Copilot is becoming Microsoft’s test of how far Big Tech can move beyond borrowed AI
Microsoft folds Copilot under Snap exec to build AI autonomy📷 Scraped: Mar 17, 2026
- ★Jacob Andreou takes over the experience division as EVP, while Mustafa Suleyman leads model development including superintelligence
- ★Four functional pillars — Copilot experience, platform, Microsoft 365 apps, and AI models — are now synchronized under one roof
- ★The move follows Nadella's Build 2024 emphasis on 'in-house scale' for AI and signals strategic decoupling from OpenAI's GPT-4 models
Satya Nadella has quietly consolidated Copilot's scattered leadership under Jacob Andreou, a former Snap Inc. executive who built the social platform's core growth and camera products. The move, confirmed by Windows Central, replaces the previous fragmented structure with a single executive running the experience division as Executive Vice President. Andreou's appointment signals something sharper than a routine reorganization: Microsoft wants Copilot to feel less like a bolt-on service and more like native infrastructure across Windows, Office, and Edge.
The restructuring synchronizes four functional pillars—Copilot experience, platform, Microsoft 365 apps, and AI models—under one roof for the first time. Mustafa Suleyman, who joined from Inflection AI in 2024, retains control of model development including the company's superintelligence efforts. The division of labor is deliberate: Suleyman builds the engines, Andreou designs the cockpit. This mirrors how Nadella has historically operated, pairing technical depth with product intuition.
The timing is hardly accidental. At Build 2024, Nadella emphasized "in-house scale" for artificial intelligence—a phrase that landed differently once analysts parsed the organizational charts. Microsoft's Azure AI services had relied heavily on OpenAI's GPT-4 for Copilot's core interactions through 2023. The new structure accelerates a decoupling that was already underway in engineering backchannels.
Nadella consolidates four pillars under one roof with a clear mandate: less OpenAI, more Microsoft
Signals long-term independence from third-party AI models📷 Scraped: Mar 17, 2026
Industry observers note Andreou's consumer-facing background could prove decisive. At Snap, he oversaw products used by hundreds of millions daily; that sensibility for frictionless interaction is exactly what Copilot has lacked. Early iterations of the assistant required users to context-switch deliberately—opening a sidebar, formulating prompts, interpreting responses. The Snap veteran's mandate appears to be embedding AI so seamlessly that the interface itself disappears.
The shake-up arrives as Microsoft's OpenAI partnership faces mounting friction. Data privacy concerns have dogged enterprise deployments, and model access limitations have frustrated developers seeking deeper customization. Analysts at SemiAnalysis reported in March that Microsoft's internal AI teams were already prototyping rival architectures, driven by licensing costs and competitive differentiation pressures. The reorganization formalizes what was emerging informally: a strategic hedge against dependency.
The hardware layer reinforces this trajectory. Microsoft's Maia AI accelerator, unveiled alongside its Cobalt CPU family, offers a path to run inference on custom silicon rather than licensed cloud capacity. If Copilot eventually ships primarily on Maia rather than OpenAI's infrastructure, the economic and strategic implications reshape Microsoft's entire AI calculus.
The broader pattern is unmistakable. Google and Meta have poured billions into proprietary model stacks; Amazon has accelerated Trainium and Inferentia development. Microsoft's move signals the final phase of a transition from partnership-dependent experimentation to vertically integrated competition. The quiet arms race within Big Tech is no longer about who can access the best models—it's about who controls the complete stack from silicon to interface.
