Four programs, three billion dollars in combined commitments, and one underlying argument: the infrastructure was never the bottleneck. Here is what each announcement changes for a CIO deciding where the next budget cycle goes.

Most enterprise AI projects never make it to production. That was the number Dave Levy opened with, and it reframes everything AWS announced after it. The infrastructure argument was already settled going into this Summit. What follows is a guide to the four programs aimed at the harder problem: getting agencies from pilot to production, and what each one is worth evaluating for.

Forward Deployed Engineering: Buying Capacity, Not Consulting

The headline investment of the day was AWS Forward Deployed Engineering (FDE), a new global organization backed by a billion dollars that embeds AWS engineers directly inside a customer's environment to co-develop and deploy AI systems. The distinction AWS is drawing matters for procurement: this is explicitly framed as different from traditional consulting, with engagements built around shared outcomes rather than billable hours, and a stated goal of leaving the customer self-sufficient with their own AI engineering capability, workflows, and a knowledge graph of their own architecture and operational patterns, not just a delivered system.

The engineers staffing FDE come from AWS's own frontier AI teams, the same group that builds AWS's internal AI tooling, and the approach runs on what AWS calls an AI-Driven Development Lifecycle, where purpose-built agents handle much of the design, build, and deployment work under human oversight. Commercial customers already running FDE engagements include the Allen Institute, Cox Automotive, the NBA, Ricoh, Southwest Airlines, and the NFL. No public sector engagement has been named yet, but the program was launched specifically at the public sector Summit, and the design constraints AWS built around it, working inside the customer's own data governance and security posture, are the same constraints regulated industries and government agencies already operate under.

For a CIO, the practical question is not whether FDE is available. It is whether your organization has the basic readiness AWS requires to engage this way: an executive sponsor who owns the transformation, not just the technology decision, and systems and data in a state that can support rapid iteration. If those aren't in place, the realistic first step is a professional services engagement to get there, not an FDE engagement itself.

AWS Secret Cloud for Industry: Closing the Access Gap for Contractors

This is the most structurally significant announcement for the defense industrial base specifically. Historically, cleared defense contractors working on classified programs had to build and maintain their own on-premises infrastructure to handle classified workloads, an approach that requires significant upfront capital, cannot scale on demand, and makes it difficult to adopt newer capabilities like generative AI at all.

AWS Secret Cloud for Industry changes that calculus directly. For the first time, defense contractors can run contractor-owned classified workloads on AWS's classified infrastructure, the same regions trusted by the Department of War, inside their own physically and logically isolated environment. The service holds a Provisional Authorization at Impact Level 6 from the Defense Information Systems Agency, meeting the Secret-classification authorization standard, and uses the existing Defense Counterintelligence and Security Agency compliance framework that cleared contractors already operate under, so there is no new security model to adopt. AWS is also committing up to twenty million dollars in credits over three years specifically to help defense companies get started.

Northrop Grumman is the first contractor live on the platform. Without it, the company's initial classified workload would have required months of hardware procurement to stand up on premises; with it, the path to authorized operation compressed to a fraction of that timeline.

For a CIO inside a defense contracting organization, the relevant evaluation isn't whether to adopt cloud, it's whether your current on-premises classified infrastructure spend could instead fund engineering work against a mission, now that the capital-intensive infrastructure buildout is no longer a prerequisite to start.

The IC Accelerated Modernization Framework: Removing the Cost Excuse

For intelligence community agencies specifically, AWS announced the IC Accelerated Modernization Framework (ICAMF), a billion-dollar program designed to remove the single most commonly cited reason workloads stay on premises: the upfront cost of migration itself. The mechanism is straightforward. Agencies migrate qualified workloads to AWS and receive credits tied directly to the successful migration, with up to a billion dollars available through October 2030 across all intelligence community agencies already on the existing AWS contract.

CIA Director John Ratcliffe, in remarks from the keynote stage, confirmed the agency will use ICAMF as part of what he described as a broader technology renaissance inside the CIA: acquisition timelines compressed from a previous range of twelve to twenty-four months down to under six, the agency's cyber function elevated to a full Mission Center, and a stated intent to go further on AI adoption across the organization's IT backbone.

What ICAMF buys an agency CIO is room in the budget. The credits are explicitly designed to offset the upfront hardware costs, ongoing power and facility expenses, and vendor lock-in that have kept workloads on premises, freeing budget that would otherwise be consumed by maintenance for the AI tooling that changes how analysts work.

AWS Transform: What Modernization Looks Like When It Works

The fourth piece is less a single announcement than a working pattern, demonstrated through customer examples rather than a new program name. AWS Transform is the company's modernization tooling for moving legacy code and infrastructure, mainframes, outdated runtimes, monolithic applications, onto modern cloud architecture with AI-assisted migration rather than manual rewrites.

The clearest public sector demonstration came from the United Kingdom. UK Government Chief Technology Officer Sonia Patel told the keynote audience that more than a quarter of UK government IT systems run on infrastructure she described as out of date, and that legacy technology, not AI capability, remains the single biggest brake on modernizing public services. HMRC, the UK's tax authority serving fifty million taxpayers, is investing more than four hundred fifty million pounds to migrate three legacy data centers to AWS, using AI and data to help close a forty-seven billion pound tax gap while improving the experience for citizens filing returns.

Patel also pointed to GOV.UK Chat, an AI assistant inside the UK government's app that answers citizen questions directly instead of routing them through a call center queue. Before release, the assistant tested with more than ten thousand users at roughly sixty-seven percent accuracy. Eighteen months later, in production, it is running at ninety percent. The lesson she drew from both examples was the same: government AI does not have to mean years of planning followed by an indefinite pilot phase. With the right infrastructure and the right economic model behind it, AI improves citizen-facing services on a timeline measured in months, not years.

What Ties the Four Together

Each of these programs is aimed at a different point of friction, talent, classified access, migration cost, legacy modernization, but the underlying diagnosis is identical across all four. AWS is no longer selling infrastructure capacity as the primary value proposition for public sector customers. It is selling the removal of specific, named obstacles that have kept agencies stuck at the pilot stage, and pricing each removal as a distinct investment rather than bundling it into a general cloud pitch.

That is a meaningfully different sales motion than the one that defined the last several years of public sector cloud adoption, and it is worth a CIO's attention less for the dollar figures than for what the figures imply: AWS's own data suggests that the customers who are still stuck are not stuck on infrastructure.