Astrolab’s upcoming rover will run the first LLMs on the moon
Large language models are set to go lunar. A computer capable of running AI software is one of the payloads aboard Astrolab’s FLIP lunar rover, which is slated to head to Earth’s celestial companion as early as November 2026.
Astrolab’s FLIP (FLEX Lunar Innovation Platform) rover, scheduled for launch as early as November 2026, will carry the first large language models (LLMs) to operate on the moon’s surface. These LLMs are a type of artificial intelligence software designed to understand, generate, and process human-like text based on vast amounts of data. The rover’s mission includes testing whether these models can function reliably in the moon’s extreme conditions, such as temperature swings between 80°C in sunlight and -200°C in shadows, as well as high levels of cosmic radiation. The goal is to enable greater autonomy for future space missions, reducing the need for real-time communication with Earth, where delays can be up to 20 minutes for Mars and still significant for the moon. This could prevent errors caused by outdated data, such as those that led to the loss of Soviet lunar rovers in the past.
The FLIP rover weighs 480 kilograms and is designed to survive at least 150 hours of the lunar night, a period when temperatures drop drastically and solar power is unavailable. It will carry 20 scientific payloads, including a lidar (light detection and ranging) 3D mapper to create detailed maps of the lunar terrain and a sensor to measure helium-3, a rare isotope found in lunar soil. The rover is part of a series of vehicles NASA has contracted from Astrolab, including future buggies intended to transport Artemis astronauts across the moon’s surface. FLIP’s deformable wheels are engineered specifically for the rugged lunar landscape, allowing it to navigate safely while carrying its heavy payload. The mission is planned to last 10 days, focusing on testing the rover’s endurance and the performance of its onboard systems.
The rover will be equipped with a high-performance computer developed by Hewlett Packard Enterprise (HPE), called the Spaceborne-4. This computer combines a 64-bit CPU (central processing unit) with a GPU (graphics processing unit) capable of running many tasks simultaneously, similar to advanced edge computers used on Earth. Unlike the weak, space-hardened computers on older Mars rovers, which are comparable to late 1990s laptops, the Spaceborne-4 is designed to handle complex AI workloads. It uses software to monitor and adjust for radiation spikes, lowering power and speed when necessary to protect its electronics. HPE has previously tested similar computers on the International Space Station (ISS), where they successfully operated for over two years, proving their reliability in space environments.
During FLIP’s mission, custom-built LLMs will run experiments to demonstrate their ability to operate in the moon’s harsh environment. These models will analyze data from the rover’s cameras and lidar mapper to assist with navigation and decision-making, reducing the amount of information that needs to be sent back to Earth. This is part of a broader effort to make space exploration more autonomous, as human operators on Earth cannot respond instantly to challenges due to communication delays. The LLMs will also help distill insights from the data collected, making it easier to identify important findings without overwhelming ground control with raw information. This technology could be crucial for future missions to Mars and beyond, where delays will be even more significant.
The moon presents extreme conditions that make operating technology difficult. Temperatures can range from 80°C in sunlit areas to -200°C in shaded regions, and cosmic radiation is nearly three times more intense than inside the ISS, which is partially protected by Earth’s magnetic field. The ISS operates at around 20°C, a far more stable environment than the moon’s surface. To cope with these challenges, the Spaceborne-4 computer relies on the rover’s radiators to dissipate excess heat and uses software to detect and mitigate radiation-induced upsets. Despite these hurdles, HPE expects the computer to last for the entire 10-day mission. The lunar south pole, where FLIP is set to land, is of particular interest due to the potential presence of water ice in permanently shadowed regions.
Experts like Roser Almenar, an AI and space law researcher at the University of Valencia, highlight the enormous potential of LLMs for future space missions, especially for Mars and beyond, where communication delays will make real-time Earth-based assistance impractical. However, she also notes that these systems can produce inaccurate or misleading information, which could have serious consequences in space. The article emphasizes that while LLMs could revolutionize space exploration by enabling smarter, more autonomous decisions, their reliability and safety in extreme environments remain a concern. The FLIP mission aims to address some of these concerns by testing the technology in real lunar conditions.
