Large language models are preparing for their lunar debut. Astrolab’s 480-kilogram FLIP (FLEX Lunar Innovation Platform) rover is scheduled to head to the Moon as early as November 2026, carrying a high-performance computer capable of running AI software on the lunar surface.
Autonomy Amidst Communication Gaps
The primary driver for lunar LLMs is the need for real-time decision-making. Communication delays between Earth and space robots—which can reach 20 minutes for Mars missions—often lead to operational errors. By running the HPE Spaceborne-4 edge computer, featuring a 64-bit CPU and a GPU, the FLIP rover can process data from onboard cameras and lidar 3D mappers locally, reducing the volume of information that must be transmitted back to ground control.
Software-Defined Resilience
While NASA’s older Mars rovers rely on radiation-hardened hardware with processing power comparable to 1990s laptops, HPE is taking a different path. Their system uses specialized software to monitor for disruptions caused by charged cosmic particles. When radiation spikes are detected, the software automatically throttles power and speed to protect the electronics. This allows the use of advanced, commercially available chips that can handle the parallel computational tasks required by modern AI.
“In space, we don’t have the redundancy, we don’t have the repairman, we don’t have the spare parts,” said Mark Fernandez, chief scientist at HPE.
The mission faces a brutal environment at the lunar south pole, where temperatures swing from 80°C in the sun to -200°C in the shade. Despite the risks of AI hallucinations or errors in a high-stakes environment, researchers view this as a critical step for future missions where immediate assistance from Earth will be impossible.
