IBM and NASA have expanded their open-source artificial intelligence partnership to the Moon, launching a foundation model designed to identify lunar ice, craters and volcanic terrain.
The NASA-IBM Lunar Foundation Model was trained on more than 30 data layers from nine instruments across four NASA missions, including the Lunar Reconnaissance Orbiter. Benchmark testing showed accuracy improvements of as much as 23% over existing techniques, according to Reuters.
The timing makes the model more than another experimental AI project. NASA is simultaneously accelerating plans for a longer-term presence near the Moon’s south pole, where finding water ice and safe terrain could become crucial.
AI Could Help Decide Where Humans Land
Water ice is one of the most valuable resources NASA hopes to locate on the Moon.
It could provide water for astronauts and potentially support oxygen or propellant production. Better terrain analysis could also help planners identify safer landing and operating areas.
NASA’s latest lunar technology program is seeking systems for oxygen extraction, power generation, energy storage and manufacturing as part of plans for infrastructure near the lunar south pole.
The AI project builds on exploration already underway. Our coverage of the Artemis II Moon mission showed how NASA’s crewed lunar program is laying the groundwork for later landings and longer stays.
IBM Is Taking Foundation Models Beyond Chatbots
The lunar model is part of IBM and NASA’s wider Prithvi family of scientific foundation models.
Rather than competing directly with consumer chatbots, these systems are built to analyze large scientific datasets across areas such as Earth observation, climate and space research.
That fits IBM’s broader strategy of focusing AI on enterprise and scientific computing. We have also tracked the company’s push into advanced computing through its $1 billion quantum investment story.
The Moon Becomes AI’s Next Scientific Test
NASA ultimately wants lunar exploration to support longer stays on the Moon and prepare technology for future Mars missions.
The Lunar Foundation Model could help researchers extract more useful information from decades of spacecraft observations without building a new AI system for every scientific task.
That makes the 23% accuracy improvement important, but the larger opportunity is scale.
If the model continues to perform across lunar datasets, AI could become part of the infrastructure NASA uses to decide where astronauts land, where resources are located and where future Moon facilities are built.
