Google Tests AI Data Center in Earth Orbit, an Early Step for Space Computing
Baca dalam 60 detik
- Google akan meluncurkan satelit prototipe pada 1 Oktober 2025 untuk menguji kelayakan chip AI Tensor Processing Unit (TPU) di luar angkasa.
- Inisiatif bernama Project Suncatcher ini bertujuan memanfaatkan sinar matahari konstan di orbit rendah Bumi yang dapat menghasilkan listrik delapan kali lebih besar daripada di darat.
- Jika berhasil, proyek ini dapat membuka jalan bagi pusat data AI orbital, meskipun Google memprediksi komersialisasi masih butuh bertahun-tahun.

Google is reported to be sending artificial intelligence hardware into space on a mission scheduled to launch on 1 October 2025 from Vandenberg Space Force Base, California. The prototype satellite will fly aboard SpaceX's Falcon 9 rocket, marking an early step in the company's ambition to place AI data centers in Earth orbit.
The initiative is part of Project Suncatcher, first introduced last November. Google is targeting two prototype satellites in early 2027 to test its hardware in orbit, with the long-term goal of building an era of large-scale computing in space. This first launch is not a full data center deployment, but an experimental version to measure the performance of its AI chips, Tensor Processing Units (TPUs), under extreme conditions.
According to Google's official blog, low Earth orbit offers near-continuous access to sunlight, allowing satellites to generate up to eight times more solar energy than on Earth's surface. This advantage is seen as capable of supporting large AI workloads, especially if several satellite constellations are interconnected. However, the technical challenges are considerable, from chip resilience to radiation to cooling without airflow.
Google revealed that early tests showed its Trillium TPUs could withstand higher total ionizing radiation than they would receive during a five-year space mission. The team has also tested chip cooling in a thermal vacuum chamber that simulates the space environment. Because there is no air in orbit, cooling becomes a particular challenge, given that TPUs generate a lot of heat in a small area.
For satellite-to-satellite connectivity, Google will use high-speed lasers. The company noted that its laser system must operate at very high bandwidth over very short distances, unlike existing space technology optimized for long distances at low bandwidth. The precision required is equivalent to hitting a coin-sized target from miles away while both points are moving. This laser connection test will take place in 2027 when two satellites are placed in orbit.
"We don't expect, honestly, to have something that's usefully operating in the next few years," James Manyika, Senior Vice President at Google, told The New York Times. He compared it to Google's self-driving car project, which took 15 years of research before bearing fruit.
Google's move is not the first in the industry. Last December, The Wall Street Journal reported that Jeff Bezos and Elon Musk, founders of Blue Origin and SpaceX, are also developing AI data centers in orbit. Sources said the Blue Origin team has been working on the technology for more than a year, while SpaceX plans to launch an upgraded version of its Starlink satellites to provide AI computing payloads.
For Indonesia, this development opens a window on the future of digital infrastructure. As conventional ground-based data centers face constraints on land, energy, and cooling, orbital computing could be a long-term alternative. However, reliance on space technology also raises questions about data sovereignty and access for developing countries. Indonesia, which is building an AI ecosystem and national data centers, needs to watch this trend closely, including the potential for cooperation or international regulations governing computing in orbit.
Google itself stresses that this journey requires methodical engineering. "This first launch is about seeing what works, identifying failure points, and applying those findings to future missions," the company wrote. In other words, the results of the 1 October mission will be the basis for determining whether the vision of AI data centers in space can truly be realized or is merely an expensive experiment with no commercial payoff in the near term.



