AWS lays out new funding and frameworks to move government and classified workloads to cloud-based AI

Amazon Web Services used its AWS Summit Washington, D.C. keynote to argue that the next wave of government AI depends on getting more workloads off aging on-premises systems and onto secure cloud infrastructure. In remarks tied to national security, energy and health, AWS said it is pairing new classified cloud capabilities with funding and engineering support designed to help public-sector agencies and their partners deploy AI faster.

A central theme was classified computing. AWS said defense contractors historically have had to build and maintain separate on-premises infrastructure for cleared programs, a model it described as expensive, rigid, and not suited to newer cloud capabilities such as generative AI. AWS introduced Secret Cloud for Industry (ASCI) as a way to let contractor-owned classified workloads run on the same AWS infrastructure trusted by the Department of War, while staying inside a physically and logically isolated environment purpose-built for stringent security and compliance needs.

AWS framed ASCI as changing the practical economics and deployment timeline for sensitive programs. It said organizations can move to the cloud without adopting an entirely new security model, with AWS handling authorization and private connectivity from existing secure facilities. In the company’s telling, the goal is to keep mission work moving even when engineers need to run AI inference on classified data or train models on sensitive operational information.

Northrop Grumman was identified as the first partner to deploy ASCI. AWS also said it is investing up to $20 million in credits over three years to help customers in the defense sector take advantage of cloud benefits, a form of financial support meant to reduce adoption friction as workloads transition.

The keynote also focused on the U.S. Intelligence Community. AWS said the Intelligence Community has been its longest-standing cloud partner, citing its ability to handle classified data across multiple security levels. But AWS added that many workloads have not migrated yet, and it announced the IC Accelerated Modernization Framework, or ICAMF, described as a $1 billion program intended to eliminate migration costs that have kept some agencies on premises.

ICAMF, AWS said, is tied to migration progress. AWS described a straightforward structure: organizations move qualified workloads to AWS and receive credits. The company said up to $1 billion is available through October 2030 for all Intelligence Community agencies operating under the existing AWS contract, and it characterized the credits as addressing upfront hardware costs, ongoing power and facility expenses, and constraints from rigid vendor lock-in. AWS also said the more workloads agencies migrate, the more they can save, freeing budgets for AI deployments intended to help analysts work faster and surface insights while responding to evolving threats.

For outside government, AWS highlighted a push to make “production AI” delivery more repeatable. It announced AWS Forward Deployed Engineering (FDE), a new global organization that the company said will embed thousands of expert engineers with customers to co-develop and deploy AI solutions. AWS backed the initiative with a $1 billion investment and said the teams partner across business, engineering and security functions to build AI systems that run on top of an organization’s data and governance processes.

AWS described FDE’s approach as using purpose-built agents to compress development and launch timelines “from months into days,” with each deployment helping the next one go faster. At the center of the effort is an AI-Driven Development Lifecycle that combines AI-powered execution with human oversight and dynamic team collaboration. AWS said customers leave with new solutions as well as engineering capabilities, including agentic systems running in the customer’s own AWS environment, and that engagements are designed to leave organizations self-sufficient rather than dependent on ongoing billable consulting.

Beyond the AI delivery mechanics, the keynote included government and research examples intended to show how cloud deployments can support large national programs. AWS said it is working with Idaho National Laboratory on the Department of Energy’s Genesis Mission to harness AI for energy, science and national security, including using AWS technology to compress nuclear reactor design cycles and develop digital twins of a small modular reactor. AWS also said the National Nuclear Security Administration announced a Secret/Restricted Data (S/RD) Enterprise Cloud environment in collaboration with AWS, described as the first cloud environment with enterprise authorization to process S/RD, intended to host Genesis Mission workloads. Separately, AWS said UK Chief Technology Officer Sonia Patel described AWS’s role in deploying AI at national scale across hundreds of government departments, including HMRC’s plan to migrate three legacy data centers and invest over £450 million, with an emphasis on using AI and data to improve taxpayer services and address a reported tax gap.

What remains unclear from the company’s remarks is the precise scope and selection criteria for what counts as a “qualified” workload under ICAMF, as well as timelines for ASCI deployments and the size or cadence of credits for particular contractors or agencies. AWS also did not provide detailed performance metrics comparing AI deployment outcomes before and after migrations, nor did it specify how many workloads have been moved so far across either the Intelligence Community or classified defense programs.

For AWS and public-sector customers, the immediate watch items are whether ASCI adoption expands beyond the initial partner, whether ICAMF credits translate into measurable migration milestones through October 2030, and how quickly FDE teams can move from pilots to stable production deployments. AWS’s pitch is that AI for sensitive missions will be constrained as long as organizations remain tied to on-premises infrastructure, and that its combination of funding, secure cloud architecture and embedded engineering support can reduce that bottleneck.