World’s First: U.S. Weather Agency Replaces Supercomputers with Google Cloud Infrastructure

The U.S. National Oceanic and Atmospheric Administration (NOAA) has officially selected Google Cloud as the backbone for its Weather and Climate Operational Supercomputing System (WCOSS), marking one of the world’s first operational weather centers to fully transition to cloud-based infrastructure. This move is expected to deliver faster, more precise forecasts and warnings for extreme weather events.

NOAA historically relied on its own massive supercomputers — like Dogwood and Cactus — to run weather and climate models, enabling high‑resolution forecasts, hurricane tracking, and climate studies before shifting operations to Google Cloud.

Dogwood and Cactus are two twin supercomputers located in Virginia and Arizona, each capable of 12.1 petaflops, tripling NOAA’s previous computing power. Together, the two supercomputers powered the Weather and Climate Operational Supercomputing System (WCOSS), running daily weather, ocean, and climate forecast models.

NOAA's replacement of supercomputers by Google Cloud represents a world-first transition from traditional HPC to cloud-based forecasting.

Key Highlights of the NOAA–Google Cloud Partnership

• Cloud-first transition: NOAA is moving away from on-premises supercomputers to Google Cloud’s high-performance computing (HPC) environment.
• H4D Virtual Machines: Powered by 5th Gen AMD EPYC processors, these VMs will run massive, tightly coupled atmospheric simulations in real time.
• Operational scope: Weather and Climate Operational Supercomputing System (WCOSS) includes daily weather, ocean, and climate forecast models, now supported by elastic, on-demand cloud computing.
• Timeline: NOAA will transition major systems like the Global Forecast System (GFS) and Global Ensemble Forecast System (GEFS) to the cloud by December 2027.
• Scientific impact: Cloud HPC eliminates bottlenecks, enables rapid updates, and boosts resilience during peak demand (e.g., hurricane season).

Supercomputers vs Cloud Transition

Feature	NOAA Supercomputers (Dogwood/Cactus)	Google Cloud HPC
Capacity	Fixed at 12.1 petaflops each	Elastic scaling, on‑demand
Upgrades	Every 3–5 years	Continuous access to latest chips
Forecast detail	High resolution, but limited by hardware	Enhanced by real‑time scaling
Resilience	Downtime risk higher	Lower, resilient cloud
Research integration	Slower adoption	Faster, seamless integration

Benefits for Weather Forecasting

• Earlier and more precise warnings for extreme weather events, improving disaster preparedness.
• Elastic scaling: Computing power can instantly increase during storm seasons.
• Faster research-to-operations pipeline: Scientists can integrate new models without hardware delays.
• Resilience: Reduced downtime compared to traditional supercomputers.

Comparison: On-Prem vs Cloud HPC

Feature	On-Prem Supercomputers	Google Cloud HPC
Upgrade cycle	Every 3–5 years	Continuous access to latest chips
Scalability	Fixed capacity	Elastic, on-demand scaling
Downtime risk	Higher	Lower, resilient cloud
Forecast precision	Limited by hardware	Enhanced by real-time scaling
Research integration	Slower	Faster, seamless

Strategic Context

• NOAA has collaborated with Google since 2011, using Google Workspace and later Google DeepMind’s WeatherNext model, which successfully predicted hurricane landfalls days in advance.
• This partnership positions the U.S. as a global leader in cloud-based numerical weather prediction, strengthening its mission to protect life and property.

Risks & Considerations

• Dependence on a single provider: NOAA’s reliance on Google Cloud raises questions about vendor lock-in.
• Data sovereignty: Sensitive environmental data will reside in commercial cloud infrastructure.
• Cost management: Elastic scaling could lead to unpredictable expenses if not tightly controlled.