Air Force Could Spend $49M on WindBorne Balloon Weather Data

What the Air Force is actually buying

Not balloons, and not a subscription to a forecast. The Air Force is buying measurements.

Under the agreement, the military can request atmospheric readings for specific places and time windows rather than committing to a fixed batch of data upfront. That flexibility is the point: a $49 million ceiling is not a $49 million check.

So far, orders total roughly $1.7 million, according to TechRadar's reporting. The rest depends on what the Air Force decides it needs — which makes this closer to a framework for future purchases than a headline contract.

The deal was awarded without competing bids. WindBorne co-founder and CTO Andrey Sushko told TechCrunch the government concluded the company was the only one able to deliver global data collection in the form the Air Force required. Whether that stays true as rivals scale up is an open question.

How a balloon without an engine goes where you want it

WindBorne's balloons don't fly to a target. They sink and rise until the wind carries them there.

Each balloon adjusts its buoyancy by releasing air or dropping sand ballast, changing altitude to catch winds blowing in a useful direction. Software models predict where those winds will be, and the fleet is large enough — about 700 balloons — that some are usually in position to be steered toward a target.

That combination paid off in August 2026, when WindBorne guided an unpowered balloon into the eye of a typhoon and collected data there. Sushko called it a novel step and pointed to the two ingredients: good wind models and a constellation big enough to work with.

He also framed it as a milestone for machine learning in operations: "It's one of the first times that a deep-learning model was used for actual operational decisions." That is the company's characterization, not an industry-wide verdict.

The forecast model claim needs outside testing

WindBorne says WeatherMesh 6 outperforms the European Centre for Medium-Range Weather Forecasts. That claim is the company's own.

The ECMWF's approach simulates atmospheric physics directly. Machine-learning systems like WeatherMesh 6 instead learn patterns from historical observations and generate forecasts from them — faster and cheaper to run, but dependent on the quality of the data they were trained on.

Neither method removes the need for real measurements. Every forecast, physics-based or learned, starts from a picture of the current atmosphere, and that picture is only as good as the sensors feeding it. This is where the balloon fleet and the Air Force money connect: better inputs, potentially better outputs.

Independent comparisons across seasons and regions are what would settle the question. Until those exist, treat the performance edge as a claim worth watching rather than a settled result.

What it means for you

This is a US government procurement, so there is no consumer product to buy and no price tag for readers in the US, UK or EU. The practical effect is indirect: more balloon measurements feeding forecasting systems could eventually improve the models behind the weather apps you already use.

WindBorne sells data to paying customers, but it has not published consumer pricing or a public availability date for its forecasts. The Air Force agreement itself is US-only.

The contract ceiling is the number that gets quoted; the order book is the number that tells you what is happening. For now, the Air Force is buying small, testing the product, and keeping a $49 million option open. WindBorne's job is to make that option worth exercising.