The U.S. Air Force will buy data collected by Windborn Systems’ fleet of autonomous weather balloons under a new contract said to be worth nearly $50 million to the company.
WindBorne currently flies approximately 700 weather balloons around the world, collecting atmospheric sensor readings that it sells to customers and feeds into its proprietary AI weather prediction models. Founded in 2015 and based in Redwood City, California, WindBorne was one of the first organizations to feed empirical data directly into deep learning weather models.
The company has worked closely with the U.S. government, particularly the National Weather Service, and the U.S. Air Force and Navy, and this is the largest single contract the company has participated in. The contract is worth up to $49 million, but the Air Force will only pay if it needs the data. You can order specific areas at specific times. So far, orders worth $1.7 million have been placed, and executives expect more to follow.
“This is primarily a way to support the continued growth of the balloon, and also work on research and development to continually improve the balloon’s hardware and performance,” WindBorne co-founder and CTO Andrey Sushko told TechCrunch. He added that the contract was awarded without competitive bidding and was “a government decision that we are the only company that can deliver this capability to collect global data in a way that meets the needs of these operations.”
WindBorne CEO John Dean told TechCrunch in August that the company had raised $37 million in a Series B round to continue expanding its business. Earlier this summer, the company released WeatherMesh 6, the latest version of its predictive model. It says it provides more accurate predictions than previous models run by the European Center for Medium-Range Weather Forecasts (ECMWF), which use the laws of physics to simulate the atmosphere.
Since the introduction of transformer model architectures a few years ago and ECMWF releasing large-scale weather datasets, AI weather forecasting has become one of the more productive applications for deep learning technology. Meteorologists have traditionally relied on expensive supercomputers that run detailed equations that approximate the behavior of the atmosphere. Now, with end-to-end learning, we can ingest decades of weather data and issue forecasts faster.
While models built by Frontier Institute and government agencies are proving more accurate than traditional forecasts, it’s still a jagged frontier, and both types of forecasts are needed to meaningfully understand the weather. Google DeepMind’s latest weather models offer important new capabilities, and its cyclone forecasting tools are playing an increasingly central role in how the U.S. government responds to hurricane season.
In August, WindBorne became the first organization to send an unpowered flight system (a flight system without an engine) to collect data in the eye of a typhoon. The company steers the balloons by adjusting their altitude, removing air and dropping sand carried as ballast, sending vehicles into the wind and carrying them to their destinations.
“This is a pretty novel step, and the two ingredients for that are: one, we have good wind and weather models and know how to navigate them, and two, we have a constellation large enough that we can put the balloons in the right places,” Sushko said. “This is the first time a deep learning model has been used for actual operational decisions.”
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
