Nvidia is showcasing five projects that apply artificial intelligence and accelerated computing to climate, conservation and sustainability challenges, timed to coincide with Earth Day.
The centrepiece is Earth-2, an open family of AI models, libraries and frameworks for weather and climate prediction that the company describes as the first fully open, accelerated weather AI software stack.
Earth-2 covers the full forecasting pipeline, from processing raw observational data to generating 15-day global forecasts and short-range storm nowcasts, tasks that traditionally require supercomputers running for hours.
Weather agencies and energy companies including the Israel Meteorological Service and TotalEnergies are already using the platform operationally, with the Israeli service reporting a 90% reduction in compute time compared to conventional numerical weather prediction.
In wildlife conservation, GPU-accelerated deep learning models are being used to automate the detection of orangutan nests from drone imagery in Borneo.
One study processed 1,800 aerial images in under five minutes on a single GPU after training on 800 high-resolution images using eight Nvidia GPUs, while a separate model achieved over 99% accuracy in identifying nests.
"Using Nvidia-enabled deep learning, we can now train models that detect and count orangutan nests efficiently from aerial images," said Song-Quan Ong, a computational ecologist at Universiti Malaysia Sabah's Institute of Tropical Biology and Conservation.
In recycling, AMP, a member of Nvidia's Inception Sustainable Futures initiative, is building AI-native sorting facilities that the company says have diverted more than two billion pounds of material from landfills.
AMP estimates the effort has prevented approximately 739,000 metric tonnes of carbon dioxide equivalent emissions.
Its plants achieve a 90% recovery rate compared to roughly 75% at conventional facilities, and the company says it has halved AI inference energy consumption by running workloads on Nvidia Hopper GPUs with TensorRT and Triton Inference Server at the edge.
Nvidia also highlighted work in natural disaster preparedness, where a GPU-accelerated method for modelling Cascadia subduction zone rupture scenarios completes tsunami forecasts in under two-tenths of a second.
Related reading
- Nvidia launches open-source quantum AI models as it deepens grip on chip design workflows
- Nvidia says the open versus proprietary AI debate is the wrong argument
- From Vera Rubins, Blackwell, Nemotron, NemoClaw, to Isaac and Cosmos - here's a whistlestop guide to NVIDIA's…
In satellite imagery, a three-month collaboration between Planet, the Earth observation company, and Nvidia brought raw satellite data processing closer to the sensor, delivering wildfire detection insights in seconds rather than the longer turnaround times associated with conventional pipelines.
The Sustainable Futures initiative within Nvidia's Inception start-up programme now supports more than 750 companies globally working on agriculture, carbon capture, clean energy, climate science, environmental analysis and waste management.
The recap
- NVIDIA spotlights five AI projects for climate and sustainability.
- AMP reports diverting 2 billion pounds from landfills.
- Earth-2 Global Data Assimilation is now available to download.
