AWS Launches $1 Billion Forward-Deployed Engineers Unit to Embed With Customers
Pods of six engineers embed inside customer organizations for 45-day stints - writing production code, not slide decks.
AWS committed $1 billion to a forward-deployed engineering unit on June 30, a program that sends AI engineers to live inside customer organizations for 45-day periods and write production-grade code. Amazon's cloud division describes the AWS forward-deployed engineers as pods of five to six people who embed with a company, navigate internal workflows, and build toward specific AI outcomes. First customers include the NBA and Ricoh.
Each Pod Spends 45 Days On-Site Writing Production Code, Not Recommendations
Forty-five days is short. Long enough, AWS is betting, to ship something real. Over each engagement, the pod accesses internal systems, writes code to production standards, and works within whatever org-chart reality the customer has - rather than diagnosing problems from outside it. For enterprise teams where AI pilots have been stuck for months between proof of concept and live deployment, having engineers inside the building is a qualitatively different kind of help than a cloud migration guide.
AWS plans to grow the unit to thousands of engineers over time. Amazon said it will pull from internal transfers and also hire externally - a signal that the company expects sustained demand rather than a short-term push.
Palantir Invented the Model. AWS Is Now the Biggest Bet on It.
Palantir built forward-deployed engineering into its core business model more than 15 years ago, sending engineers to sit inside defense agencies and banks for months or years at a time. Salesforce, Anthropic, and Google Cloud all run their own versions today, each scoped around their own software stacks. AWS joining at $1 billion and targeting thousands of engineers suggests the model has moved from a premium niche to a standard part of enterprise AI sales. For most enterprise customers, a 45-day AWS pod is also a 45-day sales engagement - engineers building on Bedrock and SageMaker inside a customer's org are the most effective possible advertisement for staying on AWS.
AWS Forward-Deployed Engineers Enter a Crowded but Growing Market
Unlike Palantir's multi-year deployments, AWS forward-deployed engineers operate on defined 45-day windows - a constraint that forces teams to prioritize shipping over planning, which may suit customers better than open-ended engagements. Microsoft's Foundry platform takes a different approach to the same problem: platform tools and hosted agents that enterprise teams deploy themselves, without embedded AWS-style engineers. Both models are bets on the same underlying fact: enterprise AI adoption is slower than vendors expected, and someone has to close the gap.
AWS has not said how many concurrent deployments the unit will run or which industries it targets beyond the NBA and Ricoh. Google Cloud and Salesforce have not publicly responded to the announcement.