Revolutionary AI Model Surya Transforms Solar Storm Prediction and Technology Protection Through NASA-IBM Partnership

The development of the Surya Heliophysics Foundation Model marks a pivotal moment in the advancement of space weather prediction. By harnessing artificial intelligence and over a decade of continuous solar observation data, IBM and NASA have created an unprecedented capability for forecasting solar storms, events that pose significant threats to global technology infrastructure. Announced in August 2025, this collaboration has yielded the first foundation model tailored to heliophysics, achieving a 16% improvement in solar flare classification accuracy and providing visual predictions of flare locations up to two hours in advance. The potential economic impact of major solar storms is immense, with global losses estimated at $2.4 trillion over five years, underscoring the necessity of accurate prediction systems for safeguarding satellites, power grids, and communication networks. By releasing this 366-million parameter model as open source, IBM and NASA have democratized access to advanced space weather forecasting, empowering researchers worldwide to build upon NASA’s extensive Solar Dynamics Observatory (SDO) dataset and create specialized applications.

The Surya model’s open-source nature, performance breakthroughs, and technical innovations not only represent a leap forward in scientific capability but also signal a new era of interdisciplinary collaboration. The model’s development and deployment have implications for global resilience, economic stability, and the future of AI-driven scientific discovery.

Background on Solar Weather and Space Storms

Solar weather, driven by the Sun’s dynamic magnetic and plasma processes, is a major natural hazard for modern technology. The Sun’s nuclear fusion core generates complex magnetic fields, occasionally releasing charged particles and electromagnetic radiation in the form of solar flares and coronal mass ejections (CMEs). When these phenomena reach Earth, they can disrupt satellites, power grids, and communications networks.

The significance of solar storms became widely recognized after the 1859 Carrington Event, which caused global telegraph failures and electrical fires. Today, the stakes are exponentially higher, as our dependence on electronic systems and interconnected infrastructure leaves us more vulnerable to space weather. The launch of NASA’s Solar Dynamics Observatory in 2010 represented a turning point, providing high-frequency, multi-wavelength solar observations that form the backbone of modern predictive efforts.

Solar activity follows an 11-year cycle, with peak periods producing frequent and intense storms. The classification system for solar flares, C, M, and X classes, reflects their increasing intensity, with X-class flares posing the greatest risk. The physical mechanisms behind these events involve the reconfiguration of magnetic field lines in the Sun’s corona, releasing vast amounts of energy that can travel through space at extraordinary speeds.

“If a Carrington-level event occurred today, the technological and economic impacts would be orders of magnitude greater than in the 19th century.”

The Economic Stakes of Space Weather Protection

The economic vulnerability to space weather has grown with the proliferation of satellite-dependent systems and digital infrastructure. According to Lloyd’s of London, a severe solar storm could expose the global economy to losses between $1.2 trillion and $9.1 trillion, with North-America, Europe, and Asia-Pacific facing the greatest risks. These projections reflect the interconnectedness of global power grids, telecommunications, and financial systems.

Electric power grids are particularly susceptible; geomagnetically induced currents from CMEs can overload transformers and trigger cascading failures. The 1989 Quebec blackout, caused by a geomagnetic storm, left millions without power and resulted in significant economic losses, a scenario that could be far worse today. Satellite operators face risks from radiation and charging effects, with past events leading to satellite failures and service disruptions.

The aviation industry is also exposed, especially on polar routes where communication blackouts and increased radiation can force costly rerouting. Telecommunications and GPS services, which underpin everything from agriculture to finance, can be degraded or disabled during severe storms. NOAA’s economic analyses estimate that space weather forecasting services help the electric power industry avoid losses ranging from $111 million to $27 billion, depending on the severity of the event.

“Space weather forecasting is not just a scientific endeavor, it’s an economic imperative for modern society.”

Technical Innovation Behind the Surya AI Model

Surya represents a technological leap in the application of AI to solar physics. The model, named after the Sanskrit word for “sun,” features 366 million parameters and a spatiotemporal transformer architecture designed to handle the massive scale and complexity of solar observation data. Its training relied on nine years of SDO data, encompassing multiple channels and instruments that capture the Sun’s behavior in unprecedented detail.

Key innovations include spectral gating and long-short range attention mechanisms, allowing the model to capture both fine spatial details and long-term temporal dynamics. Surya is the first heliophysics foundation model to use time advancement as a pretext task, enabling it to learn the physics of solar evolution by predicting changes in solar structures over time. This is further enhanced by autoregressive rollout tuning, which improves prediction accuracy over extended timeframes.

The model’s development required specialized computational infrastructure, provided by the National Science Foundation and NVIDIA, to process petabytes of high-resolution data. Surya’s architecture supports parameter-efficient fine-tuning, allowing researchers to adapt it for tasks such as active region segmentation, solar wind forecasting, and EUV spectra prediction.

“Surya’s ability to learn the Sun’s rotation and magnetic dynamics directly from data is a breakthrough for AI-driven scientific discovery.”

Performance Capabilities and Breakthrough Results

Surya’s performance sets new benchmarks in space weather prediction. Testing shows a 16% improvement in solar flare classification accuracy over previous methods, a substantial gain in a field where accuracy directly translates to protection for critical infrastructure. The model can provide visual predictions of flare locations up to two hours in advance, giving operators more time to take protective action.

The model’s versatility extends to forecasting solar wind speeds up to four days ahead and predicting the emergence of active regions that often precede major flares. Its ability to forecast EUV spectra across 1,343 channels enables more accurate modeling of atmospheric and satellite drag conditions, benefiting both scientific research and operational planning.

Surya’s architecture allows it to generalize across multiple applications, with zero-shot evaluations demonstrating its capability to forecast solar dynamics and flare events without task-specific fine-tuning. This broad applicability reduces the resources required for specialized research and operational deployment.

“Surya’s foundation model approach accelerates research and operational forecasting, setting a new standard for space weather science.”

Strategic Partnership Between IBM and NASA

The IBM-NASA partnership exemplifies the power of public-private collaboration in scientific innovation. Building on a broader initiative to develop foundational AI models for Earth and space science, the Surya project leverages IBM’s expertise in AI architecture and NASA’s deep domain knowledge in solar physics. This collaboration has produced not only technical breakthroughs but also new models for interdisciplinary research.

Key figures such as Juan Bernabé-Moreno of IBM and Kevin Murphy of NASA have emphasized the transformative potential of AI for space science. The partnership draws on expertise from multiple institutions, including NASA’s Goddard Space Flight Center, the Jet Propulsion Laboratory, the Southwest Research Institute, and others, ensuring that Surya meets both scientific and operational requirements.

The project’s open-source approach and integration with platforms like Hugging Face and GitHub reflect a commitment to transparency and global collaboration. By making Surya accessible to researchers worldwide, IBM and NASA are fostering a more inclusive and innovative scientific community.

Open Source Accessibility and Democratic Research

Surya’s open-source release is a significant departure from proprietary AI development. By publishing the model and its datasets on platforms such as Hugging Face, GitHub, and IBM’s TerraTorch, IBM and NASA have enabled researchers everywhere to access, modify, and extend Surya for their own applications. This approach supports reproducibility, accelerates innovation, and lowers barriers for institutions with limited resources.

The availability of curated datasets and benchmarks (SuryaBench) simplifies the process of developing and evaluating new heliophysics applications. Educational institutions are already incorporating Surya into curricula, ensuring that the next generation of researchers gains practical experience with advanced AI models.

The open-source model also supports commercial innovation, allowing companies to develop new services for satellite operations, power grid management, and space weather forecasting. This democratization of technology aligns with broader trends in open science and public-private partnership.

Conclusion

The Surya Heliophysics Foundation Model represents a major advancement in space weather prediction, offering improved accuracy, longer lead times, and broad applicability across scientific and operational domains. Its development through the IBM-NASA partnership demonstrates the potential of AI to address critical vulnerabilities in modern infrastructure while advancing fundamental scientific understanding.

As Surya’s capabilities are integrated into global space weather monitoring systems and adapted for new applications, the model’s open-source nature will ensure that its benefits are widely shared. The future of space weather science is likely to be shaped by continued collaboration, technological innovation, and the democratization of advanced predictive tools, transforming solar storms from unpredictable threats into manageable risks.

FAQ

What is the Surya Heliophysics Foundation Model?
Surya is an open-source AI model developed by IBM and NASA to predict solar storms and related space weather events, using high-resolution solar observation data.

How accurate is Surya compared to previous models?
Surya improves solar flare classification accuracy by 16% over earlier methods and can provide visual predictions up to two hours in advance.

Why is space weather prediction important?
Accurate space weather prediction is vital for protecting satellites, power grids, telecommunications, and GPS systems from disruptions caused by solar storms, which can have significant economic and safety impacts.

Is the Surya model available to the public?
Yes, Surya and its datasets are available as open source on platforms like Hugging Face and GitHub, enabling global access for research and development.

What are the future implications of Surya’s release?
Surya’s success paves the way for more advanced, collaborative, and open approaches to space weather prediction, with potential benefits for scientific research, industry, and global infrastructure protection.

Sources:

• IBM Blog
• Military Aerospace