Amazon Web Services launched a forward-deployed engineer organization on Tuesday and committed $1 billion in internal resources to staff it. The new group will embed AWS engineers inside customer companies to build purpose-built AI agents, run them inside the customer's own AWS environment, and hand off the workflows when the engagement ends. Francessca Vasquez, the AWS VP of Frontier AI who announced the org, framed the work around customer self-sufficiency rather than long-term managed services.
The move follows two comparable plays from frontier labs. OpenAI stood up its own FDE joint venture valued at $4 billion, and Anthropic followed with a $1.5 billion vehicle. Both labs paired with private equity partners that supplied capital and, more importantly, a portfolio of client corporations ready to deploy. Amazon's $1 billion is structured differently — it's internal Amazon resources, not a JV — but the competitive logic is the same.
The forward-deployed engineer model traces to Palantir, which spent two decades sending engineers into client sites to wire its software into customer workflows. In a typical FDE engagement, a contractor's engineer sits inside the client for the duration of a deployment, responding directly to internal opportunities and obstacles as they surface. The underlying technology gets reused across deployments while the integration layer is tailored to each customer's stack.
“Customers leave AWS FDE deployments with both new solutions and new engineering capabilities.”— Francessca Vasquez, AWS VP of Frontier AI
Key facts
- 01AWS committed $1 billion in internal resources to a new forward-deployed engineer org announced Tuesday.
- 02OpenAI's FDE joint venture is valued at $4 billion; Anthropic's sits at $1.5 billion.
- 03AWS engineers will embed inside customer companies to build and hand off agentic systems running in the client's own AWS environment.
- 04Both OpenAI and Anthropic structured their FDE ventures with private equity partners to supply capital and client pipelines.
- 05Francessca Vasquez, AWS VP of Frontier AI, announced the new organization.
The model has become the default answer to a problem that has frustrated AI buyers for two years: models that demo well in isolation but stall once they meet a real enterprise environment. Off-the-shelf APIs ship the capability but not the deployment. FDE engagements ship the deployment, and the contracting party absorbs most of the operational risk.
The downside is labor. Running an FDE practice means maintaining a permanent corps of senior engineers who can install, tune, and maintain systems on-site, then rotate to the next client. The economics only work if the underlying technology is reusable enough that each engagement gets faster and cheaper than the last. That is the bet OpenAI, Anthropic, and now AWS are all making.
AWS's structural advantage in this race is that its FDE engagements end with the customer running agentic systems inside their own AWS environment. The cloud spend is the long-tail revenue. OpenAI and Anthropic, by contrast, monetize the model inference itself, which means the FDE engagement has to drive enough token volume to justify the deployment cost. Amazon doesn't have that constraint — every successful FDE rollout grows AWS consumption.
“Along with agentic systems running in their own AWS environment, they gain lasting AI skills, workflows, and patterns they can use to innovate independently.”— Francessca Vasquez, AWS VP of Frontier AI
There's also the question of what the customer keeps. Vasquez emphasized that AWS engagements leave behind both solutions and engineering capabilities, a deliberate contrast to the consulting-firm pattern where the contractor's departure leaves the client dependent on follow-on work. Whether AWS can hold that line at scale, when its FDE org is staffed by hundreds or thousands of engineers, is a separate question.
The Palantir comparison is instructive on the limits, too. Palantir built one of the most durable FDE practices in software, but it took roughly two decades and a heavy gross-margin profile to make the unit economics work. OpenAI and Anthropic are attempting to compress that timeline by piggybacking on private equity portfolios. AWS is attempting it through scale and its existing enterprise sales motion. Neither shortcut has been proven yet.
Skeptics will note that $1 billion of internal resources is a soft commitment compared to a $4 billion joint-venture valuation with outside capital on the line. Internal allocations can be quietly trimmed; JV equity cannot. The size of the AWS FDE org six quarters from now will say more than the launch post does. There's also a recruiting question: senior engineers willing to embed on-site for months at a time are scarce, and three of the largest AI buyers in the world are now hiring for the same role.
What's clear is that the FDE model has moved from a Palantir oddity to the dominant pattern for AI enterprise deployment in under a year. The frontier labs have decided that selling tokens is not enough, and the largest cloud provider has decided that selling compute is not enough either. Both groups are converging on the same answer: put engineers in the building. For enterprise buyers who have been waiting for someone else to do the integration work, the supply of that labor just expanded considerably — and the pricing pressure on traditional systems integrators is about to follow.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
