Predicting the weather at a national scale is one of the most computationally demanding tasks in science — and NOAA just made a major infrastructure decision to keep up with it. The agency has selected Google Cloud as the primary provider of high-performance computing for its Weather and Climate Operational Supercomputing System (WCOSS), moving core forecasting operations off on-premise supercomputers and onto the public cloud for the first time.
What’s Actually Changing
NOAA will migrate operational numerical weather prediction — including systems like the Global Forecast System and Global Ensemble Forecast System — onto Google Cloud’s H4D virtual machines. That makes NOAA one of the first major global weather centers to run live operational forecasting directly on public cloud infrastructure rather than dedicated on-site supercomputers.
The shift is meant to solve a problem that’s been building for years: traditional on-premise HPC systems are powerful but rigid. Upgrading them means large, infrequent hardware refreshes rather than continuous improvement. Cloud infrastructure instead gives NOAA elastic, on-demand compute — meaning capacity can scale up for a major simulation and back down afterward, rather than sitting fixed at whatever the last hardware refresh allowed.
Why This Isn’t a New Relationship
This deal builds on more than a decade of ties between the two organizations. NOAA has used Google Workspace since 2011, and more recently has leaned on Google DeepMind’s WeatherNext model — an AI-driven forecasting system that reportedly helped predict Hurricane Melissa’s Category 5 landfall in Jamaica five days ahead of time. NOAA Fisheries also ran an earlier cloud compute accelerator pilot that gave researchers more flexible access to processing power, laying groundwork for this larger migration.
That history matters because it suggests this wasn’t a single, isolated procurement decision — it’s the product of years of smaller collaborations that gradually built trust in Google’s infrastructure and AI tooling within the agency.
The Stakes
- Faster model iteration: elastic cloud compute lets researchers test and refine forecasting models without waiting on fixed hardware cycles.
- Earlier, more accurate warnings: more compute power supports higher-resolution simulations, which can translate into earlier evacuation and public-safety alerts for severe weather.
- AI-augmented forecasting: the migration pairs traditional numerical weather prediction with machine-learning models like WeatherNext, blending both approaches rather than replacing one with the other.
Timeline
NOAA plans to begin migrating workloads to the new cloud environment in early 2027, with the transition expected to be complete by December 2027. Google Cloud will serve as the primary HPC provider for WCOSS throughout that process, supporting the National Weather Service’s operational forecasting and public decision-support systems.
The Bigger Picture
This migration is part of a broader trend of public-sector agencies moving mission-critical, real-time systems onto commercial cloud infrastructure rather than maintaining them in-house — a shift that trades direct hardware control for speed, scale, and access to newer AI capabilities as they become available. For an agency whose core job is protecting lives and property from extreme weather, the ability to scale computing power on demand, rather than waiting years for the next supercomputer refresh, could meaningfully shorten the time between a storm forming and the public being warned about it.
Read the full announcement on the Google Cloud blog.
