AivexaNewsSearch
AI news for builders and product teamsChecked every hour

WindBorne wins a hefty US Air Force contract for its balloon data

Collected Oct 12, 2026

WindBorne Systems, a Redwood City, California startup founded in 2015, has won a U.S. Air Force contract worth up to $49 million to supply atmospheric data collected by its fleet of autonomous weather balloons. The award was made without competitive bidding.

The Air Force is not paying the full amount up front. Instead, it pays as it needs data, placing orders for specific regions at specific times. So far it has ordered $1.7 million worth, and company executives expect more orders to follow. Co-founder and CTO Andrey Sushko described the agreement as a vehicle for supporting continued growth of the balloon constellation as well as research and development work to improve the hardware and performance of the balloons. He said the sole-source award reflects a government determination that WindBorne was the only company able to provide this kind of global data collection in a way that met its operational needs.

WindBorne currently operates about 700 weather balloons flying around the globe. The balloons collect atmospheric sensor readings that the company sells to customers and also feeds into its own AI weather forecasting model. The company is one of the first organizations to feed empirical data directly into a deep-learning weather model.

This is the largest single contract WindBorne has taken part in. The company has worked closely with the U.S. government before, particularly the National Weather Service and the U.S. Air Force and Navy.

WindBorne has been scaling its business. CEO John Dean said in August that the company had raised a $37 million Series B round. Earlier in the summer it released WeatherMesh 6, the latest version of its forecasting model, which the company says produces more accurate forecasts than the traditional model run by the European Centre for Medium-Range Weather Forecasts. The ECMWF model simulates the atmosphere using the laws of physics.

WindBorne also reported a notable operational milestone in August: it became the first organization to send an unpowered flight system, one with no engine, to collect data inside the eye of a typhoon. The balloons are steered by adjusting altitude, venting air, or dropping sand carried as ballast, which lets them ride winds that carry them toward their targets. Sushko said the feat depended on two things: good wind and weather models coupled with the knowledge of how to navigate them, and a constellation large enough that balloons are already positioned in the right places. He called it one of the first times a deep-learning model was used for actual operational decisions.

The contract lands in the middle of a broader shift in weather forecasting. Since transformer architectures arrived and the ECMWF released a large weather dataset a few years ago, AI forecasting has become one of the more productive applications of deep learning. Meteorologists previously depended on expensive supercomputers solving detailed equations that approximate atmospheric behavior. End-to-end learning can now ingest decades of weather data and issue predictions much faster. Models from frontier labs and government agencies are proving more accurate than traditional forecasts, though the source notes this remains a jagged frontier in which both approaches are still needed for a useful picture of the weather. Google DeepMind's latest weather model offers important new capabilities, and its cyclone forecasting tools are playing a growing role in how the U.S. government responds to hurricane season.

For developers and product teams, the interesting thread here is the data pipeline rather than the dollar figure. WindBorne's model is trained on observations its own hardware gathers, which is a different setup from models that rely solely on historical reanalysis datasets. In practice, owning the sensing layer means owning a stream of fresh observations that competitors cannot buy, and it creates a feedback loop: better forecasts make it easier to steer balloons into useful locations, which in turn produces better data. That closed loop is difficult to replicate without a fleet, and it suggests why the Air Force concluded that no other vendor could meet its requirements.

The pay-as-you-go structure is also worth noting. A ceiling of $49 million is not the same as $49 million in revenue; it is an option the government can draw on. That gives the Air Force flexibility to task coverage for particular regions and windows, and it gives WindBorne a predictable channel that can justify expanding the constellation. A likely trade-off is that demand is lumpy and tied to government priorities, so commercial customers and forecasting contracts still matter for smoothing utilization.

Several open questions remain. Whether WeatherMesh-class models can consistently outperform physics-based systems across all conditions, rather than on select benchmarks, is still an active area of comparison. The August typhoon flight shows unpowered balloons can reach extreme environments, but scaling that reliability across a global fleet is a separate engineering problem. And sole-source awards can invite scrutiny even when they are justified on capability grounds. The company's own framing is that the contract supports hardware and performance work, which implies the constellation and its balloons are still evolving rather than finished.

Why it matters: A nearly $50 million ceiling contract signals that a private constellation of instrumented balloons is now treated as operational infrastructure for U.S. military weather needs, not an experiment. For developers building on weather data, it reinforces that observation networks and learned forecast models are converging into a single product, and that access to proprietary sensor feeds may become a competitive dividing line. Product teams that depend on forecast accuracy should expect continued improvement from learned models, but also continued uncertainty about how quickly they replace physics-based ones.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

A Silicon Valley startup that uses weather balloons and AI to predict the weather won a U.S. military contract worth almost $50 million.