The US Air Force has awarded WindBorne Systems a contract that can grow to $49 million for data from the company's autonomous weather balloon fleet. The Air Force has placed $1.7 million in orders under the deal so far, and WindBorne says it is the largest government contract the company has won.
WindBorne's Air Force Contract Can Reach $49 Million
The award is structured as a ceiling, not a guaranteed payout. WindBorne co-founder and CTO Andrey Sushko told TechCrunch the funding is "a vehicle through which continued growth of the constellation can be supported," meaning the Air Force is paying to help the company grow its balloon fleet and keep buying data as that fleet expands. WindBorne, based in Redwood City, California and founded in 2015, raised a $37 million Series B round that CEO John Dean disclosed to TechCrunch in August. The Air Force contract is roughly 30% larger than that entire funding round.
The Air Force Buys Data, Not Balloons
The contract does not pay for hardware. It is a pay-as-needed arrangement: the Air Force places orders for weather data covering specific regions and time windows, and WindBorne fills them from its fleet of roughly 700 autonomous balloons already aloft. Sushko described the award as a government determination about the value of WindBorne's capability to collect global weather data.
A Deep-Learning Model Already Steered a WindBorne Balloon
The most consequential detail in the deal is operational, not financial. WindBorne's balloons adjust their altitude to ride different wind layers and steer toward target regions, and Sushko said one of those in-flight adjustments was recently made by a deep-learning model rather than a traditional rules-based system or a human controller. He called it one of the first times "a deep-learning model was used for actual operational decisions" in the company's balloon operations. That detail matters beyond WindBorne. A defense-connected organization is now accepting an AI model's real-time judgment call over physical hardware in the field.
The Contract Was Awarded Without Competitive Bidding
The Air Force made the award without a competitive bidding process. That is not unusual for specialized data-collection capabilities the government has already evaluated, but it means no rival weather-data provider had the chance to bid against WindBorne's price or performance claims for this specific contract. Combined with the AI-driven balloon maneuver, it puts a fairly novel question in front of military and civilian procurement officers alike: how to evaluate a vendor whose core product now includes decisions made by a model as well as measurements made by a sensor.
WindBorne's Balloon Data Feeds Its Own AI Forecasting Model
WindBorne doesn't just sell raw balloon readings. The company feeds that data into its own deep-learning weather model, called WeatherMesh 6, which it has previously supplied to the National Weather Service and the U.S. Navy. That puts WindBorne in the same general field as AI-driven forecasting work from Google DeepMind and the European Centre for Medium-Range Weather Forecasts (ECMWF), both of which have built their own machine-learning weather models in recent years. The Air Force contract extends that same data-to-model pipeline into defense weather forecasting specifically.
Why This Matters for Engineering Leadership
This deal is a preview of a procurement problem that will show up well outside the Air Force. When a vendor's product includes a model making real-time operational calls, the usual procurement checklist, price, uptime, data accuracy, stops covering what actually matters. Engineering and procurement leaders evaluating any AI-driven data or sensing vendor should be asking who can replay a specific automated decision after the fact, what the model's error cases look like under pressure, and whether the vendor can produce an audit trail for the exact moment an algorithm made a call with physical consequences. WindBorne's sole-source, ceiling-based contract structure, paying for capacity and data on demand rather than a fixed deliverable, is also a template worth watching: it lets a buyer scale spending with an AI vendor's growth without locking in a fixed scope up front, but it also means less public benchmarking data ever gets created for competitors or outside reviewers to check.
FAQ
How much is WindBorne's new Air Force contract worth?
The contract has a ceiling of up to $49 million. The Air Force has placed $1.7 million in actual orders against it so far.
What does WindBorne sell to the Air Force under this deal?
Weather data collected by WindBorne's fleet of roughly 700 autonomous balloons, ordered on demand for specific regions and time windows, rather than balloon hardware itself.
Was the contract competitively bid?
No. The Air Force awarded it without a competitive bidding process, according to WindBorne CTO Andrey Sushko.
What makes this contract notable beyond its size?
WindBorne says a deep-learning model recently made an actual in-flight operational decision for one of its balloons, one of the first times the company has let an AI model, rather than a human or rules-based system, make that kind of real-time call.
Is this WindBorne's biggest government contract?
Yes. The company describes it as the largest government contract it has won to date, larger than the $37 million Series B it raised, as CEO John Dean disclosed in August 2026.
Who else buys WindBorne's weather data or competes in AI weather forecasting?
WindBorne has previously supplied its WeatherMesh 6 model's output to the National Weather Service and the U.S. Navy. Google DeepMind and the European Centre for Medium-Range Weather Forecasts have each built their own AI-driven weather forecasting models.