WindBorne’s chief product officer, Kai Marshland, described the performance gains in practical terms, saying WeatherMesh 6 “is as accurate five days out as a traditional forecast is the day before,” particularly for surface temperature predictions.
The company’s approach differs from many AI weather systems because it combines model development with its own data collection network. WindBorne operates roughly 400 weather balloons at any given time from 15 launch sites around the world, continuously gathering atmospheric measurements that feed directly into the WeatherMesh platform.
Most AI forecasting models rely heavily on datasets produced by organizations such as ECMWF and the U.S. National Oceanic and Atmospheric Administration (NOAA). WindBorne says recent improvements stem largely from advances in how its proprietary balloon data is incorporated into its transformer-based forecasting model.
“I don’t understand, personally, the business model of being AI based weather company without a data set advantage,” WindBorne CEO John Dean said. The company believes direct data ingestion has become a key differentiator. Joan Creus-Costa, WindBorne’s head of AI, said data collected from the company’s balloons and other sources played a central role in the improvements introduced in WeatherMesh 6. According to the company, achieving those gains required roughly a year of tuning and redesigning the underlying model architecture while maintaining forecasting stability.
The release arrives as AI continues to gain traction across the weather forecasting industry. Traditional forecasting systems rely on computationally intensive physics-based simulations running on large supercomputers. AI models can generate forecasts much faster, although they have historically faced challenges around resolution, variable coverage, and long-range accuracy.
WindBorne argues that the gap is narrowing. The company says WeatherMesh 6 outperforms both conventional and AI-based ECMWF forecasts across several key weather variables, though broader industry validation will ultimately determine how those results compare over time and across different forecasting conditions.
Beyond forecasting, WindBorne has built a business around its atmospheric data collection network. The company sells balloon data to NOAA as well as the U.S. Air Force and Navy. It also provides forecasting services to investors and commodity traders, although Dean said the company’s primary focus remains expanding its data and model infrastructure rather than building consumer-facing products.
“When we started doing we were still very heavily reliant on ECMWF,” Dean said. “I predict today, if we removed ECMWF’s initial conditions, we would actually still do pretty good.”
The company has also taken steps to address operational challenges associated with its balloon network. After one of its balloons was struck by a United Airlines aircraft last year, resulting in minor damage but no injuries, WindBorne added transponders that broadcast balloon locations through the ADS-B aviation surveillance system to reduce the likelihood of future incidents.
WindBorne has raised $25 million in venture funding and was reportedly valued at $85 million in 2024. With WeatherMesh 6, the company is attempting to demonstrate that combining proprietary atmospheric data with modern AI techniques can compete with established forecasting systems that have long set the standard for weather prediction.
This analysis is based on reporting from BitcoinWorld.
Image courtesy of Windborne Systems.
This article was generated with AI assistance and reviewed for accuracy and quality.
