IBM and NASA Launch AI to Map Ice and Craters on Moon

IBM and NASA released an open-source AI lunar foundation model trained on 30 data layers from four missions, identifying key surface features 23% more accurately than existing methods.

The race to return humans to the Moon just got a powerful new tool. IBM and NASA have jointly released the NASA-IBM Lunar Foundation Model, an open-source artificial intelligence system designed to help scientists analyse decades of lunar observation data with a level of precision that existing methods have not been able to match.

The model was trained on more than 30 layers of data collected by nine instruments aboard four NASA missions, including the Lunar Reconnaissance Orbiter, one of the most comprehensive remote sensing platforms ever deployed around the Moon.

In benchmark tests, it identified key features on the lunar surface up to 23 per cent more accurately than widely used existing methods, a significant margin in a domain where precision can determine the safety of a landing site or the viability of a future base.

The model joins IBM and NASA’s growing Prithvi family of open foundation models, which spans geospatial, weather, and other Earth and space science applications. Its release as an open-source tool means researchers, universities, and space agencies worldwide can access, adapt, and build upon it without restriction.

The practical applications of the model are closely tied to NASA’s long-term lunar ambitions. It can identify potential ice deposits in the Moon’s permanently shadowed regions, which never receive direct sunlight and are therefore among the coldest and most scientifically promising locations in the entire solar system.

Lunar ice is of particular interest because it indicates the presence of water and oxygen, resources considered essential for sustaining a future Moon base and for producing rocket fuel for onward missions to Mars.

Beyond ice detection, the model can map craters to assist in selecting safe landing sites and study volcanic features that reveal the Moon’s geological history.

These tasks have traditionally required scientists to sift through maps and images manually or rely on lower-resolution machine learning tools, making the new model’s 23 per cent accuracy improvement a genuinely meaningful advance rather than an incremental one.

The timing aligns directly with NASA’s Artemis programme, which aims to return astronauts to the Moon in 2028, testing new technology for a sustained lunar presence and laying the groundwork for future crewed missions to Mars. For that programme to succeed, detailed, accurate lunar surface mapping is not a luxury; it is a precondition.

As Pakistan’s own Jinnah-1 rover prepares to join China’s Chang’e-8 mission to the lunar south pole in 2029, the release of open-source lunar AI tools by IBM and NASA means the scientific community heading to the Moon’s most contested region will have better data than ever before.