IBM and NASA Unveil Surya: Open-Source AI for Solar Weather Prediction

IBM and NASA made an open-source AI model for predicting solar weather

Last year, Earth experienced the most powerful geomagnetic storm in two decades, creating spectacular aurora displays visible even in unusually southern parts of the US. While these celestial light shows were beautiful, such storms typically pose a significant threat to our infrastructure. For instance, in 1989, a series of plasma ejections from the Sun caused a nine-hour blackout across Québec, Canada. This time around, power companies were better prepared, and significant service disruptions were avoided in the US and Canada.

This recent event underscored the critical need for advanced preparation against geomagnetic storms. For the past couple of years, NASA and IBM have been collaborating to provide scientists and others with a superior tool for predicting solar weather. Today, they are unveiling the culmination of their efforts: an open-source foundation model named Surya.

Named after the Sanskrit word for the Sun, Surya is described by Juan Bernabe-Moreno, director of IBM Research Europe, UK and Ireland, as an “AI telescope for the Sun.” IBM trained this model using nine years of high-resolution images from NASA’s Solar Dynamics Observatory (SDO), a satellite that has been studying the Sun since 2010.

Surya effectively applies machine learning to interpret solar images and forecast solar events, yielding impressive results. Early tests showed the model was 16 percent more accurate at predicting whether a solar flare would occur in the next 24 hours compared to previous systems. Beyond that, Surya can generate visual predictions of what the SDO might observe in advance. Using data from the most recently completed solar cycle, IBM found Surya could accurately predict the Sun’s appearance two hours ahead of time. Bernabe-Moreno noted they are exploring the accuracy of even longer lead time predictions.

While a two-hour lead time might not seem substantial, Bernabe-Moreno, who has a background with one of Europe’s largest energy companies, considers it a potential game-changer for infrastructure providers. Modern power grids have evolved to be far more responsive, making even short advanced warnings incredibly valuable. Additionally, Surya is a relatively compact 366 million-parameter model, meaning it can run efficiently on less powerful hardware.

There’s an ironic twist to this announcement: it powerfully demonstrates the value of NASA’s science team at a time when its very existence is under threat. The proposed budget cuts by the Trump administration plan to slash NASA’s science budget by nearly half, which would impact missions like the Solar Dynamics Observatory. Although the SDO wouldn’t be outright cancelled like New Horizons or OSIRIS-APEX, its operating budget is projected to be cut from $14 million annually to $8 million per year, according to an analysis by The Planetary Society.

While policymakers from both parties have pushed back against these proposals, NASA’s future remains uncertain as the Senate and House have yet to agree on the agency’s 2026 operating budget. Even if the cuts don’t materialize, NASA is still on track to lose nearly 4,000 employees, about 20 percent of its workforce, as part of broader federal government downsizing efforts.

Bernabe-Moreno emphasized that Surya would not have been possible without NASA’s invaluable assistance. “From the very beginning, the science team at NASA have been informing us what the model should do, how to validate the model, and how to ensure it’s robust,” he stated.

If there is a silver lining amidst the funding uncertainty, it’s that the crucial work of NASA’s science team will live on through models like Surya. As Bernabe-Moreno explained, “The beauty of this model is that we created a capability — an AI platform, if you want. And this capability has applications beyond NASA.”

If you are interested in exploring Surya for yourself, the model is available for download on Hugging Face.