Aeolus is building a planetary defense system for extreme weather using atmospheric manipulation
May 27, 2025 with Koki Mashita
Key Points
- Aeolus is positioning itself to fill a gap in U.S. atmospheric infrastructure by collecting proprietary storm data via instrumented aircraft, arguing that granular internal measurements from hurricanes will yield a durable edge over rivals focused on model architecture alone.
- The founder frames the business opportunity around intervention capability rather than data licensing, citing the $140 billion annual tropical cyclone insurance market and the reality that Congress spent $100 billion in disaster aid for 2024 storms that weren't Category 5.
- Aeolus traces its technical credibility to prior work as the Philippines' national weather forecast provider, where the team cut supercomputing costs by 99% using machine learning while increasing atmospheric model resolution 100-fold.
Summary
Aeolus, founded by a Teal Fellow whose family in the Philippines has weathered typhoon damage for generations, is building what it describes as a planetary defense system for extreme weather — not just better forecasting, but active atmospheric intervention. The founder is deliberately vague about the intervention mechanism, but the direction is clear: collect atmospheric data no one else has, model it more precisely, and then act on it causally.
The data collection strategy is where Aeolus diverges from incumbents like DeepMind or Nvidia, which are focused on improving model architectures. Aeolus argues that software and models are increasingly commoditized, so the durable edge is proprietary sensor data gathered inside storms. That means physically flying instrumented aircraft into hurricanes and convective clouds to measure internal dynamics — cloud density, water droplet columns, convection patterns — that satellites and ground radar can't capture. Drones are ruled out as too fragile for the turbulence.
The forecasting foundation traces to a prior venture where the team became the national weather forecast provider for the Philippines, using ML to cut the cost of running atmospheric partial differential equations by 99% while increasing resolution 100-fold. Countries currently spend roughly $2 billion per year on supercomputing just to run those equations. The commercial logic of that previous work is what convinced the founder that the bigger opportunity is acting on atmospheric intelligence, not just selling it.
The market framing
The financial case rests on the dysfunction of the current reactive model. The U.S. pays roughly $140 billion per year in tropical cyclone insurance premiums, and coastal premiums have doubled every year for the past four to five years. When Hurricane Helene and Milton hit in 2024, Congress issued $100 billion in disaster aid — and those weren't even Category 5 storms. The founder's argument is that a direct hit on a major coastal city at Category 5 intensity would render insurance insolvent across large swaths of the market.
Selling better forecast data alone faces real commercial limits. Insurers price risk using 40-year historical averages with a bias adjustment for intensification — more granular forward-looking data doesn't easily translate into premium repricing in the weeks before a storm. Catastrophe bonds and the funds that trade them are a more natural buyer, since their entire business is arbitraging expected versus actual disaster loss, but the broader data market is thinner than it might appear. The founder implies the defensible business is intervention, not data licensing.
Aeolus is early-stage and the founder declines to detail the intervention technology or current revenue. The company's NOAA reference is pointed: NOAA, the traditional source of reanalysis and satellite data, is described as "not doing too well now," which leaves a gap in national atmospheric infrastructure that Aeolus appears to be positioning to fill, at least partly.