Seeing Silicon | How to fly a balloon into a hurricane

It’s a sunny evening with a soft autumn chill in an empty concrete lot in Palo Alto. Scattered around me are 40-odd adults and a few children standing next to the office of Windborne Systems, an AI weather forecasting startup. We’re all here for an unusual product launch – a weather balloon. It’s dramatically different for Silicon Valley, where most product launches are intangible software. Sending balloons up in the air to track weather through AI? That sounds like real science.
The product launch feels like a neighbourly get-together. There’s a kiosk in the far corner which offers free pizza from Costco, fizzy drinks and chips. Geeks of mixed ages converse enthusiastically about the physics of balloons. Two five-year-olds in front of me discuss intently about the time the balloon will go up in the sky. Twenty feet ahead of them, stands WindBorne Systems co-founder and CTO Andrey Sushko, in a grey t-shirt and jeans, busy setting up Version 7 of a homemade contraption to launch the weather balloons. In front of him hangs a white balloon, the size of two cupboards, being pumped with helium. Five minutes later, it releases from the pulley floating up into the sky, to clapping, hoots and a collective cry of ‘So Say We All’.
The line ‘So Say We All’, is from American TV series Battlestar Galactica and pays homage to the start of the balloon project in Stanford Student Space Initiative (SSSI). Sushko built the balloon that became WindBorne Systems as a math and physics undergrad at Stanford in 2015. He met Windborne’s three other co-founders at SSSI and after joining an accelerator, tumbled into the beginnings of the startup. “I wanted to launch satellites but instead started working on near space balloons with Sushko,” laughs John Dean, co-founder and CEO, who tells me this story after the launch. “I thought I’d go into the space industry eventually.” But five years after starting the company, Dean’s heading the organisation and still focused on balloons. By the time they graduated, they realised that they’d solved a major problem.
Weather observations across the world, come from three places – ground observation, upper air and space. Upper air, which is where the weather balloons fall, is the biggest gap. Around the world every day, a thousand weather balloons are launched into the air, tracking only 15% of the planet. The balloons rise to an altitude of 20-30 kilometres and last for about two hours in near space, collecting data on temperature, humidity, pressure and wind speed and sending it through radio or satellite back to the station.
Suskho, Dean and the team worked with plastic used in restaurant supplies to make their balloons last in the air for over a hundred hours — a first in weather balloons. They also added capabilities to navigate the balloon in the air up and down, which could control altitude. All this while, the sensors in the balloon collected real-time weather data through a satellite network. This made it possible for the balloons to cover oceans or weather anomalies like storms — another huge weather data gap.
“We could direct a balloon with sensors into a hurricane to measure inside the atmosphere and do it cost-effectively,” says Dean adding that each launch currently costs the company $1000, which is double that of a conventional weather balloon, but the balloons last longer and collect more data offsetting this cost.
From the start, the team wanted to become a full-stack company offering not only weather data but also forecasts and hardware. For a startup on a budget, the biggest challenge was that conventional weather modelling based on tenets of physics was extremely expensive to compute when they started out. When two years ago Huawei showed an AI-based weather model that could not only produce results but beat a physics-based model and do it with less computational power required, the team jumped into weather AI research and development.
The 40-employee startup calls itself “the world’s largest atmospheric sensing system which comes with the most accurate AI weather model”. There’s proof, says Dean. The startup made big news in February 2024, when their deep learning-based weather forecast model, WeatherMesh, unseated models created by Google DeepMind, Nvidia and Huawei as the most accurate medium-range global forecast models. In the increasingly competitive space, it continues to lead in AI weather prediction in key benchmarks set up by government weather models. Armed with a Series A of $15 million led by Khosla Ventures, they want to further work on the model, expand the technical team and launch more balloons.
Currently, the startup is launching three-four weather balloons a day from six different launch sites. They want to eventually build up to 300 launches per day, so they can have 10,000 balloons up in the air all at once to cover the entire earth — that’s one-third of the total weather balloons up in the air at any given time today. In keeping with startups of the Silicon Valley, it’s an ambitious plan to use a cheap home-built product to collect massive weather data.
There is also a lot of plastic in the ocean, choking marine life, and spreading microplastics. Though Windborne balloons are more efficient in collecting data since they last longer in the air and some of them can be navigated back to a launch site, there’s still a long way for them to be sustainable. It’s something that concerns the founders, and though they’ve set offsets by partnering with nonprofits that remove plastic from oceans, they acknowledge it’s not a long-term solution. Meanwhile, commercially, they’re racing ahead. WindBorne has already received contracts from US government agencies like the National Oceanic and Atmospheric Association and the US Air Force. Now they aim to expand into the commercial market.
As the sun sets, I stare at the balloon, a tiny spec in the air and a spec in a constellation of balloons on the startup’s AI system. “It’s a conflation of balloons,” giggles a team member. A geeky joke to end a geeky autumn launch.
Shweta Taneja is an author and journalist based in the Bay Area. Her fortnightly column will reflect on how emerging tech and science is reshaping society in the Silicon Valley and beyond. Find her online with @shwetawrites. The views expressed are personal.