The race to establish a sustained human presence on the Moon is increasingly becoming a search for one thing: water. Now, artificial intelligence could give scientists a faster and more detailed way to find it.NASA and IBM have released an open-source artificial intelligence model designed specifically to help researchers study the lunar surface and identify geological features that could be important for future exploration.The NASA-IBM Lunar Foundation Model has been trained using decades of observations collected by NASA spacecraft. By bringing together data from multiple lunar instruments, the model is intended to help scientists analyse the Moon at a scale and speed that would be difficult to achieve through manual examination alone.For future Moon missions, that could prove particularly important in areas where sunlight never reaches the surface.Why Moon water matters so muchWater ice is one of the most valuable resources scientists hope to find on the Moon. Researchers have long known that ice may exist in permanently shadowed regions, particularly around the lunar poles, where extremely low temperatures can preserve water for very long periods.Finding and mapping those deposits could have consequences far beyond supplying drinking water to astronauts.Water can potentially be separated into hydrogen and oxygen. Hydrogen can be used as fuel, while oxygen is needed for breathing and can also form part of a rocket propellant system.That means lunar water could eventually help support a space economy in which some resources are obtained directly from the Moon rather than transported from Earth.The new AI model is intended to help scientists identify areas that deserve closer investigation.Rather than relying solely on individual maps or relatively low-resolution machine-learning systems, researchers can use the model to examine different types of lunar observations together. That could make it easier to detect patterns and geological characteristics associated with areas of scientific or operational interest.AI could help find safer places to landWater is not the only target.Future lunar missions will need to identify landing areas that are scientifically valuable while also being suitable for spacecraft and astronauts. The Moon's surface is covered with craters, slopes, rocks and other geological formations that can complicate landing operations.NASA and IBM say the new model can assist with mapping craters and other surface features, potentially helping researchers evaluate candidate landing locations more efficiently.It can also be used to study volcanic structures and other geological formations, giving scientists another tool to investigate how the Moon formed and evolved.The model was trained on more than 30 layers of data from nine instruments across four NASA missions, including the Lunar Reconnaissance Orbiter. Combining these observations gives the AI a broad view of the lunar environment instead of limiting analysis to a single type of measurement.In benchmark testing, NASA and IBM said the model identified important lunar surface features with accuracy up to 23% higher than widely used approaches.From lunar maps to a future Moon baseThe timing of the technology is significant as NASA's Artemis programme moves toward returning astronauts to the lunar surface.The long-term goal is not simply to repeat the Apollo-era visits. Artemis is intended to develop the technology and operational experience needed for a more sustained human presence around and on the Moon, while also laying groundwork for future deep-space missions.That makes detailed lunar mapping increasingly important.Scientists need to understand where resources may be located, where spacecraft can safely land and which parts of the lunar environment could support future scientific and infrastructure projects. AI tools could reduce the time required to analyse the enormous amount of information already collected by lunar missions.In that sense, the new model does not replace spacecraft exploration. Instead, it could help scientists decide where spacecraft and astronauts should look next.Why open-source AI could change lunar explorationThe NASA-IBM model is also part of a broader effort to make advanced AI tools available to researchers rather than keeping them restricted to individual organisations.It joins the Prithvi family of open foundation models developed by IBM and NASA for applications including geospatial and weather analysis.Making the lunar model publicly available could allow universities, research institutions and other scientists to experiment with the technology and develop new applications for lunar data.That could become increasingly valuable as more missions collect observations from the Moon.The bigger story, however, is what AI could mean for the economics of future lunar exploration. If scientists can more accurately identify water ice, understand terrain and locate useful geological features before a mission launches, spacecraft may be able to target promising areas with greater precision.For astronauts heading back to the Moon, that could make the difference between simply visiting the lunar surface and learning how to use it as a stepping stone for deeper exploration.With NASA planning an astronaut return to the Moon in 2028, the search for lunar water is no longer just a scientific curiosity. It is becoming part of the infrastructure challenge behind humanity's next phase of space exploration—and AI may become one of the most important tools in that search.Frequently asked questionsWhy is NASA looking for water on the Moon?Water could be an important resource for future lunar missions. It can provide drinking water and, after being separated into hydrogen and oxygen, could potentially support breathing systems and rocket fuel production.Where could water ice be found on the Moon?Scientists are particularly interested in permanently shadowed regions near the lunar poles. These areas receive little or no sunlight and can remain extremely cold, creating conditions where water ice can survive.How will NASA and IBM’s AI model help find lunar water?The NASA-IBM Lunar Foundation Model can analyse multiple layers of lunar observation data and help researchers identify surface features and areas that may warrant further investigation. It can also assist with mapping craters and other geological formations.Could lunar water help future missions to Mars?Potentially, yes. If water ice can be extracted and processed efficiently, its hydrogen and oxygen could eventually be used to produce rocket propellant. This could make the Moon a useful staging point for future deep-space missions, including journeys to Mars.