AI makes weather prediction better. Can WindBorne make it lucrative?
TechCrunch Tim Fernholz
WindBorne, which flies giant weather balloons and runs its own AI forecasts, just raised $37M at a $250M valuation. AI cut the cost of weather modeling so much that a startup can now out-forecast satellites with balloons.
WindBorne Systems started in 2019 with a fairly niche pitch: build cheap, long-endurance weather balloons and gather data that nobody else bothers to collect, like conditions inside a typhoon's eye. That was the whole business. But over the past four years, deep learning has rewritten what's possible in atmospheric science, and a forecasting model that once demanded a room full of supercomputers can now run on hardware a small company can actually afford. That shift turned WindBorne from a data-collection outfit into a full forecasting company, and it just landed a $37 million Series B, co-led by Khosla Ventures and Galvanize, that values the business at $250 million.
The company now keeps roughly 600 balloons aloft at any moment, launched from 20 sites worldwide, and CEO John Dean calls the resulting network a 'planetary nervous system.' The pitch to investors is straightforward: balloon data, fed into WindBorne's own model alongside government feeds, produces measurably better forecasts than satellite data alone, and each data point is worth more because it fills gaps satellites can't reach. WindBorne is also pushing into the ocean, deploying sensor packages that fall from the sky and keep transmitting as floating buoys, and it's working to swap out satellite links for a mesh radio network to cut costs further.
Right now the customer list is dominated by governments. The National Weather Service buys WindBorne's data outright, while the Air Force and Navy fund research partnerships, including one aimed at building forecasting models that ships can run without a steady internet connection. That's the safe, proven revenue. The harder problem, and the one this funding round is really aimed at, is getting private companies to pay for weather intelligence the way governments already do.
History isn't especially kind to that ambition. Earth-observation satellite startups spent the last decade learning that collecting good data is the easy part; getting businesses to actually change decisions based on it is the hard part, because most private forecasting firms just repackage government data for niche uses like plane de-icing or shipping routes. WindBorne is betting AI closes that gap by making it cheaper to turn raw forecasts into usable business signals, which is why its early commercial customers are investment funds trading on commodity prices tied to weather. Galvanize partner Saloni Multani put it plainly: integrating forecasts into business decisions used to be expensive and clunky, and AI is what finally makes the math work.
Whether that thesis holds will depend less on balloon count and more on whether WindBorne can build the sales muscle to sell forecasts to people who've never bought them before. That's the part of this round earmarked for a go-to-market team, and it's arguably a bigger bet than the hardware ever was.
My take
The technology here is genuinely cool, but the real story is a startup betting that AI's cheapening of compute finally unlocks a market that satellite companies spent a decade failing to crack. Selling data to governments is easy money; convincing a hedge fund or a shipping company to trust your forecast over the National Weather Service's free one is a sales problem, not an engineering one. If WindBorne pulls it off, expect a wave of copycats trying to turn niche sensor networks into subscription businesses — and expect most of them to underestimate how hard that pivot actually is.
Read more about this at: TechCrunch
