AI makes weather prediction better. Can WindBorne make it lucrative?
TechCrunch Tim Fernholz
WindBorne just raised $37 million to turn AI-powered weather forecasting into an actual business, not just better science. The startup uses balloon data and AI models cheap enough to run without supercomputers — and now it wants to sell that to companies, not just governments.
Based on reporting by TechCrunch, Tim Fernholz — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Weather forecasting has quietly become an AI story. The same deep learning breakthroughs that gave us chatbots have also shrunk atmospheric simulation from supercomputer-only territory to something that can run on a laptop. WindBorne Systems is betting that the harder problem isn't generating better forecasts anymore — it's getting people to actually use them.
The company just closed a $37 million Series B, co-led by Khosla Ventures and Galvanize, with TransLink Capital, Lux Capital, and existing backers also participating. That puts WindBorne's valuation at $250 million. Founded in 2019, the company built its business around long-flying weather balloons and cheap sensors, collecting data from places satellites and ground stations struggle to reach, like the interior of a typhoon. CEO John Dean says roughly 600 of these balloons are aloft at any given moment, launched from 20 sites worldwide, and the company is now adding sensor packages that drop into the ocean and keep transmitting as floating buoys.
What changed recently is that WindBorne no longer just gathers data for other forecasters. New AI weather models let it run its own forecasts, something that used to require the kind of supercomputing budget only national weather agencies could afford. Dean argues the balloon data itself becomes a moat here, layered on top of government datasets already in wide use, and that each balloon-collected data point outperforms satellite data on accuracy. He also credits growing revenue with making the funding round an easier sell to investors, since it showed real demand rather than just technical promise.
Right now, that demand comes almost entirely from governments. The National Weather Service buys WindBorne's data, and the Air Force and Navy are funding research partnerships, including work on forecasting systems that could run aboard ships with patchy connectivity. The next test is commercial customers, starting with investment funds that trade on weather-driven swings in commodity prices. WindBorne plans to use the new funding for compute, for replacing satellite links in its balloon network with a mesh radio system, and for building a sales team aimed at the private sector.
That private-sector push is the part with no guaranteed outcome. Plenty of earth-observation startups, satellite companies among them, have struggled over the past decade to move beyond government contracts, because turning raw data into business decisions takes workflows and expertise that most companies don't have in-house. The private weather firms that do exist mostly repackage government forecasts for media outlets, aviation de-icing, shipping routes, or trading desks. Galvanize partner Saloni Multani, who co-led the round, thinks AI is what finally makes that integration cheap enough for ordinary businesses to bother with — better forecasts paired with tools that make acting on them less of a specialist job.
My take — AI-written commentary, not fact-checked reporting
The interesting bet here isn't the balloons, it's the assumption that AI removes the real bottleneck in weather data, which was never accuracy but distribution. Plenty of hardware-heavy data startups have learned the hard way that governments will pay for novelty while private companies want turnkey answers, not raw feeds. If WindBorne actually cracks that commercial workflow problem instead of just collecting cooler data, that's the story worth watching — not another balloon.
Read more about this at: TechCrunch
