The U.S. Air Force has entered into a contract to acquire data from WindBorne Systems’ fleet of autonomous weather balloons, a deal that could potentially generate nearly $50 million for the startup. WindBorne currently operates approximately 700 weather balloons worldwide, which gather atmospheric sensor readings that the company markets to clients and integrates into its proprietary AI weather forecasting model.
Founded in 2015 and headquartered in Redwood City, California, WindBorne is among the pioneers in directly incorporating empirical data into a deep-learning weather model. The company has collaborated extensively with the U.S. government, particularly with the National Weather Service, as well as the U.S. Air Force and Navy. However, this contract represents the largest engagement the company has undertaken to date.
The contract is valued at up to $49 million, although the Air Force will only make payments as data is required. Orders can be placed for specific regions at designated times. To date, the Air Force has ordered $1.7 million worth of data, with company executives anticipating additional orders in the future.
“This contract primarily serves as a means to support the ongoing expansion of our constellation, along with facilitating research and development to enhance the hardware and performance of the balloons,” stated WindBorne co-founder and CTO Andrey Sushko, according to hinews. He noted that the contract was awarded without competitive bidding, indicating that the government determined WindBorne was uniquely positioned to fulfill their data collection needs for these operations.
In August, WindBorne CEO John Dean informed hinews that the company had successfully secured a $37 million Series B funding round to further scale its operations. Earlier this summer, WindBorne unveiled the latest iteration of its forecasting model, WeatherMesh 6, which it claims delivers more precise forecasts compared to the traditional model utilized by the European Centre for Medium-Range Weather Forecasts (ECMWF), which relies on physical laws to simulate atmospheric behavior.
With the advent of transformer model architectures and the ECMWF’s release of a substantial weather dataset in recent years, AI-driven weather forecasting has emerged as a highly effective application of deep-learning technology. Meteorologists, who previously depended on costly supercomputers executing intricate equations to approximate atmospheric behavior, can now utilize end-to-end learning to process decades of weather data and generate predictions more swiftly.
Models developed by leading laboratories and government agencies are demonstrating greater accuracy than traditional forecasting methods, although both approaches remain essential for a comprehensive understanding of weather patterns. Google DeepMind’s latest weather model introduces significant new functionalities, with its cyclone forecasting tools becoming increasingly integral to the U.S. government’s response during hurricane season.
In August, WindBorne made history by becoming the first organization to deploy an unpowered flight system—one devoid of an engine—to collect data from within the eye of a typhoon. The company navigates its balloons by adjusting their altitude, releasing air, or dropping sand used as ballast, allowing the vehicles to harness winds that guide them to their intended locations.
“This represents a significant innovation, relying on two key factors: having robust wind and weather models, along with the expertise to navigate them; and maintaining a sufficiently large constellation to ensure balloons are positioned appropriately,” Sushko remarked. “It marks one of the initial instances where a deep-learning model has been applied to actual operational decision-making.”
