IBM and NASA release Surya, an open-source AI “telescope” that predicts solar weather with higher accuracy
- Sep 24, 2025
- 2 min read
Updated: Sep 24, 2025
This Article Is Available In
Solar storms once caused citywide blackouts and threatened power grids, but now IBM and NASA are putting the Sun on notice with Surya—a breakthrough open-source AI model trained on nearly a decade of solar satellite imagery. Surya (“Sun” in Sanskrit) acts like an “AI telescope,” interpreting images from NASA’s Solar Dynamics Observatory to predict solar flares up to two hours in advance, and even visually forecast what the Sun will look like next.
Why is this a big deal? Even a few hours’ notice can mean everything for energy providers and infrastructure operators trying to prevent outages. Surya is light enough (366 million parameters) to run on average hardware but robust enough to boost solar flare prediction accuracy by 16 percent compared to previous methods. In early testing, the model not only predicted “Will there be a solar flare in the next 24 hours?” better than the old guard, it also generated accurate visual snapshots for scientists to study possible changes before they occur.
The new model comes at a moment when NASA’s budget and science staff face federal cuts—potentially threatening the very missions that yield breakthrough data. But Surya, built on nine years of SDO satellite data and NASA researchers’ expertise, is open-sourcing its tools so that innovation can live on in the research community and industry. This means grid managers, scientists, and even startups can freely access Surya on Hugging Face, bringing next-generation space weather prediction to everyone.
