A new AI weather forecasting tool from WindBorne Systems promises more accurate and frequent predictions than traditional methods. Founded by Stanford students in 2019, the team realised their unique combination of model-building and data collection could capture more value.

Their latest version, WeatherMesh 6, delivers hourly forecasts with a resolution down to 3km across Europe and North America – outstripping established systems. The company attributes this success to improvements in how sensor readings are fed into their models, reducing dependency on external data sources like the ECMWF.

Despite its advancements, WindBorne remains focused on building out its model infrastructure rather than developing consumer-facing products. The startup has raised $25 million and now deploys 400 balloons at 15 global sites to gather sensor readings. Its latest innovation includes ADS-B transponders to avoid mid-air collisions.

But can this AI weather prediction really outmanoeuvre government agencies? WindBorne CEO John Dean believes it’s only a matter of time before direct data ingestion from their balloons makes them the go-to source for accurate forecasts.