DeepMind loses its CEO and independence as Google centralizes AI in the US

The era of DeepMind as an independent research lab inside Google is over. A Guardian report cites a former Google manager declaring as much, while a current DeepMind employee now describes the legendary AI lab as “just another subdivision” of the parent company. The structural shift is most visible at the top: founder Demis Hassabis is being moved to a chairman role, and AI researcher Koray Kavukcuoglu will take over day-to-day operations from the U.S. without the CEO title. Hassabis, who reportedly wanted to leave alongside departing executive Jeff Dean, is said to be planning an orderly exit to focus on his pharmaceutical startup, Isomorphic Labs.

All Gemini-related development is now being centralized in the Bay Area, with Google co-founder Sergey Brin stepping into a more influential role after coming out of retirement during the company’s initial “red alert” over ChatGPT. This completes a process that began with the controversial merger of Google Brain and DeepMind, a shotgun wedding that reportedly frustrated Hassabis and his team from the start. The loss of a dedicated chief executive and the geographic shift signal that DeepMind’s voice is being absorbed into Google’s broader machinery.

Industry analysts are split on what this restructuring actually means, and they might both be right. SemiAnalysis delivers a brutal read in their newsletter, arguing that Google is losing the AI model race due to a risk-averse, bureaucratic culture that starved its own researchers of compute while chasing short-term cloud revenue. They compare it to IBM abandoning the PC business or Intel protecting legacy chips—historic blunders. The financial incentives support this cynicism: Gemini’s annualized revenue hit $12 billion in Q2 2026, a rounding error next to the projected $73 billion in Google Cloud AI infrastructure revenue and $120 billion in TPU sales expected by the end of 2027. Selling the picks and shovels is simply more profitable than winning the AI race.

Tim O’Reilly offers a counter-narrative, arguing that Google is playing a different game entirely. In a major tech shift, he points out, the inventors rarely win—the ones who democratize the technology do. He uses the example of Thomas Edison, who invented the lightbulb, losing the electricity war to George Westinghouse’s superior AC grid. Google’s efficiency-optimized Flash models fit this thesis: smart enough to be useful, cheap enough to be everywhere. But the rumor mill complicates this rosier interpretation. Gemini 3.1 Pro remains stuck in preview, and the next-gen 3.5 Pro, announced for a June release, has reportedly been shelved. That smells less like a deliberate infrastructure play and more like a team that simply couldn’t keep up with the frontier.