Starcloud Demonstrates First Large Language Model Operation on Orbital Satellite
December 12, 2025
In a significant milestone for the nascent space-based computing industry, startup Starcloud has successfully run and queried a sophisticated artificial intelligence model aboard its satellite in orbit. This achievement marks a critical proof-of-concept for deploying advanced computational infrastructure beyond Earth, a frontier attracting billions in investment from major technology and aerospace players.
The company announced that its 'Starcloud-1' test satellite, launched last month and equipped with a high-performance Nvidia H100 GPU, is now hosting and operating Google's open-source Gemma large language model (LLM). According to a CNBC report, this represents the first instance of an LLM running on a powerful Nvidia GPU in space. Starcloud CEO Philip Johnston described the operation as definitive "proof" that the core concept is viable. "This very powerful, very parameter-dense model is living on our satellite. We can query it, and it will respond in the same way that when you query a chat from a database on Earth," Johnston stated, highlighting the model's ability to deliver sophisticated responses from orbit.
While Starcloud-1 is a demonstrator, the company has plans to launch substantially larger orbital data centers should testing proceed successfully. This development occurs amid intensifying competition in the sector. Industry titans Jeff Bezos and Elon Musk have both publicly championed the vision of space-based data centers. Bezos predicted gigawatt-scale orbital facilities within a decade, while Musk confirmed SpaceX's ambitions in the field, with reports suggesting a planned $30 billion initial public offering to fund the venture. Other entities, including Google's Eric Schmidt through 'Project Suncatcher,' NTT, and several specialized startups, are also actively pursuing similar capabilities.
The successful in-orbit AI demonstration is more than a technical feat; it validates a key assumption underlying the economic and logistical rationale for space data centers. Proponents argue that processing data directly in orbit, particularly for Earth observation and satellite communications, can reduce latency and the massive bandwidth costs associated with transmitting raw data to ground stations. As the race to establish a permanent computing presence in space accelerates, Starcloud's test provides tangible evidence that the necessary hardware and software can function in the harsh environment of space, potentially paving the way for a new layer of global digital infrastructure.
Source: datacenterdynamics
Starcloud Demonstrates First Large Language Model Operation on Orbital Satellite
