In nine days the hyperscaler consensus on how to sell AI has visibly changed. On June 30, Amazon Web Services announced a $1 billion Forward Deployed Engineering organisation seeded with what Francessca Vasquez, AWS vice president of frontier AI engineering and services, called “thousands” of engineers, working in pods of five or six on roughly 45-day cycles inside clients including the Allen Institute, Cox Automotive, the NBA, the NFL, Ricoh and Southwest Airlines. Two days later Microsoft unveiled Microsoft Frontier Company, backed by $2.5 billion and 6,000 industry and engineering experts. Today, OpenAI told Axios that its majority-owned OpenAI Deployment Company, launched in May with $4 billion earmarked for acquisitions, has agreed to buy the applied AI firm Northslope, its second purchase after Tomoro, expanding its bench to hundreds of forward-deployed engineers.

The vendors are cannibalising the systems-integrator layer. Model access is the commodity; the margin is in wiring models into customer operations, work Accenture and Deloitte used to book.

Judson Althoff, Microsoft’s Commercial Business chief executive, described his unit as “the largest, most capable, outcome-driven engineering organization in the industry,” pointedly avoiding the “Forward Deployed Engineer” phrase that Palantir turned into a category and OpenAI adopted. The vocabulary is contested because the positioning is.

Downstream of the hyperscalers, agentic-deployment platforms like LemonLime are picking up the small and mid-size customers the 45-day pods won’t touch, which is most of the market. The Northslope terms weren’t disclosed. They rarely are when the point of the deal is headcount.

Sources