On September 24 local time, U.S. company Alphabet (NASDAQ: GOOG) announced that on October 1 it will launch an experimental satellite carrying its self-developed chip, beginning the first in-orbit space test of its "Sun Catcher" initiative.

The launch will ride on a SpaceX Falcon 9 rocket, with the core aim of studying whether large-scale computing equipment that supports artificial intelligence operations can be moved from the ground into space for construction and operation.

The satellite will carry four self-developed tensor processing units, or TPUs, which are AI-specific chips designed by Alphabet (NASDAQ: GOOG) to accelerate machine learning.

During the test, staff will focus on testing the chip's space adaptability, including checking whether it can withstand the enormous impact force during rocket launch, space radiation and extreme temperatures in low Earth orbit.

Compared with the ground, low Earth orbit can receive sunlight almost continuously, which can provide stable solar power for extremely power-hungry artificial intelligence computing.

Under the plan, Alphabet (NASDAQ: GOOG) will also launch two test satellites in 2027.

Space computing power is a new generation of space-based information service capability following communications, navigation and remote sensing, and countries are accelerating constellation networking and in-orbit computing verification.

In November 2025, Alphabet (NASDAQ: GOOG) unveiled the "Sun Catcher" initiative, envisioning launching a constellation of 81 satellites into low Earth orbit.

Regarding this entirely new technical direction, industry experts analyzed that, constrained by high rocket launch costs, complex space engineering technical limitations and bottlenecks in satellite production capacity, truly building a space artificial intelligence data center will be difficult to achieve in the short term.