A new contract with the U.S. Air Force will see the military branch purchase data from WindBorne Systems‘ autonomous weather balloon network, a deal potentially valued at close to $50 million for the company.
WindBorne presently operates approximately 700 weather balloons worldwide, gathering atmospheric sensor data. This data is supplied to clients and also integrated into WindBorne’s proprietary AI weather forecasting model. Established in 2015 and headquartered in Redwood City, California, WindBorne is among the pioneers in directly incorporating empirical data into a deep-learning weather prediction system.
While WindBorne has a history of collaboration with U.S. government entities, including the National Weather Service, Air Force, and Navy, this new agreement represents their most substantial contract to date. Valued at up to $49 million, the contract functions on an as-needed basis, allowing the Air Force to procure data for particular areas and periods. To date, orders totaling $1.7 million have been placed, with company leaders anticipating further requests.
Andrey Sushko, WindBorne’s co-founder and CTO, informed TechCrunch that this contract primarily serves as a means to facilitate the ongoing expansion of their balloon constellation and to fund research and development aimed at enhancing the hardware and performance of their balloons. He also noted that the contract was awarded non-competitively, indicating the government’s assessment that WindBorne was the sole provider capable of delivering the required global data collection for their operational needs.
In August, WindBorne’s CEO, John Dean, informed TechCrunch of a $37 million Series B funding round, secured to further scale the company’s operations. This past summer, the company unveiled WeatherMesh 6, the newest iteration of its forecasting model. According to WindBorne, this model delivers more precise forecasts than the conventional physics-based atmospheric simulations employed by the European Centre for Medium-Range Weather Forecasts (ECMWF).
With the advent of transformer model architectures and the European Centre for Medium-Range Weather Forecasts’ release of a substantial weather dataset several years prior, AI weather forecasting has emerged as a highly effective application of deep-learning technology. Previously, meteorologists depended on costly supercomputers to execute complex equations simulating atmospheric behavior. Today, end-to-end learning systems can process decades of weather data to generate predictions at an accelerated pace.
Forecast models developed by leading research laboratories and government bodies are demonstrating greater accuracy compared to conventional methods. However, this field remains dynamic, with both traditional and AI-driven forecasts essential for a comprehensive meteorological understanding. Notably, Google DeepMind’s most recent weather model introduces significant new functionalities, and its tools for cyclone prediction are playing an increasingly critical role in the U.S. government’s hurricane season preparedness.
Last August, WindBorne achieved a first by deploying an unpowered flight system – an engineless vehicle – to gather data directly within the eye of a typhoon. The company directs these balloons by altering their altitude, releasing air, or dropping sand ballast, thereby guiding them into specific wind currents for precise positioning.
Sushko remarked, ‘This represents a considerably innovative advance, reliant on two key factors: first, access to accurate wind and weather models coupled with the expertise to navigate them; and second, maintaining a sufficiently expansive constellation of balloons to ensure optimal placement.’ He further emphasized, ‘This marks one of the initial instances where a deep-learning model has been applied to real-world operational decision-making.’