The U.S. Air Force has awarded WindBorne Systems a contract valued at up to $49 million to purchase atmospheric data gathered by the startup's fleet of autonomous weather balloons, according to a report published October 12, 2026 by TechCrunch. The agreement represents the single largest contract the Redwood City, California-based company has secured since its founding in 2015. WindBorne operates roughly 700 weather balloons worldwide that collect sensor readings, which the company sells to clients and also uses to power its own artificial intelligence-driven weather prediction system.
The Air Force contract functions as a pay-as-you-go arrangement, with the military branch ordering data for particular geographic areas at particular times rather than paying the full amount upfront. So far, the service has placed orders totaling $1.7 million, with company leaders anticipating additional purchases ahead. The agreement was granted without competitive bidding, reflecting what the government determined was WindBorne's unique capability to deliver global atmospheric data matching operational requirements. WindBorne is among the first entities to directly incorporate empirical measurements into a deep-learning weather model, and it has maintained close relationships with federal agencies including the National Weather Service and both the Air Force and Navy. In August, the company completed a $37 million Series B funding round to expand its operations.
Co-founder and CTO Andrey Sushko explained that the contract serves "largely as 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 noted that the government's decision to award the contract without competition reflected "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 released its newest forecasting model, WeatherMesh 6, earlier this summer, claiming it delivers more precise predictions than the conventional model operated by the European Centre for Medium-Range Weather Forecasts, which relies on physics-based atmospheric simulations.
AI weather forecasting has emerged as one of the more productive applications of deep-learning technology since transformer model architectures appeared and the ECMWF published an extensive weather dataset several years ago. Traditional meteorology depended on costly supercomputers running complex equations that approximate atmospheric behavior, but end-to-end learning can now digest decades of historical weather information and generate forecasts more rapidly. Models developed by leading research labs and government bodies are demonstrating greater accuracy than conventional predictions, though both approaches remain necessary for comprehensive weather understanding. Google DeepMind's most recent weather model has introduced important new functions, and its hurricane prediction tools are becoming increasingly vital to how the federal government manages storm season. WindBorne steers its balloons by changing altitude, releasing air, or discarding sand ballast to position the craft in wind currents that carry them to desired locations. In August, the company became the first to fly an unpowered system—one lacking an engine—into a typhoon's eye to gather data. Sushko described it as "one of the first times that a deep-learning model was used for actual operational decisions," explaining that success requires both effective wind and weather models plus a sufficiently large balloon constellation with craft positioned correctly.
The contract will support expansion of WindBorne's balloon network while funding hardware improvements and performance enhancements, positioning the company to deliver more comprehensive atmospheric coverage as military and civilian demand for precise weather intelligence grows. The combination of empirical balloon data and machine learning represents a shift from purely physics-based forecasting toward hybrid approaches that blend real-world measurements with algorithmic pattern recognition. The Air Force's willingness to structure the agreement as an on-demand service rather than a fixed purchase suggests federal weather data needs are becoming both more targeted and more urgent. Companies that can position sensors in hard-to-reach locations stand to capture significant government spending as extreme weather events intensify and traditional forecasting gaps become more costly.