NASA and IBM have introduced a new artificial intelligence model designed to help scientists analyse the vast amount of data collected about the Moon.

Called the NASA-IBM Lunar Foundation Model, the open-source system is available through Hugging Face and can support research tasks such as mapping impact craters, identifying relatively young volcanic formations and estimating where water ice may be present near the lunar poles.

Turning decades of lunar data into usable insight

NASA has spent decades collecting detailed information about the Moon through multiple missions. One of the most important sources is the Lunar Reconnaissance Orbiter, or LRO, which has been studying the lunar surface for 17 years.

During this time, LRO has produced an almost complete high-resolution image mosaic of the Moon. The volume of information collected by the spacecraft reportedly exceeds the combined datasets of NASA’s other planetary missions.

This creates a significant opportunity for scientific research, but also a practical challenge. Reviewing, classifying and interpreting such large datasets requires considerable time and specialist effort.

The new foundation model is intended to make this process faster and more accessible.

How the Lunar Foundation Model can support researchers

Primarily trained on data from the Lunar Reconnaissance Orbiter, the model can analyse lunar imagery and help researchers recognise important geological patterns.

Potential applications include:

Detecting and mapping lunar craters
Identifying recent volcanic features
Analysing the structure and composition of the lunar surface
Supporting models that estimate the presence of water ice
Accelerating the classification of large image collections

By automating parts of the initial analysis, the system can help scientists focus their attention on the areas most likely to produce meaningful discoveries.

The model does not replace scientific interpretation. Instead, it acts as a research tool that can process information at a scale that would be difficult to manage manually.

Part of a broader collaboration between NASA and IBM

The Lunar Foundation Model is the latest result of an ongoing collaboration between NASA and IBM focused on applying AI to scientific data.

The two organisations have previously developed foundation models for Earth observation and heliophysics. These initiatives share a common objective: making complex scientific datasets easier to explore and use.

NASA’s wider strategy reflects an important shift in how research institutions approach data. Collecting information is only one part of the process. The ability to organise, analyse and connect that information is increasingly essential for turning it into scientific value.

Making the model open source could also expand its impact. Researchers, universities and technology teams can experiment with the system, adapt it to specific use cases and contribute to its continued development.

AI is becoming part of the scientific infrastructure

The project is also part of a broader movement across planetary science and space exploration.

Space agencies and private companies are increasingly using AI to analyse satellite imagery, detect unusual patterns, optimise mission planning and automate repetitive research workflows. These capabilities can improve scientific productivity while reducing the time and resources required to examine complex datasets.

The same principle applies beyond space research. Many organisations already hold large volumes of valuable data but lack the practical systems needed to extract insight from it efficiently.

Foundation models trained for specific scientific or operational domains can help close that gap. Their value comes not only from processing more information, but from helping experts identify connections and opportunities that might otherwise remain hidden.

For lunar science, this could mean faster geological mapping, better identification of potential exploration sites and a deeper understanding of the Moon’s history and resources.

The NASA-IBM Lunar Foundation Model demonstrates how domain-specific AI can transform extensive archives into practical research tools. As scientific datasets continue to grow, this type of collaboration may become an increasingly important part of how discoveries are made.

Source