Better Forecasts Are the Easy Part

WindBorne just raised $37M at a $250M valuation. Its real challenge isn't predicting the weather, it's selling the prediction.

Welcome Back,

The new deep learning techniques behind LLMs have quietly rewritten meteorology too, giving us weather simulations that can run on a laptop instead of a supercomputer, according to TechCrunch. But the more interesting question isn't whether AI can forecast the weather better, it increasingly can, it's whether anyone can actually turn that better forecast into a business.

WindBorne Systems is betting yes. The startup, which collects atmospheric data with the world's longest-flying weather balloons and feeds it into its own forecasting model, just closed a $37 million Series B, co-led by Khosla Ventures and Galvanize, valuing the company at $250 million.

Today we look at how WindBorne actually builds its data advantage, why government agencies remain its bread and butter, the graveyard of failed "sensor data" startups this company has to avoid becoming, and why AI might finally be the thing that opens up the private weather market.

🎈 A Planetary Nervous System, Built From Balloons

WindBorne was founded in 2019 with a plan to acquire a novel set of weather data using low-cost sensors and endurance balloons. The company now has 20 launch sites around the world and roughly 600 balloons in the air at any given time, gathering data from places traditional weather infrastructure simply doesn't reach, including the eye of a typhoon.

The AI weather forecasting models that emerged over the last four years changed what WindBorne could actually do with that data. Previously, running full atmospheric simulations required supercomputers most private companies couldn't afford. Deep learning models made that kind of forecasting newly accessible, and WindBorne pivoted from just selling data to building its own model on top of it.

That proprietary dataset, what Dean calls a "planetary nervous system," is the moat behind WindBorne's model, which also ingests data from government weather agencies worldwide. The company is now deploying aerial sensor packages that can fall into the ocean and keep collecting measurements as floating buoys, extending coverage even further into the gaps other systems can't reach.

WindBorne's balloons and sea-based sensors reach places traditional weather infrastructure can't, including storm systems forming over open ocean.

🏛️ Government Agencies Are Still the Bread and Butter

For all the talk of a new commercial market, WindBorne's current customer base is overwhelmingly public sector. The U.S. National Weather Service purchases the company's data outright, while the U.S. Air Force and U.S. Navy are paying WindBorne through research partnerships, including an effort to build forecasting models that can run onboard ships with intermittent connectivity to the rest of the world.

That's a meaningfully different growth story than most AI startups tell. Rather than chasing consumer scale or enterprise logos first, WindBorne built its case on institutions that already understood weather data's value and had budgets set up to pay for it, then used that traction to fund the harder, unproven part of the business.

📉 The Graveyard WindBorne Has to Avoid

Here's the part of this story that gives it real stakes. In the last decade, a variety of startups tried to scale up sensing businesses like earth-observing satellite networks, and found it genuinely difficult to break through to the private sector, because extracting value from raw sensor data requires experience and established workflows most commercial buyers simply don't have. Most of those companies ended up selling almost entirely to government agencies already used to working with that kind of data.

That's the exact trap WindBorne is trying to route around, and it's precisely why this funding round matters more than the dollar figure alone suggests. Part of the new capital is earmarked specifically for building out a go-to-market team focused on expanding into the private sector, alongside continued spending on compute and an effort to replace the balloon network's satellite communications with a mesh radio network.

Investment funds using weather data to predict commodity prices are WindBorne's first serious commercial customer segment.

💹 Who Actually Pays for a Better Forecast

WindBorne's initial commercial push is focused mainly on investment funds that use weather data to predict commodity prices and other business outcomes, a genuinely lucrative but narrow slice of the private sector. That's a sensible starting point, financial firms already have the analytical infrastructure and the direct financial incentive to pay for a marginal edge in forecast accuracy.

Beyond finance, existing private weather forecasting companies mostly make their money in a handful of specific ways, repackaging or refining government forecasts for news media, meeting specialized operational needs like aircraft de-icing decisions and ship routing, or serving the same kind of commodity speculators WindBorne is now targeting directly.

Multani's point cuts to the real thesis behind this round. The forecast itself was never really the bottleneck for most businesses, the ability to actually wire that forecast into a supply chain decision, a pricing model, or an insurance calculation was. That's the layer AI is arguably better positioned to solve than the atmospheric physics ever was.

🌪️ A Crowded, Increasingly Serious Field

WindBorne isn't operating in a vacuum. Google DeepMind's GraphCast, trained on 40 years of weather data, was found to outperform the European Centre for Medium-Range Weather Forecasts, long considered the gold standard in weather modeling. WindBorne has said its own WeatherMesh model has since surpassed GraphCast's key benchmarks, and the company's most recent version, WeatherMesh-2, claims to beat both AI and physics-based gold standards on longer time-horizon forecasts.

That competitive field is exactly why the go-to-market spending in this round matters as much as the modeling improvements. A better forecast is a defensible technical edge for a while, but not indefinitely, several serious, well-capitalized competitors are chasing similar accuracy gains at the same time.

The competition to build the best AI weather model has grown crowded fast, from tech giants to well-funded startups.

🧠 AI Spotlight Analysis

The headline of this story, "AI makes weather prediction better," undersells what's actually interesting here. The forecasting breakthrough is genuinely real, but WindBorne's harder problem was always going to be commercial, not technical. Plenty of companies have built impressive sensing or prediction technology and then struggled for years to find paying customers outside of government.

What makes this round worth watching is that WindBorne seems to understand that distinction clearly. Rather than treating better forecasts as automatically self-selling, the company is explicitly investing in the go-to-market muscle needed to translate accuracy into revenue, precisely the step where earlier sensing startups fell short.

That's the honest takeaway. AI has genuinely changed what's technically possible in weather forecasting. Whether it also changes what's commercially possible, turning better predictions into a business that reaches beyond government contracts, is the actual experiment WindBorne is now running with $37 million in fresh capital.

🔗 Sources and Further Reading

Thanks for reading AI Spotlight.

Our mission is simple: deliver clear, trustworthy, and actionable AI insights that help professionals stay ahead without the hype.