An alert pinged: Demis Hassabis is stepping back from the day-to-day at DeepMind. I read the announcement twice. People in meeting rooms across the company paused and reopened their calendars.
I’ve followed Google’s AI moves for years, and you should pay attention here: this is not a cosmetic shuffle. You’ll see talent, money, models, and reputation all being weighed at once.
Posters of AlphaGo still line the DeepMind hallways.
Hassabis will stop running the lab every day and become chair of Google DeepMind while taking on a new role as Alphabet’s chief scientist. I know him as someone who has framed his life’s work around artificial general intelligence, and he told employees that AGI feels “close at hand.”
He’ll keep leading Isomorphic Labs, the company’s AI-driven drug-discovery arm, while Koray Kavukcuoglu, the current CTO, moves into the senior vice president spot to run daily operations.
Why did Demis Hassabis step down as DeepMind CEO?
Internal pressure and external pace played together. Google delayed Gemini 3.5 Pro after it missed internal goals, and reports point to low morale in parts of the lab. You should read that as a technical shortfall plus a management signal: the company needs focused operational leadership while Hassabis pursues broader scientific aims.
Outside the lab, Google Cloud dashboards tell a different story.
Revenue from Google’s cloud unit jumped 82% year over year to $24.8 billion (€22.8 billion) in Q2. That surge sits beside a forecasted capital spend between $195 billion and $205 billion (€179 billion–€189 billion) this year, aimed largely at AI infrastructure.
That math explains the pressure cooker: heavy spending, high expectations, and a market that already prizes execution as much as research. Sundar Pichai framed the rework as a speed and focus play — “We have to accelerate all this work and stay focused on the AI frontier,” he wrote — but acceleration costs cash and patience.
Who will run DeepMind now?
Koray Kavukcuoglu is the new senior vice president in charge of daily operations. He’s a long-time DeepMind technologist with deep engineering cred. For you, the change signals a pivot toward operational rigor and product delivery rather than public-facing research theater.
On campus and across the valley, people are changing addresses.
Jeff Dean, a 27-year Google veteran and most recently Google DeepMind’s chief scientist, is leaving to start Discovery Loop, a public-benefit corporation aimed at automating machine learning and scientific discovery. Google will be a founding investor and cloud partner for the new company.
That move removes a heavyweight from the building—and hands a new set of expectations to remaining leaders. It feels less like abandonment and more like a redistribution of influence across academia, startups, and corporate labs.
What does this mean for Gemini, Google Cloud, and the AI race?
Expect three intertwined outcomes: product timelines will be pressure-tested; infrastructure bets will keep eating capital; and talent flows will determine which labs set the pace. Google faces a real-time rivalry with OpenAI, Anthropic, and Microsoft (where Mustafa Suleyman now leads Microsoft AI). If Gemini 3.5 Pro needs more months, that delay will shape customer sentiment and partner deals.
Hassabis’s move reads as a strategic recalibration—like a chess grandmaster stepping away from the clock—retaining influence while freeing time for long-range scientific direction. Passing day-to-day leadership to Kavukcuoglu is a relay race handoff aimed at speed and continuity.
For you tracking where tech power concentrates, this is where compute, capital, and culture collide: Google has world-class talent and massive compute through Google Cloud, but rivals are nimble and hungry. The question now is whether the new structure accelerates product delivery or only reshuffles responsibility—will it be enough to keep Google leading the next chapter of AI?