The world’s space and defense industries are embarking upon an age of perilous moon exploration, in-situ resource exploitation, and deep space infrastructure development. For space agencies and aerospace companies to be well prepared for long-term human settlement, programs such as NASA’s Artemis have made it necessary to identify safe landing zones, identify dangerous terrain, and pinpoint important resources such as ice at the poles.
Yet, aerospace engineers and planetary scientists confront a staggering volume of data: decades’ worth of disjointed and multimodal satellite telemetry data.
Previous missions have gathered petabytes of data from orbiters in the form of high resolution pictures, topography, temperature, and gravimetry information.
The analysis of this disconnected and incompatible data set usually involved manual review by geologists of the planet or custom machine learning algorithm.
Sifting through maps by hand or building single-use algorithms slows down mission planning and risks missing critical geographic features.
To bridge this data gap, the aerospace sector requires open, multi-modal foundation models capable of unifying petabytes of space data into actionable site selection, hazard assessment, and resource mapping intelligence.
Solving this multi-sensor data challenge, technology giant IBM and NASA announced the open-source release of the NASA-IBM Lunar Foundation Model.
Hosted publicly on Hugging Face and integrated into the open-source TerraTorch geospatial toolkit, the AI model combines over 30 spatially aligned data layers from nine space instruments across four major lunar missions (including NASA’s Lunar Reconnaissance Orbiter, GRAIL, and Japan’s SELENE/Kaguya). The open-source model establishes a unified AI foundation to map water ice deposits, classify craters, and analyze volcanic formations for future crewed bases and commercial lunar operations.
Unifying Multi-Sensor Telemetry for Lunar Operations
This release provides the global aerospace community with pre-trained AI model weights, open datasets, and benchmarking tools. Using the foundation model trained using approximately 2 million image tiles and multi-spectral layers, the performance is enhanced by 23% against conventional machine learning baselines for geographical feature recognition, with lower fine-tuning computation cost.
Some of the important technical and operational pillars of this open-source release are:
Polar Water Ice Prospecting: An improvement of up to 22% in accuracy against conventional vision models for identifying shadowed polar craters which contain ice – an important source of water, oxygen, and rocket fuel.
Precise 1-Meter Scale Crater Identification: Helps in identifying and classifying craters at a 1-meter scale, thus assisting in selecting safe landing sites, avoiding slopes and mapping infrastructure build-outs.
Identification of Volcanic Features and IMPs: An increase of 3% in detection of Irregular Mare Patches (IMPs) and volcanic terrains which helps in understanding the thermal history of the moon and selection of operation zones.
Also Read: Northrop Grumman and SandboxAQ Partners to Advance Lumberjack Quantum Navigation for Uncrewed Aircraft
The Open-Source Prithvi Model Legacy: Expands NASA and IBM’s Prithvi open foundation family-joining Earth observation, weather, and heliophysics models-on Hugging Face and GitHub to democratize space science.
“Uncovering the mysteries of the Moon requires an ability to learn from an extraordinary volume of scientific data,” said Juan Bernabe-Moreno, Director of IBM Research Europe, UK and Ireland. “The NASA-IBM Lunar Foundation Model gives scientists a foundation to explore the Moon at scale, connecting observations across instruments, revealing patterns that are difficult to see in isolation, and providing an open platform the global research community can build on.”
Impact on the Aerospace Industry
The release of the NASA-IBM Lunar Foundation Model signals structural developments across the broader Aerospace landscape:
1. Shifting from “Single-Task Algorithms” to “Reusable Open Foundation Models”
Historically, space missions built custom, single-use algorithms for individual instruments, driving up software R&D budgets.
Open-sourcing the Lunar Foundation Model formalizes the transition toward Domain-Specific Geospatial Foundation Models. Aerospace contractors can adapt a single pre-trained base model using lightweight adapters (LoRAs), reducing AI fine-tuning expenses and accelerating software deployment cycles.
2. Commercialization of In-Situ Resource Utilization (ISRU)
Establishing permanent bases on the Moon requires harvesting local materials rather than launching all supplies from Earth.
Accurately pinpointing polar ice and rare mineral deposits enables commercial space companies to build viable business models around In-Space Resource Extraction, providing oxygen for habitats and hydrogen fuel for deep-space Mars transit.
Overall Effects on Businesses Operating in the Sector
For aerospace prime contractors, commercial lunar lander developers, satellite operators, and defense tech vendors, the NASA-IBM model delivers clear operational benefits:
Key strategic benefits across the commercial aerospace sector include:
Reducing Risk in Commercial Lunar Landings: Automated mapping of 1-meter craters and slopes enables the commercial lander vendors (like Intuitive Machines and Astrobotic) to improve autonomous hazard detection software.
Reducing Entry Barriers in Space Tech Companies: Making weights, code, and datasets available on Hugging Face helps small satellite companies to develop sophisticated lunar analytics applications without investing millions in AI training.
Standardizing Autonomous Surface Navigation: Autonomous rovers and lunar haulers can leverage unified terrain models to navigate rugged polar environments safely.
Conclusion
IBM and NASA’s open-source release of the Lunar Foundation Model marks a vital milestone in planetary science and aerospace engineering. By unifying petabytes of multi-mission data into a reusable, high-precision AI model, these two organizations are providing the digital infrastructure for permanent lunar expansion. For the global aerospace industry, this announcement confirms that the future of space exploration relies on open, collaborative AI frameworks capable of converting raw satellite telemetry into actionable discovery at scale.
