Microsoft's New Bet: "We Won't Feed Your Data to the Model"

The numbers behind Microsoft's new business unit, Frontier Company, are large: a $2.5 billion budget, 6,000 engineers embedded inside customer companies, and a claim to be "the largest engineering organization in the industry." But what stands out isn't the numbers — it's the line Microsoft draws. It puts in writing that it won't use a customer's data and IP to train models in ways that erase what sets that company apart.

In brief

• What launched: Frontier Company, a new Microsoft business unit that scales up its "Forward Deployed Engineering" practice. Led by Rodrigo Kede Lima, a 30-year Microsoft veteran. No new product or model — this is an org and vision announcement.
• The numbers: $2.5 billion invested, 6,000 industry and engineering experts embedded at customer sites, working alongside existing FDE partners like Accenture, Capgemini, EY, KPMG, and PwC.
• The framework — Intelligence + Trust: Intelligence is a company's own data, expertise, and decision-making; Trust is observing, governing, managing, and securing that AI at every layer, on Azure and Microsoft 365.
• Model choice is open: customers aren't locked into one model — they pick per use case across OpenAI, Anthropic, Microsoft, and open source. Early customers include LSEG (embedded in LSEG Workspace for finance professionals), Unilever, Novo Nordisk, and Land O'Lakes.

Our take

What's being sold here isn't a product — it's a guarantee. Microsoft has clearly decided that "we'll set up AI for big companies" is no longer enough; the real question is who ends up owning that company's data along the way. This $2.5B move doesn't touch a small business's day-to-day directly, but the question underneath applies to everyone: does the AI tool you use keep the knowledge that sets you apart in your hands, or quietly pour it into a shared pool? Pick a tool without asking that, and sooner or later you'll find the answer sitting in someone else's hands.

Kaynak: Microsoft