Google is expanding its AI empire — and losing the people who built it

Google is currently navigating a period of significant transition within its artificial intelligence division. The company remains a dominant player in the industry with substantial revenue growth in its cloud sector, yet it faces internal friction as prominent researchers depart for new ventures. Longtime chief scientist Jeff Dean is leaving after 27 years to launch a public benefit corporation, and Demis Hassabis is transitioning from his role as CEO of Google DeepMind to become the unit's chairman.

These personnel shifts occur alongside a strategic pivot in how the organization allocates its resources. While Google continues to invest heavily in proprietary AI chips known as Tensor Processing Units, these assets are in high demand across both internal research teams and external cloud clients. Some researchers have expressed frustration regarding access to this computing capacity, which is essential for developing frontier models. This tension highlights the ongoing conflict between prioritizing pure research and meeting the needs of profitable commercial products.

Management is working to bring research and commercial operations closer together to maintain competitive footing against rivals like OpenAI and Anthropic. CEO Sundar Pichai has emphasized that securing compute for future development remains the company's first priority, even as it balances that goal with the demands of consumer products. As leadership roles shift, Koray Kavukcuoglu is taking on increased responsibility for the management of the DeepMind division and the development of the Gemini model.

Market analysts suggest that the broader industry is entering a phase where many enterprise tasks do not require the most powerful models available. This shift in market demand may influence Google's future investment strategies. While the departure of key technical talent draws attention, the company is still positioned as a full-stack player in the AI market, providing everything from specialized hardware to advanced software models for a wide range of global customers.