FLYING ABOVE THE STORM

California start-up WindBorne Systems offers help to the National Weather Service (NWS) after DOGE layoffs

“Hearing about the suspension of the Kotzbue radiosonde station in Alaska and the concerns around the future of weather forecasting reinforced our commitment to support NWS and help offset imminent observation losses. Part of our broader strategy is to offer free observations where possible.”

-- John Dean, Founder & CEO WindBorne Systems

Founded in 2019, WindBorne is a planetary intelligence company with a mission to build a planetary nervous system. WindBorne’s approach is two-fold: 1) to collect critical atmospheric data via Atlas, WindBorne’s constellation of long-duration Global Sounding Balloons and 2) to fuse this data with WeatherMesh, an AI weather forecasting model, to create the most powerful and efficient global weather forecasts in the world.

Interview with John Dean, Founder & CEO WindBorne

By Suzanne Forcese

WT: Please introduce yourself to our viewers highlighting your background and the journey to founding WindBorne.

 Dean: I dropped out of Stanford’s electrical engineering master’s program to co-found WindBorne in 2019. At WindBorne, I’ve leveraged my background in machine learning and convex optimization to build AI weather forecasts and autonomous balloon flight software. While earning my BS in electrical engineering at Stanford, I served as co-president of the Stanford Space Initiative, the student group that gave rise to WindBorne. Before WindBorne, I worked on Falcon 9 main engine controls at SpaceX and autonomous systems at Lyft.

WindBorne was a project that I started with a group of peers at the Stanford Space Initiative. We were tasked with extending the flight duration of conventional weather balloons, which can typically only fly for about 2 hours to gather a single vertical profile above land before they pop and return to earth.

Our core team is composed of former SpaceX engineers who have been developing our technology since it started as a Stanford research project in 2015. Between us, we've worked at giants like IBM and NASA and successful startups like LeoLabs and Lyft.

During our research at Stanford we realized that the limited flight capabilities of conventional weather balloons means that roughly 85% of the globe lacks reliable atmospheric data, as conventional balloons can only fly above land, and not oceans without difficulty and excessive costs. This inspired us to create Atlas, our constellation of long-duration Global Sounding Balloons (GSBs).

These lightweight, AI-powered GSBs can self-navigate and intelligently surf wind currents, allowing them to collect critical in-situ atmospheric data above some of earth’s most remote locations. The goal is to close the atmospheric data gap, which currently limits traditional weather forecasting in terms of speed and accuracy.

WT: What is the importance of better weather forecasting and how does it accelerate the adoption of renewables?

Dean: Weather forecasting affects everything from disaster preparedness and response, to business, travel, and military operations.

When it comes to renewables, uncertainty and variability are key obstacles to their widespread adoption. Better forecasts can help us better prepare for extreme weather events, which significantly strains the grid. 

One specific application is that better forecasts can lead to reducing the usage of carbon-intensive peaker plants by helping grid operators optimize energy dispatch.

WT: Following the announcement on February 28, 2025, from the National Weather Service in the wake of recent DOGE activity, Windborne has made a very generous offer.

Why you are confident that Windborne is poised to make a positive impact?

Dean: Hearing about the suspension of the Kotzbue radiosonde station in Alaska and the concerns around the future of weather forecasting reinforced our commitment to support NWS and help offset imminent observation losses. Part of our broader strategy is to offer free observations where possible.

It’s disappointing to see what is going on, but I also am optimistic about how new tech like WindBorne’s balloons can both save money and improve weather observations.

The timing of this offer was apt, as we just happened to be almost done setting up a new balloon launch site in Fairbanks, Alaska.

WT:Describe the Windborne system. How is the system superior to other traditional weather tracking methods?

Dean: We operate as a full-stack weather intelligence company, owning data collection, proprietary AI-based forecast modeling, and forecast productization. This vertical integration enables each part of the stack to be better optimized. Our AI model can be tuned to take optimal advantage of the data, and vice versa. The result is higher quality data collection and forecasting accuracy than we could achieve if each technology operated independently.

More data equates to stronger forecasts. Because our GSBs are able to collect 10x more data per dollar over land and 150x more data per dollar over ocean when compared to conventional weather balloons, we are able to surpass traditional models in terms of novel data collection.

Our global, medium-range AI-based weather forecasting model, WeatherMesh -launched in February, 2024 which broke the world record for the most accurate global weather forecasts -- is then able to use this data to generate hi-resolution forecasts in a fraction of the time of traditional models. WeatherMesh can compute a two-week forecast at 0.25 degree in just 12 seconds on a single RTX 4090 - a rate that is roughly 100,000 times faster than traditional physics-based weather models.

As a result, we have consistently beat out both physics-based models (GFS, ECMWF) and AI-based models (Google GraphCast, Huawei) to become the most accurate forecasting model in the world.

WT: What is Windborne Trader?

Dean: WindBorne Trader is a platform that allows energy traders to use WindBorne’s record-breaking weather forecasts to make better decisions faster. Our forecasts update every 10 minutes, offering real-time intelligence from Atlas and WeatherMesh. WindBorne Trader’s visual interface - which incorporates everything from macro-level continental patterns to specific wind turbine sites - allows traders to analyze data geographically, supporting more intuitive strategy development. This means traders can drill down to local pockets of weather risk without losing the global context.

WT: Your website refers to WeatherMesh-2. Is this a new iteration of WeatherMesh?

Dean: Yes, just launched in January 2025 WeatherMesh-2 (WM-2) - which brokenew forecasting records - is the latest iteration of our global, medium-range AI-based weather forecasting model. It predicts core forecasting metrics across the Earth’s surface and atmosphere, including geopotential, wind speed and direction, temperature, precipitation, cloud cover, solar radiation, pressure, and humidity. It predicts these metrics about 8% to 24% more accurately than the GFS, ECMWF’s HRES, and Google DeepMind’s GraphCast over equivalent time horizons.

Other achievements and capabilities include:

• WM-2 forecasts 0.25-degree global grid forecasts for 14+ days at both hourly and six-hourly intervals.
• Offers high-resolution local forecasts anywhere in the world for plots as small as 1 km by 1 km.
• Produces complete 10-day forecasts in nine seconds and complete 14-day forecasts in 13 seconds, about 143,000 times faster than the gold standard global weather model, the ECMWF’s HRES

WT: Internationally what are your breakthroughs to date?

Dean: We have balloon sites in South Korea and Southeast Asia

In September, we partnered with The Gates Foundation to set up a balloon launch site in Kenya. The goal is to collect data in Sub-Saharan Africa - a region that critically lacks reliable atmospheric data - to improve agricultural processes and efficient energy and water usage on the continent. Further, most hurricanes that hit North America begin as disturbances over Africa, and increased data collection can help people in North American countries better prepare for these storms.

WT: What stage are you currently at with your start-up? What's next?

Dean: We are working on a version of WeatherMesh that can do full AI-based data assimilation and is not reliant on initial conditions from conventional weather models. We are continuously improving WeatherMesh with a new significant upgrade every few months.

We’re also working on doubling the number of balloons we have aloft every 3-6 months. Currently we have around 100 aloft at any point in time.

One of the biggest threats to humanity and our environment is weather uncertainty. This is why it is crucial to close the atmospheric data gap that currently exists for 85% of the globe, so that we can adapt as we move into a future with more extreme weather.