Microsoft hires top AI researchers from Google DeepMind to boost Copilot and its new London AI lab

A new wave of talent aims to accelerate Microsoft’s proprietary models for Copilot, Bing, and Windows.

Microsoft has confirmed the hiring of at least 24 high-profile engineers and researchers from Google DeepMind, marking the largest single group transition between the two AI giants to date.

This strategic move signals Microsoft’s intent to rapidly expand its in-house AI capabilities for Copilot, Bing, and Windows, reducing its dependence on OpenAI and setting the stage for a new era of proprietary, consumer-first AI research.

Microsoft’s AI ambitions sharpen with the creation of a London reseach hub

The new hires will be based at Microsoft’s dedicated AI lab in London, working directly under Mustafa Suleyman, the former DeepMind co-founder brought in to lead Microsoft AI in 2024. This lab’s mission is clear: build foundation models and generative systems for direct integration with Copilot, the Windows OS, Surface devices, and Azure AI Services. By establishing a major European presence, Microsoft positions itself to tap into a world-class research ecosystem and compete head-to-head with DeepMind on its home turf.

The London team will focus on the next wave of multimodal language models, advanced agent frameworks, and fast, privacy-safe AI services for consumers and enterprise clients alike.

The new team includes Gemini project leaders and RLHF specialists

The most prominent name in the group is Amar Subramanya, who spent sixteen years at Google and most recently led engineering for Gemini—Google’s flagship foundation model. Subramanya joins Microsoft as Corporate VP of AI. Alongside him are Adam Sadovsky, former Senior Director at DeepMind, as well as Sonal Gupta and Jonas Rothfuss, renowned for their work in reinforcement learning from human feedback (RLHF) and advanced reasoning.

Reports suggest these researchers are joining on highly competitive packages—including equity grants exceeding two million dollars—reflecting the extraordinary value placed on top AI talent in the ongoing “talent war” among tech giants.

Why Microsoft is investing in homegrown models alongside OpenAI’s partnership

While Microsoft remains a core investor and customer of OpenAI, the company’s strategy is clear: hedge reliance on external partners by developing its own proprietary models, optimized for integration within Microsoft products and tailored for the privacy, latency, and cost needs of its global user base.

The new lab is tasked with rapid experimentation—delivering native Copilot agents for calendars, email, spreadsheets, and specialized verticals like healthcare, finance, and gaming. By building “sovereign compute” capabilities and training models on Microsoft-owned infrastructure, the company hopes to future-proof Copilot and offer differentiated experiences in competition with both OpenAI and Google Gemini.

Mustafa Suleyman’s vision: building the leading European AI powerhouse

Since joining Microsoft in early 2024, Suleyman has doubled down on European AI recruitment, targeting a doubling of headcount on the continent by 2026. The new London hires are at the center of this plan, tasked with advancing efficiency, long-context reasoning, and jailbreak resistance in next-generation AI models.

The group will collaborate closely with global teams in Redmond and San Francisco but operate with autonomy, leveraging the depth of European academic and engineering talent to push Copilot’s capabilities far beyond current boundaries.

What’s next for Copilot, Bing, and Microsoft’s AI ecosystem

If onboarding proceeds smoothly, the first new models developed by the ex-DeepMind team could be previewed as soon as the next Microsoft Ignite event. These advances will power Copilot’s expansion across consumer and enterprise workflows, strengthen Bing’s search capabilities, and help ensure that Microsoft remains at the forefront of the global AI race.

_______

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org