Recently, NVIDIA launched three new open-source artificial intelligence models, aiming to completely transform the production of weather forecasting. These models challenge traditional and expensive numerical simulation methods through AI-driven algorithms, striving to significantly reduce operational costs while improving forecast speed.

For a long time, conventional weather simulations have been time-consuming and required massive computing power. However, NVIDIA's newly released "Earth-2" series models demonstrate the great potential of AI in this field. According to official statements, once these AI models are trained, their running speed and cost efficiency will far exceed traditional methods, and their accuracy can match or even surpass existing standards.

Specifically, these three models have clear divisions of labor: one focuses on providing global weather forecasts up to 15 days in advance; another targets severe storms within the United States, offering precise short-range forecasts for the next 6 hours. The third model acts as a "data integration expert," capable of processing messy data from various meteorological sensors, providing higher-quality initial data points for other weather prediction technologies.

Currently, NVIDIA has published these models online, showcasing its latest progress in the intersection of high-performance computing and climate science. This marks that the meteorological industry is accelerating its transition from purely physical simulations to a new era driven by AI and data.