Google unveils a sharper, hourly AI weather model

Google announced an updated AI‑driven weather model that promises sharper forecasts and faster updates.

The new system, called WeatherNext 3, builds on Google’s earlier WeatherNext 2 model but delivers predictions at a resolution five times finer.

Whereas WeatherNext 2 generated forecasts every six hours on a 25‑kilometre grid, WeatherNext 3 produces hourly forecasts on a 5‑kilometre grid.

The higher spatial and temporal granularity comes from training the model on live satellite observations rather than relying solely on historical data.

“One of the main developments is for [WeatherNext 3] to go beyond what data most global AI models train on,” says Samier Merchant, a research engineer at Google Research.

“We’re able to leverage fresher and richer observational data sets.”

Traditional weather forecasting still depends on supercomputers that solve physical equations, a process that introduces a time lag between observation and prediction.

AI models such as WeatherNext 3 bypass part of that lag by recognizing patterns in the combined historical and real‑time inputs.

The hourly update cycle is especially useful for tracking fast‑moving systems that can bring sudden rain or snow.

Google claims that the new model can improve precipitation forecasts by up to 50 percent when looking at least a day ahead.

The benefit is most pronounced in regions with sparse ground‑based rain gauges, many of which lie outside the United States and Europe.

By filling observational gaps with satellite‑derived moisture and temperature data, WeatherNext 3 can generate more reliable forecasts for those underserved areas.

How WeatherNext 3 Improves Resolution and Timeliness

The model visualizes temperature and moisture fields at 5‑kilometre resolution, a scale comparable to many local weather services.

This resolution allows Google to display temperature variations within a single metropolitan area that previously would have been averaged together.

Hourly forecasts mean that a user checking the service at 2 p.m. receives a prediction that incorporates satellite data captured minutes earlier.

Such immediacy can help users plan activities, agricultural tasks, or emergency responses with fresher information.

Implications for Renewable Energy and Public Services

Google also tuned WeatherNext 3 to predict wind speeds at 100 metres, the typical hub height of modern turbines.

Accurate wind forecasts at turbine height support grid operators and renewable developers in balancing supply and demand.

“As the energy needs of Google, but [also] entire humanity, is increasing its energy needs, to make sure that we make renewable a very appealing opportunity is very important for us,” Ferran Alet, a research scientist at Google DeepMind tells The Verge.

The model’s outputs are already embedded in Google Search, Maps, and the Gemini AI assistant, extending the reach of the forecasts to billions of users.

Google has partnered with the U.S. National Hurricane Center and several Asian meteorological agencies to test the model’s performance on severe weather events.

Despite the improvements, Google notes that WeatherNext 3 continues to be trained on data from physics‑based models, and agencies typically consult multiple forecasts before issuing warnings.

This collaborative stance underscores that AI forecasts are intended to complement, not replace, traditional numerical weather prediction.

Early internal evaluations suggest that the model’s finer grid reduces false alarms for localized thunderstorms while catching more genuine rain events.

For consumers, the integration means that a simple voice query to Google Assistant can now return a hyper‑local rain probability for the next hour.

For developers, the API access to WeatherNext 3 data opens possibilities for building sector‑specific applications, such as precision farming dashboards.

The rollout arrives as data‑center operators, including Google, grapple with rising electricity consumption from AI workloads.

By improving renewable energy forecasts, WeatherNext 3 could help offset some of that demand through better grid planning.

Analysts observe that the model’s reliance on satellite data may inspire other tech firms to invest in similar observation pipelines.

As more companies adopt AI‑enhanced weather services, the overall ecosystem of forecasting could become more resilient to data scarcity.

Why This Matters: Google’s WeatherNext 3