ORLANDO, Fla. — The emergence of China’s DeepSeek has shaken up the artificial intelligence sector, promising new opportunities for space companies beginning to explore ways to leverage AI in space.
AI is seen as key to unlocking true autonomy in orbit and managing an increasingly congested space domain. Yet, while satellite operators like Loft Orbital are making strides to integrate AI into their operations, widespread adoption across the industry remains in its early stages.
Enter DeepSeek, which claims to achieve high performance with significantly lower computational demands than other generative AI — a category of deep-learning models that analyze vast datasets to generate content, answer questions and infer likely outcomes based on learned patterns.
This efficiency is a critical advantage for space applications, where bandwidth and onboard processing power are limited. Notably, DeepSeek is open source, positioning it as a potential catalyst for broader AI innovation.
Details are still emerging, but if DeepSeek can run efficiently on devices at the edge of a network, or within smaller models, it could make real-time AI decision-making more feasible for autonomous satellites, deep-space exploration and other resource-constrained environments.
China’s Sputnik moment
DeepSeek innovations could help large language models get off the ground and into space to increase autonomy, said Douglas Marsh, CEO of Martian Sky Industries, which is leveraging AI to develop orbital debris mitigation technology.
“If you can get systems that are talking to each other — that are trained and have some sort of wrapper around them that’s essentially defining what they’re looking at, what they do, and what they process — you can begin to further automate systems,” Marsh said Jan. 30 on the sidelines of the SpaceCom conference here.
While DeepSeek is not more capable than other generative AI platforms, its main breakthrough lies in the apparent speed at which it was trained to compete with them.
“It means that you can solve problems faster,” said Isaac Passmore, technologist fellow at U.S. government contractor ASRC Federal.
Speaking on a SpaceCom panel, Passmore predicted more frequent “DeepSeek moments” that disrupt the AI landscape, with the next likely to happen within 18 months.