Google’s AI leadership is changing, and this one feels bigger than a routine executive shuffle.
Koray Kavukcuoglu, a longtime DeepMind researcher who became one of Google’s most important AI executives, is taking on expanded responsibility for Google DeepMind as the company reorganizes the people building Gemini and its next generation of frontier AI systems.
Demis Hassabis isn’t leaving. That distinction matters.
Hassabis is moving away from much of the day-to-day operation of DeepMind and into the role of chair of Google DeepMind and chief scientist at Alphabet. Kavukcuoglu, meanwhile, is moving closer to the machinery: model development, research teams, Gemini products and the increasingly urgent job of turning Google’s enormous AI research operation into products people actually use.
Koray Kavukcuoglu Is Suddenly One of Google’s Most Important Executives
Suddenly might be the wrong word.
Kavukcuoglu has been inside DeepMind for well over a decade. Long before Gemini became a consumer brand, he was working on the kind of research that helped establish DeepMind’s reputation in machine learning.
He was involved in the early Deep Q-Network work that showed a neural network could learn to play Atari games directly from pixels. That research became an important milestone in deep reinforcement learning.
Google describes Kavukcuoglu as one of its leading AI experts. Before his latest expansion in responsibility, he served as Google DeepMind’s CTO and as Google’s chief AI architect, overseeing generative AI model development and helping bring those models into Google products.
That last part is probably the more interesting one now.
Google does not need another research trophy.
It needs execution.
Gemini Is Becoming the Center of Google’s AI Operation
DeepMind spent years being associated with ambitious research projects: AlphaGo, AlphaFold, reinforcement learning, scientific discovery.
Gemini changed the center of gravity.
Google now needs its frontier models to compete inside search, smartphones, productivity software, developer tools, cloud services and standalone AI assistants at roughly the same time.
That is a very different job from running a prestigious research lab.
Under the new structure, Kavukcuoglu is expected to oversee frontier AI research and Gemini model development while holding responsibility across Gemini’s product and developer teams. He will report directly to Alphabet CEO Sundar Pichai.
There is an obvious message buried in the org chart: research and product development are getting pulled closer together.
Google has incredible AI talent. Nobody seriously disputes that.
The uncomfortable question has been whether Google can turn all of that talent into products quickly enough.
Demis Hassabis Isn’t Really Stepping Away From AI
Calling this simply a Hassabis departure would miss what Google is doing.
Hassabis is moving into a broader scientific role as Alphabet’s chief scientist while becoming chair of Google DeepMind. That gives him room to concentrate more heavily on artificial general intelligence and scientific research instead of carrying the full operational burden of one of the world’s largest AI organizations.
And he still has another major project on his plate: Isomorphic Labs, Alphabet’s AI-driven drug discovery company.
So this looks less like Google pushing Hassabis aside and more like separating two jobs that have become enormous.
One person focuses on long-range science.
Another has to make the AI machine move.
Kavukcuoglu gets the second job.
The Timing Is Hard to Ignore
The reshuffle comes while competition at the frontier of AI is relentless.
OpenAI and Anthropic continue pushing new models and products. Google has responded by embedding Gemini across more of its ecosystem while investing heavily in the infrastructure needed to train and serve increasingly capable models.
At the same time, Google has dealt with departures from some unusually prominent AI researchers.
Jeff Dean, one of the architects of Google’s modern AI organization, is leaving after 27 years to start Discovery Loop. Several other researchers are joining him. Google plans to support the new company as an investor and cloud partner.
That makes Kavukcuoglu’s expanded authority more consequential.
Google isn’t just changing titles while everything else stays still. Some of the people who shaped its previous AI era are changing roles or walking out the door entirely.
Google Has the Resources. Speed Is the Harder Problem.
There is something strange about describing Google as an AI underdog.
It owns global-scale data centers. It builds custom AI chips. It has billions of users. Search, YouTube, Android, Gmail, Workspace and Google Cloud give the company distribution that most AI startups could never reproduce.
Then there is DeepMind itself.
Few organizations have contributed as much foundational AI research.
Yet the generative AI boom created a market where possessing great research isn’t necessarily enough. A competitor can launch a model, wrap it in a simple product and have millions of people testing it before a larger company has finished deciding where the feature belongs.
Google appears increasingly determined to remove that gap.
Kavukcuoglu’s rise makes sense in that context. His role sits directly between model research and deployment — exactly where Google has the most to gain.
This Is Really About What Google DeepMind Becomes Next
The fascinating part isn’t whether Kavukcuoglu receives a CEO-style title. It’s what happens to DeepMind under him.
DeepMind began as an unusually research-heavy AI laboratory. After its merger with Google Brain in 2023, it became the center of Google’s broader AI development effort. Now Gemini is pulling the organization even closer to Google’s commercial products.
That could make Google faster. It could also change the culture that made DeepMind unusual in the first place.
Kavukcuoglu now has to navigate both. He needs researchers willing to chase ideas that may not become products for years, while Google simultaneously needs Gemini improvements that can ship this quarter. That tension isn’t going away. And with AI competition moving at its current pace, Google probably doesn’t want it to.
Sources
- CNBC — Google DeepMind: Koray Kavukcuoglu
- Axios — Google’s AI leadership shuffle
- Google — Koray Kavukcuoglu official author profile
- MarketWatch — Alphabet stock and Google AI leadership changes
- Playing Atari with Deep Reinforcement Learning
