In December 2025, startup Starcloud trained the first large language model ever trained in orbit, using an NVIDIA H100 — the same class of GPU built for Earth's AI data centres, now running roughly 500 kilometres above the planet. - ScienceBlog.com

In December 2025, startup Starcloud achieved a milestone by training the first large language model in orbit using an NVIDIA H100 GPU — the same high-performance chip used in terrestrial AI data cente... Read original →

Starcloud’s in-orbit training of a large language model using an NVIDIA H100 marks a pivotal shift in AI computing beyond Earth. The successful deployment of data-center-grade chips in low-Earth orbit validates the feasibility of space-based AI infrastructure, driving semiconductor vendors to enhance radiation-hardened and thermal-efficient designs for space applications. This milestone intensifies global regulatory scrutiny over space AI compute standards, particularly concerning data sovereignty and supply chain resilience. Competitors such as Amazon, Google, and SpaceX are likely to accelerate their own orbital AI platforms to capture emerging market share. In the short term, orbital AI training will attract significant capital investment, while long-term implications include the development of autonomous AI systems for deep-space exploration, fundamentally reshaping the satellite and remote sensing industries.