Exclusive: Google Cloud unveils AI-powered weather predictions
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Shoshana Gordon/Axios
Google's Cloud division is taking a major step toward making operational recent gains in AI weather forecast models and marketing them for the energy industry, the company tells Axios exclusively.
Why it matters: This is a prominent example of a tech company that invested in developing AI models to make the transition from research to applications.
- AI weather models are in their infancy but have demonstrated remarkable accuracy. Those advances have come as certain extreme weather events are becoming more intense and frequent due to human-caused global warming.
Driving the news: Google Cloud is marketing two AI forecast models to its enterprise cloud customers.
- Both were developed by Google DeepMind, and used historical weather data to make predictions about future conditions out to 10 to 15 days in advance.
- One model, previously known as GenCast, bested some of the world's most accurate modeling systems.
- It generates probabilistic projections to allow companies to plan for high impact, low probability scenarios as well as the most likely forecast outcomes.
The big picture: The tech industry has largely led the charge on AI modeling given its expertise working with large datasets and access to significant computer resources.
- Google, Microsoft and Nvidia have each pursued the development of AI weather models despite none of them being a strictly weather and climate company.
- However, Google is now out in front when it comes to bringing its models to market.
The intrigue: Google Cloud is bringing two models, branded as "WeatherNext," to its Cloud enterprise customers to try to help them plan for extreme weather.
- The energy industry is a key customer given companies' needs to plan for changing weather conditions, Pete Battaglia, director of research for sustainability at Google DeepMind, told Axios in an interview.
- Energy companies, Google hopes, will find the new tools useful for everything from planning for supply and demand swings to anticipating the need to tap into battery storage resources.
- Google also hopes it can lead them to make decisions on where to build renewable energy infrastructure.
Google's Cloud division also sees future demand for its new weather models coming from the logistics and retail sectors, as companies seek to optimize shipping routes and stores try to stock their shelves with weather-appropriate gear.
Zoom out: Google made its announcement in the run-up to the annual CERAWeek energy conference in Houston, which features top oil and gas CEOs and representatives of the renewables sector as well as utilities.
- The announcement also comes as NOAA, the nation's top weather and climate agency goes through rounds of cuts and an uncertain future.
- Most private sector weather providers obtain original weather data for free from NOAA and other global centers, then use it to feed into their proprietary weather models.
- AI weather models work differently, since they are trained on historical weather data and don't involve computationally-intensive physics equations, enabling them to be run far faster and cheaper than traditional models.
NOAA's approach to AI weather modeling is still developing, and Battaglia said he is open to collaboration opportunities between the agency and GoogleDeepMind.
The bottom line: AI weather models are going mainstream, tailored to specific use cases. For now, they are supplementing, rather than replacing, traditional physics-based models.
