On 5 August 2026, the Google DeepMind reshuffle put Koray Kavukcuoglu in charge of the lab’s day-to-day operation while Demis Hassabis moved into a wider scientific role across Alphabet. Google says Kavukcuoglu, previously DeepMind’s chief technology officer and Google’s Chief AI Architect, will oversee Gemini model development, frontier AI research, the Gemini app and developer teams. Hassabis becomes Chair of Google DeepMind and Chief Scientist of Alphabet while continuing to lead Isomorphic Labs, according to Google’s leadership announcement.
The timing matters because Gemini is already operating at enormous scale. Alphabet told investors on 22 July that the Gemini app had reached 950 million monthly active users, AI Mode in Search had passed one billion monthly active users, more than nine million developers were building each month with Google’s models, and its model APIs were processing about 22 billion tokens per minute. Those are company-reported figures from Alphabet’s second-quarter 2026 earnings call, not independent audience measurements.
Google DeepMind reshuffle: What Actually Changes
Google has not removed Hassabis from DeepMind. It has separated the strategic and operating responsibilities that had increasingly sat with the same person. Hassabis told staff that he wanted more time and space to focus on the bigger questions around artificial general intelligence, science and Isomorphic Labs. Kavukcuoglu now holds the operating brief for the models, research organisation, Gemini app and developer teams, while Hassabis remains connected as Chair and adviser. Google’s own messages to employees spell out that division.
What This Means for You
If you use Gemini, the announcement does not promise an immediate change to the app. The more important signal is organisational: the executive responsible for frontier-model development is also responsible for the product and developer teams that turn those models into services. For businesses and developers, that could shorten the distance between research, product decisions and deployment. It also makes Kavukcuoglu one of the people most directly accountable for whether Google’s AI research translates into dependable products.
That matters because the next phase of AI competition is increasingly about systems that act, not only chat. LiveAIWire’s guide to why major technology companies are racing to build agentic AI explains how model capability, tools, permissions and enterprise deployment are converging into one platform contest. Google is putting more of that chain under one DeepMind leader.
Koray Kavukcuoglu Was Already at the Centre of Google’s AI Machine
Kavukcuoglu is not an outside operator brought in to impose a new culture. Google’s current executive profile says he previously served as Vice President of Research at Google DeepMind, founded its deep-learning team and helped pioneer work including DQN and WaveNet. The same profile now lists him as Senior Vice President of Google DeepMind and Chief AI Architect at Google, leading development of DeepMind’s generative AI models and their integration into products at scale. Google’s executive biography provides the continuity.
That history matters because DeepMind’s identity has always depended on research depth. The leadership change does not look like a conventional product executive taking over from scientists. It looks more like Google asking one of DeepMind’s longest-serving technical leaders to own the difficult handoff between frontier research and mass-market execution.
Hassabis Is Moving Upstream, Not Walking Away
Google’s announcement says Hassabis will remain closely connected to Kavukcuoglu and other DeepMind leaders, advising across models and research while working with Sundar Pichai on strategic and global AGI matters. Hassabis also said he would lean further into Isomorphic Labs, the Alphabet-backed company applying AI to drug discovery. That makes the change an elevation into a broader scientific role rather than a clean departure from DeepMind. Google now lists Hassabis as Chair of Google DeepMind and Chief Scientist of Alphabet.
The opportunity for Google is clear. DeepMind’s daily organisation can move without requiring its most prominent scientific leader to personally carry every operating responsibility. The risk is equally clear: separating long-range strategy from execution only helps if the two sides remain tightly connected. Google’s structure now depends on that relationship working well.
Gemini Is Too Large to Be Managed Like a Research Experiment
Gemini is no longer a lab project searching for an audience. Google’s 950 million monthly-active-user figure puts the app close to the billion-user threshold, while AI Mode in Search has already crossed it. LiveAIWire has tracked that growth in its analysis of Gemini’s 950 million monthly active users. Whatever caveats apply to platform-reported usage, Google’s published Q2 remarks confirm the same 950 million figure, and the scale creates an operating problem very different from running a research lab.
A model decision can now become a product decision affecting hundreds of millions of people. Latency, safety, reliability, cost, developer tooling and release cadence all matter alongside benchmark performance. LiveAIWire’s comparison of ChatGPT, Gemini and Claude shows how users increasingly judge assistants as products, not research demonstrations. DeepMind’s leadership structure has to serve both worlds.
The Word That Matters Most Is Accelerate
Google’s own announcement says the company needs to accelerate its AI work and stay focused on the frontier. In management terms, that turns the reshuffle into a speed decision. One executive now owns more of the day-to-day path from frontier research to Gemini’s app and developer ecosystem, while Hassabis has more room for long-horizon science. The company is presenting the change as a way to move faster without abandoning the research culture behind DeepMind. Pichai’s message makes that urgency explicit.
Speed also depends on infrastructure. Frontier models require vast computing capacity, and LiveAIWire’s recent investigation into Big Tech’s long-term AI data-centre lease commitments shows how deeply model strategy is now tied to physical and financial infrastructure. Better models, lower serving costs and enough capacity to reach global users have become parts of the same execution problem.
Do Not Read the Reshuffle as Proof That AGI Is Around the Corner
Hassabis told staff that he feels AGI is close at hand. That is his judgement, not a demonstrated technical fact or a timetable. Google’s announcement provides no measurable definition of AGI and no date by which the company expects such a system to arrive. The defensible news is organisational: Hassabis has handed over day-to-day DeepMind operations while retaining a strategic scientific role. The distinction matters because the same announcement contains both the leadership facts and Hassabis’s personal assessment.
A compelling AI story becomes weaker when an organisational change is stretched into a technological prediction it cannot support. The reshuffle tells us what Google is prioritising. It cannot tell us when, or whether, a contested concept such as artificial general intelligence will be achieved.
Google Just Turned Gemini’s Next Phase Into an Execution Test
The Google DeepMind reshuffle gives the next phase of Gemini a clearer operator. Kavukcuoglu now has explicit responsibility for model development, frontier research, the Gemini app and developer teams. Hassabis has explicit room to focus on Alphabet-wide science, AGI strategy and Isomorphic Labs. Those responsibilities are now formally separated.
Whether the structure produces better models cannot be known on announcement day. What can be judged over the coming months is execution: release pace, reliability, developer adoption, product coherence and whether DeepMind can keep producing frontier research while serving one of the world’s largest AI user bases. Google has not reduced the ambition around Gemini. It has changed who is responsible for turning that ambition into a working system.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity, and the social impact of emerging technology. He publishes daily at LiveAIWire.com.