercial satellite AI services.

1 of 2

Story summary

Google launched a prototype satellite for Project Suncatcher on SpaceX’s Transporter-18 rideshare on 1 October 2026.
The satellite carries four Tensor Processing Units that will run the Gemini AI model for 15 minutes per session.
Heat-dissipation tests require chips to power down every 20 minutes, prompting trials of radiators and heat-pipes.
Solar panels in sun-synchronous orbit are claimed to generate up to eight times more power than ground installations.

1 / 2