Enterprise

Microsoft bets $2.5 billion on 'Frontier Company' to embed 6,000 engineers inside enterprise customers

The new operating business, unveiled July 2, is Microsoft's answer to Palantir-style forward-deployed engineering — and comes two days after AWS committed $1 billion to the same idea.

By Asher Glick · Enterprise software desk · 2026-07-05 · 3 min read

Photo: Unsplash / Salvatore De Lellis — Microsoft logo on a corporate building exterior

Microsoft on July 2 announced Microsoft Frontier Company, a $2.5 billion operating business that’ll place 6,000 engineers and industry specialists directly inside customer sites to build and run AI systems on the customer’s behalf. It’s the largest single commitment yet to a model the industry has spent the last eight months quietly converging on: sell the software, then send in the humans who make the software work.

The timing isn’t subtle. Amazon committed $1 billion to the same idea on June 30, two days earlier. Anthropic launched a $1.5 billion embedded-engineering venture in May with Goldman Sachs, Blackstone, and Hellman & Friedman as partners; OpenAI launched a standalone OpenAI Deployment Company the same month, backed by more than $4 billion led by TPG. Microsoft’s number is more than twice AWS’s, which reads less like a coincidence than a message.

Rodrigo Kede Lima, most recently a Microsoft sales leader across the Americas and Asia, will run Frontier as its president, reporting into a commercial org led by Judson Althoff, Microsoft’s Commercial Business CEO. Althoff says the effort “goes beyond what has been labeled as Forward-Deployed Engineering” and pitches it as “the largest, most capable, outcome-driven engineering organization in the industry.” Named launch partners include the London Stock Exchange Group, Unilever, Land O’Lakes, and Accenture. Microsoft’s $1 billion, five-year alliance with EY, struck last year, sits alongside it.

The subtext is a Palantir-shaped hole in the hyperscaler business model. Selling model access and Copilot seats assumed enterprises could self-serve their way to production. They largely haven’t. Microsoft 365 Copilot hasn’t achieved the broad adoption the pricing implied, and GitHub Copilot now competes with a stack of newer coding tools that didn’t exist eighteen months ago. Frontier is an admission that the software alone doesn’t close.

What’s being built here isn’t a consulting arm. It’s an infrastructure layer, priced like a product, staffed like a services firm, and modeled after a company Microsoft spent a decade treating as a curiosity. That every major AI vendor arrived at the same conclusion within sixty days is the story.

Sources