NOAA Moves Weather Data to Cloud as Agencies Scale AI

The National Oceanic and Atmospheric Administration is moving weather data storage and processing to the cloud to preserve high-resolution data and give forecasters more flexibility in how and where they work.

This reflects a challenge federal agencies face as they expand AI and prepare the data and technology infrastructure to make AI useful in day-to-day operations.

Speaking at the Google Public Sector Summit in Washington Tuesday, NOAA Administrator Neil Jacobs said the agency is changing how it handles massive weather datasets. Other officials from the War Department and the Defense Logistics Agency also discussed efforts to deploy AI at scale and prepare supply-chain data for AI-enabled operations.

NOAA collects data from satellites, ships, aircraft and other sources to support weather, water, climate and space weather forecasting. But the volume of that information can create challenges for moving and storing data.

“Sometimes those files are so big that we have to reduce the resolution to get the file size to a manageable size to move it from point A to point B,” Jacobs said. “When we reduce that resolution, that data is non-recoverable. We’re never getting it back.”

The agency invests in high-resolution instruments to collect detailed observations. It is changing how it stores and processes weather data by using a two-tier cloud architecture, with one environment for high-performance computing and another for storage. This reduces the need to move massive datasets between locations.

The change also builds on NOAA’s adoption of AI-based weather models. Traditionally, the agency has relied on physics-based forecasting models written in Fortran. In recent years, some of that has shifted to AI-based systems that replicate much of the work of traditional models with greater efficiency, Jacobs said.

NOAA is also moving away from workflows that require forecasters to download large files and process them locally. Jacobs said the agency aims to equip forecasters with laptops and satellite connectivity so they can eventually work from virtually anywhere.

“In a couple of years, all we do is equip every forecaster with a laptop with a satellite connection,” he said. “They can literally do their job from anywhere on the planet.”

DLA Prepares Supply-Chain Data for AI

For the Defense Logistics Agency, the challenge is preparing data across its supply-chain operations so it can be used more readily for analytics and AI.

DLA provides the joint force with the supplies and services it needs to operate globally. CIO Adarryl Roberts said the agency is digitizing supply-chain processes as it works toward greater use of agentic AI.

“Agentic AI is the north star,” Roberts said. “It will allow us to have some real continuous enablement of activity.”

Roberts said the agency sees AI as a way to augment its workforce, but achieving that vision requires more than deploying AI tools.

Roberts distinguished between transactional data used to conduct business and data prepared for analytics and AI applications. This means that as it moves across platforms, data will not have to be modified in any way to suit that platform. This will make information ready for use in real time, he said.

“We’re actually making sure the data is AI ready, AI enabled,” Roberts said.

War Department Focuses on Adoption at Scale

The War Department’s challenge is making AI capabilities available and useful across an organization with 3 million users.

Andrew Mapes, principal deputy in the Office of the Secretary of War’s Chief Digital and AI Office, said the department is using platforms such as GenAI.mil to deliver AI tools to personnel and train them to use the technology.

Mapes said the department introduced GenAI.mil in December 2025 as its first AI model deployed across the department, then used the platform to make additional models available to users.

The department is also drawing on capabilities from private-sector partners and coordinating with other government agencies, he said.

But making tools available is only part of the challenge. Personnel need to be able to incorporate them into demanding jobs without adding unnecessary complexity.

“You’re not layering on top of that, so figuring out that balance of training, adoption and integration into the process to leverage all the data [is the challenge],” Mapes said.