On October 1, our team launched the prototype satellite for Project Suncatcher, marking the beginning of an ambitious research initiative aimed at exploring AI computation in space. Over the past few weeks, we have diligently worked through our commissioning checklist, which involved activating the satellite’s host and TPU trays and conducting initial machine learning tests.
This week, we achieved a significant milestone, known as the “Hello World” moment: we received the first AI response from a data center class TPU operating in orbit. During a pass over a ground station, we sent a prompt to the onboard Gemma 4 open model, requesting it to create a haiku about space. As the satellite traveled over Svalbard, Norway, it successfully transmitted the following poem:
Blue marble below
Data streams in silent grace
Starlight feeds my core. ✨
Following this successful communication, we further tested the system with space-themed limericks and telemetry prompts to assess the payload systems’ technical status. Looking ahead, we plan to conduct a series of tests to evaluate the performance of TPUs under conditions of thermal stress and radiation exposure in orbit. Demonstrating the reliability of our TPUs in executing tasks in space is a crucial step towards our future goal of launching twin interconnected satellites by 2027.
This achievement signifies a major advancement in our journey, as our chips have now earned their ‘space wings.’ It marks an important milestone in the ongoing pursuit of this cosmic exploration initiative.