NASA and IBM launch AI Model ‘Surya’ to Predict Sun’s violent outbursts

NASA launches AI model ‘Surya’ to forecast Sun’s violent moods (Image: NASA)

When the Sun lashes out, Earth feels the shock. From power cuts to satellite loss, solar storms hit harder than many realise. NASA is now answering with artificial intelligence.

NASA, in partnership with IBM and academic teams, has unveiled the Surya Heliophysics Foundational Model, an AI system trained on nine years of solar observations. The project draws from NASA’s Solar Dynamics Observatory, which has provided an unbroken record of the Sun since 2010.

AI predicts solar flares with higher accuracy

Surya’s strength lies in analysing vast amounts of solar data. Unlike traditional models, it learns directly from raw images, enabling faster and more accurate forecasts. Early tests show it can predict solar flares up to two hours ahead, beating previous benchmarks by 16 percent.

Scientists say this leap could provide early warnings for satellite operators and power companies. Surya’s forecasts may also help anticipate how ultraviolet radiation affects Earth’s upper atmosphere, where many communication systems are vulnerable.

Why does it matter for Earth?

Solar storms can disrupt aviation, block GPS, and damage power grids. They expose astronauts to harmful radiation and increase drag on satellites in low Earth orbit. “Our society is built on technologies highly susceptible to space weather,” said Joseph Westlake, director of NASA’s Heliophysics Division.

Commercial flights, space missions, and internet satellites are especially at risk as solar activity increases. Experts stress the importance of tools that provide reliable warnings, comparing the need to how meteorology forecasts Earth’s weather.

The "ground truth" solar activity is shown on the top row. The bottom row shows solar activity predicted by Surya. (Image: NASA)

Built for science, open to the world

The Surya model and its training datasets are freely available online, hosted on HuggingFace and GitHub. NASA hopes researchers, students, and developers worldwide will adapt the tool for new applications. Surya’s design allows it to track active regions, forecast solar wind, and integrate data from other missions, including the Parker Solar Probe.

Kevin Murphy, NASA’s chief science data officer, said the model embeds NASA’s expertise into modern AI. “We are making it easier to analyse the Sun’s complexities with unprecedented speed and precision,” he said.

Part of a wider AI push

The initiative forms part of the National Artificial Intelligence Research Resource Pilot, supported by the National Science Foundation and industry partners including NVIDIA. Katie Antypas, NSF’s director of advanced cyberinfrastructure, said the project unites federal and industry resources to accelerate research.

NASA’s Marshall Space Flight Center, IBM, and universities including Princeton, Colorado, and Georgia State collaborated on Surya’s architecture. The SETI Institute and the Southwest Research Institute also contributed expertise.

While designed for heliophysics, Surya’s framework can be applied to other sciences, from Earth observation to planetary research. Officials believe this marks a step towards AI-powered science that strengthens both space exploration and technology here on Earth.