Google just rolled out WeatherNext 3, an AI weather model that ingests live satellite data every hour — meaning a company that already tracks your searches, location, and email now has a continuously updating atmospheric feed it controls end to end.
The model may well produce better forecasts. Google DeepMind claims up to 50% more accurate precipitation predictions and resolution five times sharper than its predecessor, dropping from a 25-kilometer grid down to 5 kilometers for temperature and moisture. That's a genuine technical leap. But every new data stream Google swallows raises the question the tech press won't ask: who owns the data, and what consent did Americans give?
All three outlets covering the rollout — Engadget, Gizmodo, and 9to5Google — frame WeatherNext 3 as an unambiguous good. Gizmodo leads with how "hyper-local weather is still tricky" and calls the model Google's "most advanced." Engadget highlights the renewable energy angle. 9to5Google runs the announcement essentially as a product brief. None of them mention surveillance, data ownership, or the fact that a monopoly with a documented history of hoovering up user information just added another real-time global data ingestion pipeline to its stack.
Here is what the model actually does. Previous AI weather systems, including Google's own WeatherNext 2 from November 2025, relied on numerical weather prediction models — physics simulations run on supercomputers that carry a six-hour lag. WeatherNext 3 bypasses that bottleneck by feeding "a mosaic of live, global geostationary satellite data" directly into the AI, according to Google. It also trains on sparse weather station observations and combines NASA satellite precipitation data with what Google calls its "own global precipitation reanalysis based on satellite radar."
The result: a new global forecast every hour, at resolutions detailed enough to predict 100-meter wind speeds — roughly turbine height — for wind farms, along with cloud cover and solar radiation for solar operators. Google says the model is already powering Search, Maps, Gemini, the Maps Platform Weather API, and Google Earth Engine. It also sits atop the Operational WeatherBench leaderboard, an independent ranking run by AI weather startup Brightband.
Google leans hard on the developing-world angle. The company says WeatherNext 3 "brings localized, high-fidelity forecasting to billions of people" in Latin America, Africa, and Asia-Pacific, where traditional supercomputer models have been too expensive. Fair enough. But Google's philanthropic framing papers over the commercial reality: the same infrastructure that serves a farmer in Kenya also feeds Google's ad-targeting ecosystem, its cloud services, and its mapping monopoly back home.
And then there is the tell. Despite all the hype, Google still directs users to their local meteorological agency or national weather service for official forecasts, as Gizmodo noted — buried at the end of its story. If the model is that good, why the disclaimer? Because Google is building the infrastructure of information dependency, not replacing the institutions. You come for the forecast. You stay because every Google product now runs on the same data backbone.
The question isn't whether WeatherNext 3 works. It's whether Americans should be comfortable with one corporation sitting on live satellite feeds, weather station data, search histories, location pings, and email content — all at once, with no meaningful consent mechanism and no clarity on who owns what.
