NASA and IBM Launch AI Model That Could Transform How Scientists Explore the Moon

NASA and IBM have launched a new open-source artificial intelligence model designed to help scientists analyze decades of lunar data, map craters, identify potential ice deposits and study volcanic features on the Moon.

The NASA-IBM Lunar Foundation Model brings together vast amounts of lunar observations and uses artificial intelligence to help researchers study the Moon’s surface faster and in greater detail.

The model is publicly available and was developed through a collaboration involving NASA, IBM Research and academic institutions. NASA says it is among the first open-source foundation models built specifically for lunar science.

What Is the NASA-IBM Lunar Foundation Model?

The NASA-IBM Lunar Foundation Model is an AI system designed specifically to analyze scientific data from the Moon.

Instead of creating a separate machine-learning system for every lunar research problem, scientists can start with the foundation model and adapt it for different tasks.

These tasks include:

• Mapping lunar craters
• Identifying potential ice deposits
• Studying volcanic features
• Analyzing changes on the lunar surface
• Understanding the Moon’s geological history
• Supporting future lunar mission planning

NASA says the model was trained primarily using observations from the Lunar Reconnaissance Orbiter, which has collected detailed information about the Moon for more than 17 years.

NASA and IBM Are Turning Decades of Moon Data Into AI

NASA has accumulated an enormous archive of lunar observations through multiple missions.

The Lunar Reconnaissance Orbiter alone has produced an extensive record of the Moon’s surface.

According to NASA, the new foundation model was trained using roughly 2 million image tiles, including more than 1 million high-resolution camera images and nearly 964,000 multispectral images.

The training data also incorporates observations from other missions, including NASA’s GRAIL and Lunar Prospector missions and Japan’s SELENE/Kaguya spacecraft.

This means the AI is not relying on a single type of image.

It can learn from different kinds of lunar observations and combine information from multiple sources.

The AI Could Help Scientists Find Water Ice on the Moon

One of the most important applications is the search for potential lunar ice.

Scientists are particularly interested in permanently shadowed regions near the Moon’s poles.

Some of these locations receive little or no direct sunlight and can remain extremely cold, allowing ice to survive for very long periods.

Finding and mapping lunar ice is important because water could eventually become a valuable resource for future exploration.

Water can provide drinking supplies and, after processing, can potentially provide hydrogen and oxygen for life-support systems and rocket propellant.

The NASA-IBM model can help researchers estimate where ice deposits may be stable on or beneath the lunar surface.

The Model Could Make Lunar Ice Mapping More Accurate

NASA and IBM report that the model showed significant improvements in several benchmark tests.

For lunar ice prospectivity, the model reduced the reported error by up to 22% compared with a commonly used SwinV2-B model in the cited technical work.

Reuters reported that overall benchmark testing showed the new system identifying key lunar surface features with up to 23% higher accuracy than widely used methods.

The exact improvement depends on the specific task and benchmark, so the 23% figure should not be interpreted as the AI being 23% better at every type of lunar research.

Instead, it demonstrates that the model can provide meaningful advantages for particular mapping and classification tasks.

AI Can Also Help Map Lunar Craters

Craters are among the most important geological features on the Moon.

They preserve evidence of impacts that occurred over billions of years and can help scientists estimate the relative ages of different regions.

Crater maps are also useful for future missions.

A detailed understanding of the terrain can help researchers identify safer landing locations and avoid areas containing steep slopes, boulders or other hazards.

The NASA-IBM model can be adapted to detect and classify craters at different scales. NASA says it can work with relatively small amounts of labeled data because the model has already learned from its large pre-training dataset.

The AI Could Reveal the Moon’s Volcanic History

The Moon may look geologically quiet today, but its surface preserves evidence of a much more active past.

Researchers are particularly interested in unusual volcanic structures known as irregular mare patches.

These features appear relatively young compared with much of the Moon’s surface and could provide clues about the Moon’s thermal evolution.

The NASA-IBM model can help researchers identify these features more efficiently, potentially allowing scientists to study larger areas of the Moon than would be practical through manual analysis alone.

Why Is an AI Model Needed for the Moon?

NASA has collected enormous quantities of lunar data over decades.

The challenge is no longer simply collecting information.

Scientists also need efficient ways to analyze it.

Traditional research methods can require scientists to manually examine maps and images or develop specialized machine-learning models for individual tasks.

Foundation models offer another approach.

Instead of starting from zero every time, researchers can take a model that has already learned general patterns from a large dataset and adapt it to a specific scientific problem.

That could reduce the amount of training data and computing effort required for new lunar investigations.

The Lunar AI Model Is Open Source

One of the most significant parts of the announcement is that the model is being made publicly available.

NASA says the model is hosted on Hugging Face, while the complete codebase is available through GitHub.

Researchers can therefore experiment with the system, test it on new scientific questions and develop their own applications.

NASA and IBM have also released machine-learning-ready datasets and benchmark collections alongside the model.

This open approach could allow scientists outside NASA and IBM to contribute improvements and investigate questions the original development team did not anticipate.

What Data Was Used to Train the Model?

The model was built using observations from multiple lunar missions.

The dataset includes information from:

• NASA’s Lunar Reconnaissance Orbiter
• NASA’s GRAIL mission
• NASA’s Lunar Prospector
• Japan’s SELENE/Kaguya mission

NASA says the model primarily uses data from the Lunar Reconnaissance Orbiter, whose observations cover most of the Moon’s surface in high detail.

IBM and NASA also describe a unified machine-learning dataset containing more than 30 spatially aligned data layers from nine instruments across four missions.

Bringing these different observations together is important because individual instruments can reveal different properties of the lunar surface.

Could This AI Help Build a Future Moon Base?

Potentially, yes — although the model itself is not a Moon-base construction system.

Its importance comes from helping scientists better understand the lunar environment.

Future missions will need detailed information about terrain, geology, potential resources and hazards.

Potential water-ice deposits are particularly important because water could become a valuable resource for a sustained human presence.

AI-assisted mapping could help researchers identify promising regions for future exploration and provide better information for mission planners.

NASA’s Artemis program is intended to support a sustained return to the Moon, with future missions also contributing to longer-term plans for human exploration beyond the Moon.

What Makes This Different From a Normal AI Model?

A general-purpose AI model is trained on broad information such as text, images or other common datasets.

The NASA-IBM Lunar Foundation Model is different because it is designed around scientific observations of the Moon.

It learns from specialized geospatial and planetary data rather than ordinary internet content.

That allows researchers to adapt the model to scientific problems such as crater detection, volcanic-feature mapping and lunar-ice prospecting.

The broader idea is similar to the development of specialized foundation models for weather, Earth observation and other scientific fields.

Could AI Discover Something Scientists Have Missed?

That is one of the most interesting possibilities.

The Moon has been observed extensively, but its enormous surface contains countless geological features and subtle patterns.

AI can process large datasets much faster than humans can examine them manually.

That does not mean the AI automatically discovers scientific truths.

Researchers still need to validate its predictions and determine whether apparent patterns are real.

But an AI system can help scientists identify areas that deserve closer investigation.

In that sense, the model could become a tool for directing scientific attention toward previously overlooked features.

What Happens Next?

NASA and IBM are making the model available so researchers can begin adapting it for new lunar science applications.

Future work could expand its ability to analyze different types of lunar data and improve its performance on specialized tasks.

Scientists could also use the model to investigate new geological questions as additional lunar observations become available.

The larger goal is to create a reusable scientific AI system rather than a tool designed for only one experiment.

Why This Matters for Future Lunar Exploration

The Moon is becoming an increasingly important target for scientific research and future human exploration.

Finding water ice, understanding the terrain and identifying geological resources could all influence where future spacecraft and astronauts travel.

AI will not replace spacecraft or scientific instruments.

Instead, it can help researchers make better use of the enormous amount of information those instruments produce.

The NASA-IBM Lunar Foundation Model represents an important shift toward using AI not just to generate content, but to analyze scientific evidence and accelerate discovery.

Key Facts

• Model: NASA-IBM Lunar Foundation Model
• Purpose: Lunar science and exploration
• Organizations: NASA and IBM Research
• Primary data source: NASA’s Lunar Reconnaissance Orbiter
• Training data: Roughly 2 million image tiles
• Applications: Craters, lunar ice, volcanic features and surface analysis
• Data integration: More than 30 spatially aligned layers from nine instruments across four missions
• Availability: Open source
• Public platform: Hugging Face
• Code: GitHub
• Reported benchmark improvement: Up to 23% for key lunar surface feature identification in testing

Frequently Asked Questions

What is NASA’s new lunar AI model?

The NASA-IBM Lunar Foundation Model is an open-source AI system designed to analyze lunar scientific data and help researchers study the Moon’s surface.

Can the NASA AI find water on the Moon?

The model can help researchers identify areas with a higher likelihood of stable ice deposits, particularly in permanently shadowed polar regions. It does not directly confirm the presence of water; those predictions require scientific validation.

How accurate is NASA’s new Moon AI?

NASA and IBM report improvements on several benchmark tasks. Reuters reported that the model identified key lunar features with up to 23% higher accuracy than widely used methods in testing. The improvement varies depending on the specific task and benchmark.

What can the NASA-IBM Moon AI do?

Researchers can adapt it for tasks including crater mapping, identifying potential lunar ice, studying volcanic features and analyzing changes across the lunar surface.

Is NASA’s lunar AI available to the public?

Yes. NASA says the model is publicly hosted on Hugging Face, and its code is available on GitHub for testing and experimentation.

Why is lunar ice important?

Water ice could become an important resource for future lunar exploration. Water can potentially support life-support systems and, after processing, provide hydrogen and oxygen for propellant.

Conclusion

NASA and IBM are turning decades of lunar observations into a new kind of scientific tool.

The NASA-IBM Lunar Foundation Model can analyze different types of Moon data and help researchers investigate craters, volcanic formations and potential ice deposits.

Its open-source release means scientists around the world can experiment with the technology and adapt it to new research questions.

The biggest significance may not be any single map produced by the AI.

Instead, it is the possibility of using artificial intelligence to make decades of lunar observations easier to explore — potentially helping scientists find important features and resources as humanity prepares for a new era of lunar exploration.

Sources & References

Primary source: NASA Science — NASA, IBM Launch AI Foundation Model for Lunar Science.

Additional primary source: IBM — IBM and NASA Release Open-Source AI Model to Support Lunar Exploration.

Independent reporting: Reuters — IBM, NASA launch AI model to help map ice, craters on Moon.

Editorial note: The Science Signal distinguishes between AI-generated predictions and confirmed scientific observations. Potential lunar ice locations identified by the model are research targets and should not be described as confirmed ice deposits without independent verification.