WindBorne Systems, founded in 2015 and based in Redwood City, California, operates roughly 700 weather balloons that collect atmospheric sensor readings around the globe. The company sells that data to customers and also feeds it into its own AI weather forecasting model, making it one of the first organizations to directly feed empirical data into a deep-learning weather system.
The Air Force contract is worth up to $49 million, but the military branch pays only as it needs data. It can place orders for specific regions at specific times. So far it has ordered $1.7 million worth, and WindBorne executives expect more orders to follow.
The contract was awarded without competitive bidding. WindBorne co-founder and CTO Andrey Sushko told TechCrunch that the deal "is largely a vehicle through which continued growth of the constellation can be supported, as well as some work on research and development to continue to improve the hardware and performance of the balloons." He added that it represented "a determination by the government that for this capability of collecting global data, we were the only company able to provide it in the way that met their needs for these operations."
The company has worked closely with US government agencies, including the National Weather Service and the US Air Force and Navy, but this is its single-largest contract to date.
Earlier this summer, WindBorne released WeatherMesh 6, the latest version of its forecasting model, which the company says provides more accurate forecasts than the traditional physics-based model run by the European Centre for Medium-Range Weather Forecasts (ECMWF). In August, CEO John Dean told TechCrunch the company had raised a $37 million Series B round to continue scaling its business.
AI weather forecasting has become one of the more productive uses of deep-learning technology since the arrival of transformer architectures and ECMWF's release of a large weather dataset. Meteorologists previously relied on expensive supercomputers running detailed equations approximating atmospheric behavior. Now end-to-end learning models can ingest decades of weather data and issue predictions more quickly. Models built by frontier labs and government agencies are proving more accurate than traditional forecasts, though both types remain necessary for useful understanding of the weather. Google DeepMind's latest weather model offers new capabilities, and its cyclone forecasting tools are becoming increasingly central to how the US government responds to hurricane season.