Google to Launch AI Satellite Next Week to Test TPU Performance in Space

The primary goal of the mission is to measure how the TPUs handle the physical stress of spaceflight, including extreme radiation and thermal fluctuations. While preliminary tests at the UC Davis Crocker Nuclear Laboratory showed that Trillium TPUs could withstand radiation levels exceeding a typical five-year space mission, Google noted that certain environmental factors can only be truly tested in orbit.

Project Suncatcher is a long-term research initiative exploring whether space can host scalable AI computing infrastructure. Low Earth orbit offers an advantage for AI data centers due to near-constant sunlight, which can generate up to eight times more solar power than on Earth.

The mission also aims to test specialized cooling systems, such as combinations of heat pipes and radiators, to manage the significant heat generated by TPUs in a vacuum where airflow is unavailable. Following this initial test, Google plans to launch two more satellites into orbit next year to test high-bandwidth laser communications between them.

PLUS ULTRAby Amenoyomi

The strategic driver for moving AI compute to orbit is energy efficiency. In low Earth orbit, satellites have near-constant access to sunlight, allowing them to generate up to eight times more solar power than terrestrial installations. This immense power availability is the primary motivation for scaling energy-intensive AI infrastructure in space.

Before reaching operational status, hardware must survive the extreme physical toll of launch and the orbital environment. While the overall spacecraft experiences sustained loads up to 10g, individual components like TPU chips can be subjected to forces between 50g and 100g. Additionally, hardware must withstand cosmic rays and solar radiation; initial tests on Trillium TPUs indicate they can survive radiation doses exceeding those of a typical five-year mission.

The most critical technical barrier is heat dissipation. Unlike terrestrial data centers that rely on airflow and convection, a vacuum allows heat to be diffused only through radiation via radiators. Because TPUs generate intense heat in a small area, this limitation currently creates a severe operational bottleneck. Early estimates indicate that chips can only run for approximately 15 minutes before they must be powered down to cool.