Nvidia launches ‘Earth-2’ AI platform

Nvidia has introduced its Earth-2 AI models, a suite of advanced frameworks designed to revolutionize weather forecasting. By combining generative AI with satellite, radar, and ground-station data, the system delivers ultra-fast atmospheric predictions — reportedly up to 500 times faster than conventional methods — enabling quicker climate insights and early warnings.

The world's leading tech company Nvidia has introduced AI models under the name Earth-2. These models include frameworks such as CorrDiff, FourCastNet3, Medium Range, Nowcasting, Global Data Assimilation, and PhysicsNeMo. Among these, PhysicsNeMo has been introduced to train and improve AI-physics models. All these models integrate information from satellites, radar, and weather stations to accurately assess the state of our atmosphere.

Use of Generative AI

The Earth-2 AI models utilize several frameworks that employ generative AI. This makes forecasting much faster. It is claimed that these AI models can predict weather 500 times faster than existing technology.

Different Models, Different Functions

The Earth-2 AI models have various frameworks, each with different functions. According to a TechRadar report, the CorrDiff framework uses a generative AI architecture to transform continent-level predictions into regional predictions. Similarly, FourCastNet3 is designed to accurately predict wind, temperature, and humidity. It is claimed to predict 60 times faster than traditional models. The PhysicsNeMo model is used to train AI-physics models. According to the report, it is useful for both weather forecasting organizations and scientific research.

Even Supercomputers Fail in Comparison

Earth-2 is so advanced and intelligent that it can detect changes in the atmosphere in a matter of seconds. Supercomputers take hours to perform this task. If a change can be detected in a few seconds, alerts can be sent to people much faster. Some of Nvidia's frameworks are already being used by weather departments in several countries.

Used in Israel

CorrDiff is being used in Israel's weather forecasting service. As we mentioned, this system uses regional assessments of continent-level activity to generate forecasts. Israel will soon be adopting Nowcasting, which will allow for weather forecasts to be updated eight times a day using high-resolution graphics. Besides Israel, meteorologists in Taiwan are also utilizing several of NVIDIA's frameworks.