Assessment: opportunity. The U.S. Air Force will buy atmospheric data from WindBorne Systems’ fleet of autonomous weather balloons under a contract worth up to $49 million, TechCrunch AI reports. It’s the largest single contract the Redwood City startup has taken part in. It also shows that the military now treats AI weather forecasting as working infrastructure, not just a research project.
📡 Situation Report
Here’s what’s on the table:
- Contract ceiling: Up to $49 million. The Air Force pays only when it needs data.
- How orders work: The Air Force can order coverage of specific regions at specific times.
- Orders so far: $1.7 million. Executives expect more to come.
- How it was awarded: No competitive bidding. WindBorne was the sole source.
- Fleet size: About 700 balloons flying around the globe right now.
Co-founder and CTO Andrey Sushko told TechCrunch the contract is “largely a vehicle through which continued growth of the constellation can be supported,” with some funding going to R&D on balloon hardware and performance. He called the sole-source award “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.”
🎯 Why It Matters
WindBorne doesn’t just sell balloon data. It also feeds that data into its own deep-learning forecast model, WeatherMesh 6. According to TechCrunch AI, it’s one of the first organizations to feed real-world sensor readings straight into a deep-learning weather model.
That matters because good forecasting models are getting easier to find. Frontier labs and government agencies are shipping models that beat traditional forecasts. Owning fresh observations from places other networks don’t reach is much harder to copy, and that’s what the Air Force is paying for.
Since its founding in 2015, WindBorne has also worked with the National Weather Service and the Navy. Earlier this summer it released WeatherMesh 6, which it says beats the physics-based model run by the European Centre for Medium-Range Weather Forecasts (ECMWF). CEO John Dean told TechCrunch in August that the company had raised a $37 million Series B to keep scaling.
🧭 How We Got Here
For decades, meteorologists ran detailed physics equations on expensive supercomputers to estimate how the atmosphere behaves. Two things changed that:
- Transformer architectures gave researchers models that can learn patterns across huge datasets.
- ECMWF’s release of a massive weather dataset a few years ago gave those models decades of history to learn from.
Together they made AI weather forecasting one of the most productive uses of deep learning. These models learn end to end from decades of data and produce forecasts faster than the old approach.
The old approach isn’t dead, though. TechCrunch AI calls it “a jagged frontier”: forecasters still need both AI and physics-based models to understand the weather well. Google DeepMind’s cyclone tools, for example, are becoming central to how the U.S. government handles hurricane season.
🌀 Proof in the Field
In August, WindBorne became the first organization to send an unpowered flight system into the eye of a typhoon to collect data. The balloons have no engines. Operators steer them by:
- Changing altitude
- Letting out air
- Dropping sand ballast
Each move puts a balloon into a different wind current that carries it toward its target. Sushko said it takes two things: good wind and weather models you know how to navigate, and “a large enough constellation that you have balloons in the right places.”
He added: “It’s one of the first times that a deep-learning model was used for actual operational decisions.”
This is the part that stands out to me. Here the AI model does more than predict the weather. It decides where the balloons fly, those balloons gather new data, and the new data improves the model. Each part makes the other better.
⚠️ Watch List
- More orders. Only $1.7 million of the $49 million has been spent. How fast the rest gets ordered will show how much the Air Force really relies on WindBorne.
- Fleet growth. Sushko says the contract will help pay to grow the constellation. More balloons mean better coverage, and better coverage makes the data worth more.
- Independent checks. WeatherMesh 6’s edge over ECMWF comes from WindBorne itself. Outside benchmarks will matter.
- Copycats. A sole-source defense contract tells everyone there’s a market for exclusive sensor data. Expect other AI weather players to look harder at collecting their own observations.
WindBorne is betting that whoever owns the data will win in AI weather, not just whoever has the best model. The Air Force just put its money behind that bet. Full details are available at TechCrunch AI.