WindBorne Systems has raised a $37 million Series B at a $250 million post-money valuation, betting that AI weather models can finally crack a private forecasting market that has resisted every earth-observation startup before it. Khosla Ventures and Galvanize co-led the round, with TransLink Capital, Lux Capital, and prior investors participating. CEO John Dean said the company plans to spend the money on compute, a mesh radio network to replace satellite communications on its balloon fleet, and a commercial go-to-market team.
WindBorne operates 20 launch sites around the world and keeps roughly 600 balloons in the air at any given time, collecting atmospheric data in places satellites struggle to observe — including the eye of a typhoon. The company is now deploying aerial sensor packages that fall into the ocean and continue transmitting as floating buoys, extending the sensor grid past the atmosphere.
Founded in 2019, WindBorne started as a data-collection business selling proprietary measurements to government meteorologists. The pivot into forecasting only became viable in the last four years, once deep-learning weather models made it possible to simulate the atmosphere without owning a supercomputer. The same architectural advances that produced modern LLMs now let a startup run global forecasts on commodity hardware.
Key facts
- 01WindBorne Systems raised a $37 million Series B co-led by Khosla Ventures and Galvanize, at a $250 million post-money valuation.
- 02The company operates 20 launch sites globally with roughly 600 balloons airborne at any moment, feeding data into its AI forecasting model.
- 03Existing customers include the U.S. National Weather Service, the U.S. Air Force, and the U.S. Navy.
- 04Founded in 2019, WindBorne began building its own forecasts once AI weather models emerged over the last four years.
- 05TransLink Capital, Lux Capital, and prior investors joined the round.
The proprietary balloon data set is the moat. WindBorne's forecasting model ingests public data from government weather agencies worldwide and layers its own measurements on top, and Dean argues each balloon-derived data point moves the forecast more than a satellite pixel does.
Government agencies are still the paying base. The U.S. National Weather Service buys WindBorne's data, and the U.S. Air Force and U.S. Navy fund the company through research partnerships — including work on forecasting models that can run onboard ships operating with intermittent connectivity.
The commercial pitch is harder. Over the last decade, satellite-imagery and sensor-network startups have raised large rounds on the premise that private-sector customers would eventually buy the data, and most ended up leaning back on government contracts because enterprise buyers lacked the workflows to turn raw sensor feeds into decisions. Existing private weather firms mostly repackage government forecasts for media, plane de-icing, ship routing, and commodity traders.
WindBorne's early commercial traction is with investment funds using weather signals to trade commodities — the customer segment most comfortable extracting alpha from raw data. Broadening beyond that means selling to industries that have never built a weather workflow before, which is where Galvanize's thesis comes in.
“integrating weather forecasts into broader business decision-making has traditionally been expensive and difficult. We think AI changes that equation. Better forecasts make the effort worthwhile, and AI makes it much easier to connect those forecasts to the decisions businesses are trying to make.”— Saloni Multani, Partner at Galvanize
That framing is the bet the round is underwriting: that generative AI turns weather data from a specialist input into something a mid-market operator can query in plain language and act on. If it works, the addressable market for private forecasting expands from a handful of trading desks and airlines to logistics, agriculture, insurance, energy, and construction.
The counterweight is execution risk. WindBorne still has to prove that its balloon data measurably improves forecasts at the scale enterprise customers care about, and it has to build a sales motion into industries with no existing budget line for AI weather. The satellite-network graveyard is full of companies that had better data than incumbents and still couldn't close commercial deals.
For the broader AI market, WindBorne is a useful data point on where domain-specific AI startups can defend a moat: not in the model, which is increasingly commodity, but in the physical sensor network feeding it. The companies that will monetize AI weather, AI seismology, AI oceanography, and every other physical-world domain are the ones that own proprietary measurement infrastructure the hyperscalers cannot replicate from a data center.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
