Google Launches First Orbital AI Satellite Test in Space Data Center Quest
Google has launched a test satellite carrying custom AI chips into low Earth orbit to explore whether future artificial intelligence data centers could operate in space.
Google has taken its first step toward exploring whether future artificial intelligence infrastructure could operate in space, sending a refrigerator-sized test satellite into low Earth orbit on Thursday. The spacecraft, nicknamed MVP, lifted off from Vandenberg Space Force Base in California at 11:32 a.m. PT aboard a SpaceX Falcon 9 rocket as part of the Transporter-18 rideshare mission, which carried a total of 130 payloads.
This mission marks the initial in-orbit test for Project Suncatcher, a research initiative Google unveiled in November 2025. While other entities have tested space-based AI hardware previously, MVP represents the first such orbital test conducted by a major tech company. Inside the satellite are four of Google's custom tensor processing units (TPUs), which are specialized chips designed to execute AI workloads. Together, these processors provide roughly the computing power of a single ground-based server. The satellite is expected to process basic AI queries using Google's Gemini models during its operational lifespan of about one year, though it could remain in orbit for up to six years before eventually re-entering the atmosphere and burning up.
The Power and Resource Challenge in Space
The fundamental motivation behind exploring orbital data centers is energy generation. According to Google, satellites situated in low Earth orbit experience near-constant sunlight and can produce up to eight times more solar energy than ground-mounted solar panels. This advantage becomes increasingly relevant as artificial intelligence drives up electricity demand worldwide, often encountering local opposition as rural communities push back against ground-based data center developments. Furthermore, terrestrial facilities frequently strain municipal water supplies through cooling towers that dissipate heat via evaporation.
Before reaching orbit, Google had to clear multiple engineering hurdles. The initial challenge involved surviving the launch sequence, which exposes payloads to gravitational forces up to 10 times normal Earth gravity. Prior to flight, engineers subjected the satellite to multi-axis vibration testing to simulate rocket conditions. Radiation presents another ongoing concern, as cosmic rays and solar activity can induce "bit flips" that alter binary code on the chips. To prepare, Google tested the hardware at UC Davis's Crocker Nuclear Laboratory starting in February 2025, using a cyclotron particle accelerator to blast the components with a radiation dose equivalent to five years in space. Typically, restarting the chips can resolve these errors.
Thermal management in the vacuum of space is another critical factor. Unlike terrestrial systems that utilize air, fans, or water to remove heat, spacecraft must rely entirely on radiation. Google implemented a layered cooling architecture where the chips rest on a motherboard, thermal putty transfers heat into aluminum and copper layers, and a dedicated radiator panel projects the thermal energy outward. Even with this design, MVP's processors cannot maintain continuous operation; Travis Beals, Google's senior director for Project Suncatcher, noted that the hardware can run for approximately 15 minutes before requiring a shutdown to cool down.
What it means for developers
While space-based data centers remain in the early experimental phase, the broader push toward orbital computing reflects how hard the tech industry is searching for scalable power solutions for AI workloads. For developers looking to build and test applications today without waiting for orbital infrastructure, platforms like Apixo provide an efficient bridge. Developers can try top AI models cheaply through one API at https://apixoai.online, accessing models from Claude, GPT, Gemini, Grok, and DeepSeek using a single key.
Google is not alone in pursuing this trajectory. SpaceX, which transported the MVP payload, has submitted filings to the Federal Communications Commission regarding plans for up to 1 million orbital data center satellites, a concept that contributed to interest in SpaceX's recent public offering. Additionally, startup Starcloud launched an Nvidia H100 GPU into orbit in November 2025 to train a small model, while Amazon founder Jeff Bezos has projected that gigawatt-scale orbital data centers could emerge within a decade or more, with support from figures like OpenAI CEO Sam Altman.
Source: Could AI data centers run in space? Google launches its first test — Scripps News AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
