NASA and IBM have jointly unveiled a new open-source AI model designed to help scientists predict the timing of solar storms. The collaboration leverages advanced machine learning techniques to analyze vast amounts of solar data collected by NASA’s space-based observatories. By identifying patterns in solar activity, the AI model can forecast when powerful bursts of radiation and charged particles—known as solar storms—are likely to be ejected from the Sun. These events can disrupt satellite communications, GPS systems, and power grids on Earth, making accurate predictions crucial for planetary defense and infrastructure protection. The model is now available as open-source software, enabling researchers, developers, and space agencies worldwide to access, modify, and build upon the technology. This initiative marks a significant step forward in combining space science with artificial intelligence to enhance our understanding of space weather and improve early warning systems.

NASA and IBM have unveiled Surya, a groundbreaking open-source AI model designed to enhance the prediction of solar activity, particularly solar flares that can threaten Earth’s technology and infrastructure. The model, trained on over nine years of high-resolution data from NASA’s Solar Dynamics Observatory (SDO), represents a major leap in space weather forecasting. SDO has continuously monitored the Sun since 2010, capturing images every 12 seconds across multiple wavelengths, creating a dataset exceeding 250 terabytes. This rich, multi-wavelength data allows Surya to analyze the Sun’s magnetic fields and thermal layers in unprecedented detail. Surya’s most significant achievement is its ability to predict solar flares up to two hours in advance—potentially doubling the lead time of current methods. Unlike traditional models that rely on partial observations, Surya generates predictive images of the Sun’s surface, forecasting where and how a flare might occur. Early tests show a 16% improvement in classification accuracy, enabling scientists to pinpoint flare locations and estimate their intensity with greater precision. While two hours may not be enough to fully mitigate the impact of a major flare, it provides critical time for satellite operators, power grid managers, and space agencies to implement protective measures. Solar storms, driven by solar flares and coronal mass ejections, pose serious risks. They can disrupt radio communications, cause data corruption in satellites, damage power grids, and expose astronauts to dangerous radiation. A severe solar storm could cost the global economy up to $2.4 trillion over five years, according to Lloyd’s. Despite advances in monitoring, predicting the exact timing and scale of flares remains a persistent challenge. As ETH Zurich’s Louise Harra notes, while scientists can identify flare conditions, they still lack understanding of why events occur at specific moments—making Surya’s ability to detect subtle, precursory patterns potentially transformative. Surya is not just a specialized tool; it is a foundational model, much like large language models in natural language processing. This design allows it to be adapted for various solar physics tasks beyond flare prediction. IBM researcher Juan Bernabé-Moreno envisions Surya as an “AI telescope” that can uncover hidden patterns in solar behavior, helping scientists understand the mechanisms behind solar instability. He also hopes to integrate Surya with Earth weather models to explore cross-influences, such as whether solar activity affects lightning or atmospheric conditions. The model’s open-source release on GitHub invites the global scientific community to test, refine, and expand its capabilities. While Surya was trained on data from Solar Cycle 24, researchers plan to fine-tune it with data from the current Solar Cycle 25 to ensure its relevance. This ongoing development could lead to even longer lead times and broader predictive power. The unveiling of Surya follows a major tabletop exercise earlier in 2025 that revealed significant gaps in preparedness for extreme space weather events. With Surya, NASA and IBM aim to close those gaps by providing a powerful, accessible tool for early warning. As Bernabé-Moreno emphasizes, the true value lies in the model’s potential to generate new scientific insights and applications through community use. By treating the Sun as a natural laboratory, Surya may not only improve space weather forecasts but also deepen our understanding of stars across the universe.