Google DeepMind has released two datasets, SatWeave and PARDIG, that underpin its WeatherNext 3 forecasting model, and is making live feeds and historical archives available to researchers and developers. SatWeave merges imagery from several geostationary satellites covering the Americas, Europe, Africa, the Indian Ocean and Asia-Pacific into a single global mosaic. It spans 17 visible, water vapour and infrared channels and updates hourly. Most existing global composites either carry four or fewer channels or refresh only every few hours. The company says the hourly refresh lets WeatherNext 3 produce new global predictions, rather than waiting for analysis data that is refreshed every six hours. An archive covering more than a decade is included. PARDIG, short for Precipitation AI Reanalysis: Densely Inferred GPM, is an AI-generated global rainfall dataset. It is trained on sparse spaceborne radar observations from NASA's Global Precipitation Measurement Core Observatory, using SatWeave imagery and ECMWF atmospheric fields as inputs. It produces precipitation estimates at 15-minute intervals and roughly 10-kilometre resolution, with an average latency of about five hours. DeepMind says PARDIG improves hourly detection skill over NASA's IMERG-Final product by more than 30 per cent for light-to-moderate rain, and performs better on extreme precipitation. Those results come from evaluations against ground radar networks in the US, Europe and Japan and more than 36,000 weather stations worldwide. DeepMind has not yet published independent evaluations of accuracy in regions without ground-based measurements, and the claim that PARDIG will be reliable everywhere is an expectation rather than a tested result. Historical PARDIG data runs from 1999 to 2026. Both datasets are openly available on Google Cloud Storage, and PARDIG's historical catalogues are also on Google Earth Engine. For Australian businesses and professionals, the release costs nothing to access and creates no direct obligations under the Privacy Act, since it contains weather and rainfall data rather than personal information. Its practical value is mainly for teams doing hydrology, flood or drought modelling, solar forecasting, agricultural planning or insurance risk work, particularly in regions with thin radar coverage. Anyone in those fields could reasonably test PARDIG against their own rainfall records before relying on it. The roughly five-hour latency and 10-kilometre resolution mean it is not a substitute for the Bureau of Meteorology's official warnings or radar products. For most other organisations, the release changes nothing immediately.