U.S. Air Force awards $49 million contract to WindBorne Systems for balloon data

What happened

The U.S. Air Force has awarded a contract to the startup WindBorne Systems to purchase data collected by its fleet of autonomous weather balloons [2]. The contract is worth up to $49 million, though the Air Force only pays for data as it is needed [2]. The Air Force can place orders for specific times and regions, and it has already ordered $1.7 million worth of data so far [2].

According to WindBorne co-founder and CTO Andrey Sushko, the contract was awarded without competitive bidding [2]. Sushko stated that the government determined WindBorne was the only company able to provide this specific capability for global data collection [2]. He noted that the agreement will support the growth of the balloon constellation and fund research and development to improve balloon hardware and performance [2].

WindBorne, which was founded in 2015 and is based in Redwood City, California, currently operates about 700 weather balloons globally [2]. These balloons collect atmospheric sensor readings, which the company sells to customers and uses to feed its own AI weather forecasting model [2]. In August, the company became the first organization to use an unpowered flight system—a vehicle without an engine—to collect data inside the eye of a typhoon [2]. To navigate, the balloons adjust their altitude, release air, or drop sand used as ballast to move with the winds [2].

Who is affected

The U.S. Air Force is the primary customer under this contract [2]. WindBorne’s technology also involves work with the National Weather Service and the U.S. Navy [2]. The company’s forecasting model, WeatherMesh 6, was released earlier this summer [2]. WindBorne claims this model provides more accurate forecasts than the traditional physics-based models used by the European Centre for Medium-Range Weather Forecasts (ECMWF) [2].

Previously, meteorologists relied on supercomputers to run complex equations to approximate atmospheric behavior [2]. Now, deep-learning technology—a type of AI that learns from large amounts of data—can ingest decades of weather information to produce predictions more quickly [2]. While frontier labs and government agencies are developing these models, the industry remains a developing field where both AI and traditional forecasts are still required [2].