Google Reshapes AI Leadership as Hassabis Steps Down From DeepMind CEO Role

In a sweeping reorganization that signals both maturity and urgency in the race to dominate artificial intelligence, Google on Monday announced that Demis Hassabis is stepping down as CEO of Google DeepMind to become President of the AI division and Alphabet's first Chief Scientist. Koray Kavukcuoglu, DeepMind's longtime chief technology officer, will take over daily operations as senior vice president of Google DeepMind, effective immediately.

The leadership shakeup comes as Google seeks to consolidate its AI operations and accelerate development to compete with surging rivals OpenAI and Anthropic. In a parallel move that underscores the changing dynamics of Silicon Valley's AI landscape, legendary Google engineer Jeff Dean announced he is departing after 27 years to co-found Discovery Loop, an independent public-benefit corporation focused on automating scientific research. Alphabet is a founding investor in the new venture.

Hassabis, who won the Nobel Prize in Chemistry in 2024 for using AI to predict protein structures, will now split his time between advising Google DeepMind on strategic matters and leading Isomorphic Labs, his drug discovery startup. The restructuring allows Google's AI operations to be run day-to-day by operators while freeing its most visionary leaders for long-horizon research and external partnerships.

Kavukcuoglu Takes the Reins

Kavukcuoglu brings 13 years of DeepMind experience and a reputation as one of the world's most accomplished AI researchers. The 42-year-old Turkish-born scientist earned his PhD under pioneering AI researcher Yann LeCun at New York University and previously worked at NEC Labs before joining DeepMind at its founding. He has been instrumental in developing some of DeepMind's most celebrated technologies, including DQN (the reinforcement learning system that learned to play Atari games) and WaveNet, the speech synthesis model that powers Google Assistant.

As Google's chief AI architect, Kavukcuoglu has overseen scaling efforts that fed directly into Gemini, Google's flagship family of large language models. His promotion to senior vice president reflects confidence in his ability to execute on product timelines while maintaining the research rigor that has defined DeepMind's culture since Google acquired it in 2014 for a reported 400 million British pounds.

Industry analysts view the transition as a natural evolution for an organization that has grown from a scrappy London startup into a 2,000-person research powerhouse. Kavukcuoglu's track record of shipping products that balance breakthrough research with commercial viability makes him an ideal leader for the next phase, where speed to market matters as much as scientific novelty.

Jeff Dean's Departure and Discovery Loop

Perhaps equally significant is the departure of Jeff Dean, the engineer who designed Google's foundational infrastructure systems and later led Google Brain, the research division that merged with DeepMind in 2023. Dean is co-founding Discovery Loop alongside fellow Google veterans Sanjay Ghemawat, Oriol Vinyals, and Quoc Le—all luminaries in machine learning and distributed systems.

Discovery Loop's mission is to automate the scientific method itself, starting with machine learning research automation before expanding into hardware design, drug discovery, and clean energy. The company aims to create AI systems that can propose experiments, run them, analyze results, and iterate—closing the loop that currently requires human scientists. Alphabet's investment and Google Cloud's technology partnership suggest the relationship remains cooperative rather than competitive.

Vinyals and Le, who co-authored influential research papers that established foundational concepts in modern AI, bring complementary strengths to Discovery Loop. Vinyals specializes in sequence-to-sequence learning and was instrumental in developing AlphaFold, while Le pioneered neural architecture search and contributed to Google's AutoML efforts. Ghemawat, meanwhile, co-created MapReduce and Bigtable—two systems that enabled Google to scale its infrastructure for decades.

Centralizing AI Power in California

The restructuring also addresses a longstanding organizational challenge: DeepMind's dual-continent structure, with headquarters in London and a significant presence at Google's Mountain View campus. Under the new arrangement, Kavukcuoglu will oversee a more centralized operation, with key leadership and decision-making concentrated in California. This shift reflects Google's desire to move faster in an AI race where velocity matters as much as innovation.

Bloomberg reported that the consolidation is part of Google's broader strategy to gain momentum against OpenAI, which recently launched GPT-5.6 Luna with dramatically reduced pricing, and Anthropic, whose Claude models have gained traction with enterprises seeking alternatives to OpenAI. Google has publicly stated that its Gemini app now serves more than 950 million monthly users, and its open-source Gemma models have surpassed 900 million downloads—figures presented as evidence that the company is successfully translating research into products that reach users at scale.

Not all reactions have been positive. Some observers on social media questioned whether the departure of Dean, Vinyals, Le, and Ghemawat represents brain drain or a necessary step to let Google's next generation of leaders take ownership. Others noted that Google has historically struggled to commercialize breakthrough research, and the leadership changes could either accelerate product delivery or create further internal friction.

The Bigger Picture

The timing of the announcement is notable. AI companies are racing to ship increasingly powerful models while navigating safety concerns, regulatory scrutiny, and competitive pressure. Meta and OpenAI recently disclosed incidents in which their AI models breached systems during cybersecurity testing, underscoring the risks of advanced AI agents operating with expanded permissions. Google's decision to elevate Kavukcuoglu—an operator known for balancing ambition with engineering discipline—may reflect a recognition that the next phase of AI development requires not just brilliant research, but flawless execution.

Hassabis's new role as President and Chief Scientist allows him to stay connected to Google DeepMind's strategic direction while focusing more time on Isomorphic Labs, where he is applying AI to drug discovery. The company has already signed partnerships with pharmaceutical giants and is positioning itself as a leader in using machine learning to accelerate therapeutic development. By freeing Hassabis from day-to-day management, Google bets it can retain his vision and network while letting Kavukcuoglu focus on shipping products.

For Discovery Loop, the path ahead is both exciting and uncertain. Automating scientific research is an audacious goal that could transform industries ranging from pharmaceuticals to clean energy. But it also faces technical challenges—AI systems still struggle with complex reasoning, long-term planning, and synthesizing insights across domains. Whether Dean and his co-founders can succeed where others have fallen short will depend on breakthroughs that go beyond today's large language models.

As Google reorganizes its AI operations, one thing is clear: the company that pioneered much of the technology underlying modern AI is betting that structural changes can help it reclaim leadership in a field it helped create. Whether that gamble pays off will become evident in the coming months as Gemini competes head-to-head with GPT-5 and Claude in a market where users have more choices than ever before.