# WindBorne Systems

> WindBorne Systems is building a planetary nervous system for Earth's atmosphere. The Redwood City company operates a constellation of autonomous long-duration balloons that gather high-resolution atmospheric data from remote and ocean regions previously invisible to weather models - then feeds that data into WeatherMesh, its in-house AI forecasting model that has surpassed Google DeepMind's GraphCast as the world's most accurate medium-range global weather forecast. With backing from Khosla Ventures, the Gates Foundation, and U.S. military contracts, WindBorne is making a credible run at replacing century-old weather infrastructure.

- **Founded:** 2019
- **Headquarters:** Redwood City, California, United States
- **Founders:** John Dean (CEO), Kai Marshland (CPO (Chief Product Officer)), Andrey Sushko (CTO (Chief Technology Officer)), Joan Creus-Costa (Engineer / Co-founder)
- **Team size:** ~86 employees
- **Products:** Atlas Balloon Constellation, WeatherMesh AI Forecasting Model, Weather Data API & Solutions
- **Notable:** WeatherMesh-1 became world's most accurate AI global weather forecast (February 2024), surpassing Google DeepMind's GraphCast, WeatherMesh-2 released January 2025, achieving 8-24% greater accuracy than GraphCast, Named in TIME Magazine's Best Inventions of 2025 for Atlas balloon constellation

## Products & services

- **Atlas Balloon Constellation** — Long-duration autonomous weather balloons that weigh under 2 kg, fly for 50+ days, and navigate by riding different wind currents at various altitudes. Each balloon delivers ~50 vertical atmospheric profiles over its lifespan vs. one for conventional balloons. The constellation has launched 2,000+ global sounding balloons and targets 10,000 concurrent balloons by 2028 for complete global coverage.
- **WeatherMesh AI Forecasting Model** — An AI global weather forecast model trained on WindBorne's proprietary atmospheric data. WeatherMesh-1 (2024) surpassed Google DeepMind's GraphCast as the most accurate medium-range global forecast. WeatherMesh-2 (January 2025) is 8-24% more accurate than GraphCast. WeatherMesh-3 source code is publicly available on GitHub. Predicts temperature, wind, precipitation, humidity, solar radiation, and more.
- **Weather Data API & Solutions** — Real-time atmospheric data and industry-specific forecast solutions delivered as a service for agriculture, aviation, logistics, insurance, energy/utilities, and government clients.

## Achievements

- WeatherMesh-1 became world's most accurate AI global weather forecast (February 2024), surpassing Google DeepMind's GraphCast
- WeatherMesh-2 released January 2025, achieving 8-24% greater accuracy than GraphCast
- Named in TIME Magazine's Best Inventions of 2025 for Atlas balloon constellation
- Named 2024 Inc. 5000 honoree
- Conducted largest-ever in-situ atmospheric data collection campaign over Arctic Circle
- Delivered first-ever African atmospheric observations via the U.S. National Mesonet program
- Launched 2,000+ global sounding balloons globally including from Uruguay
- Offered free atmospheric data to support National Weather Service following staffing cuts
- Expanded to 50,000 sq ft manufacturing facility in Redwood City (2026)
- WeatherMesh-3 released as open source on GitHub

## Latest updates

- **2026-02** — Expanded manufacturing with new 50,000 sq ft balloon factory in Redwood City, California.
- **2025-01** — WeatherMesh-2 released - achieves 8-24% higher accuracy than GraphCast across key atmospheric variables.
- **2025-01** — Bill & Melinda Gates Foundation invests $5M strategic round.
- **2025** — NOAA begins purchasing Atlas sensor data for official U.S. weather forecasting.
- **2025** — Named to TIME Magazine's Best Inventions of 2025.
- **2024-06** — Series A funding of $15M closed led by Khosla Ventures.
- **2024-02** — WeatherMesh-1 breaks world record for most accurate global weather AI forecast, surpassing GraphCast.

## Links

- Website: https://windbornesystems.com
- LinkedIn: https://www.linkedin.com/company/windborne
- Twitter/X: https://twitter.com/WindBorneWx
- GitHub: https://github.com/windborne
- Instagram: https://www.instagram.com/windbornewx/

---

Profile page: https://yespress.io/windborne-systems
Published by YesPress — https://yespress.io
Last updated: 2026-05-18