The U.S. Air Force has signed a contract with WindBorne Systems to purchase data collected by a network of approximately 700 autonomous weather balloons, with a maximum value of nearly $49 million. The agreement supports the expansion of the company’s system and the development of AI-based weather forecasting models.
The U.S. Air Force chose to purchase data collected by WindBorne Systems’ network of autonomous weather balloons under a contract worth up to $49 million. This does not mean that the company will automatically receive the full amount; the Air Force pays for the data when it requests it, and it has so far issued orders worth $1.7 million.
WindBorne, which was founded in 2015 and is headquartered in Redwood City, California, operates approximately 700 balloons that fly around the world and collect readings from atmospheric sensors. The company sells this data to its customers and also uses it to feed its own AI-based weather forecasting model.
A Flexible Contract to Support Expansion and Development
The contract allows the Air Force to request data from specific areas and at specific times. Andrej Suchko, WindBorne’s co-founder and chief technology officer, said the agreement would go a long way toward financing the growth of the balloon constellation, along with research and development work to improve the hardware and system performance.
According to Suchko, the contract was awarded without competition after the government concluded that WindBorne was the only company capable of providing global data collection in the manner suited to the needs of the operations concerned. The company also works with the National Weather Service and the U.S. Navy, but this is the largest contract in which it has participated so far.
AI Enters the Data-Collection Cycle Itself
WindBorne’s importance is not limited to using AI to analyze weather data. In August, the company became the first entity to send an unpowered aerial system into the eye of a tropical cyclone to collect data. The balloons control their paths by changing altitude, venting air, or dropping a payload of sand, allowing the winds to carry them toward the required locations.
This approach requires good wind and weather models, along with a network large enough to have balloons in the appropriate locations. Suchko said this was one of the first times a deep-learning model had been used to make actual operational decisions, rather than only to produce forecasts.
What Is Changing in Practice?
The deal illustrates the transition of AI weather models from experimentation and analysis to an operational role that combines sensing and decision-making. This summer, WindBorne launched the sixth version of its WeatherMesh 6 model, and says it provides more accurate forecasts than the traditional model operated by the European Centre for Medium-Range Weather Forecasts (ECMWF).
However, the source describes the field as still advancing unevenly, with traditional and AI models remaining necessary together to understand the weather in a useful way. The announced contract value is also a potential ceiling rather than fully guaranteed funding, while actual revenue depends on future purchase orders.