On January 26, 2026, NVIDIA unveiled its Earth-2 family of open AI weather models at the American Meteorological Society’s Annual Meeting in Houston, Texas, marking a significant shift in meteorological forecasting capabilities. The company’s new Earth-2 Medium Range model claims to outperform Google DeepMind’s GenCast across more than 70 atmospheric variables for predictions extending up to 15 days ahead.
The announcement comes as artificial intelligence transforms weather prediction, with GenCast having achieved 97.2% accuracy compared to traditional models in December 2024. NVIDIA’s approach focuses on computational efficiency alongside accuracy, with its CorrDiff model operating 500 times faster and 10,000 times more energy-efficiently than conventional numerical weather prediction systems.
Real-world implementations are already demonstrating practical benefits. Amir Givati of the Israel Meteorological Service reported that “NVIDIA Earth-2 models give us a 90% reduction in compute time at 2.5-kilometer resolution compared with running a classic numerical weather prediction model without AI on a CPU cluster.” This efficiency gain allows meteorological agencies to generate higher-resolution forecasts with existing infrastructure.
The energy sector is particularly interested in these developments. Emmanuel Le Borgne of TotalEnergies described the technology as “groundbreaking for our business because they improve short-term risk awareness and decision-making in energy systems where minutes and local impacts matter.” The ability to generate rapid, localized forecasts addresses critical operational needs in renewable energy management and grid optimization.
Despite these advances, AI weather models face inherent limitations. These systems cannot forecast extreme weather events that fall outside their training data scope, a constraint that traditional physics-based models don’t share. This limitation raises questions about reliability during unprecedented climate conditions.
The commercial implications are substantial. The global weather forecasting systems market is projected to grow from $3.92 billion in 2025 to $7.50 billion by 2035, representing a compound annual growth rate of 6.70%. The AI-specific segment shows even more dramatic expansion, with forecasts projecting growth from $197.1 million in 2025 to $926.3 million by 2033 at a 21.3% CAGR.
NVIDIA’s decision to release Earth-2 as an open model family accelerates this market development. By making these tools accessible to meteorological agencies, research institutions, and commercial entities, the company is positioning itself at the center of the AI weather prediction ecosystem while potentially establishing its models as industry standards.
The technology’s implications extend beyond traditional weather services into agriculture, transportation, disaster preparedness, and climate modeling. However, the reliance on historical training data means these systems perform best for typical weather patterns rather than the extreme events that often matter most for disaster response and long-term planning.
