Why Demis Hassabis Is Stepping Back and What It Means

Demis Hassabis is moving out of the daily CEO grind at DeepMind. The change feels less like a departure and more like a relocation of influence. After leading the lab since 2010, Hassabis will leave the CEO role to take on the chairmanship of DeepMind's board and a new title as chief scientist at Alphabet. He will also keep running Isomorphic Labs, the AI drug discovery company he founded.

What actually changed inside Google and DeepMind

Sundar Pichai, Google’s CEO, says the day-to-day leadership of DeepMind will pass to Koray Kavukcuoglu. Once DeepMind’s chief technology officer, Kavukcuoglu now assumes a senior vice president role responsible for operations and will report directly to Pichai. The shift signals a tighter alignment between DeepMind and the broader Alphabet machine learning effort, with Hassabis stepping back to advise on longer-term strategy, especially around artificial general intelligence and health applications.

Hassabis has been explicit about his priorities. He argues that AI should prove its worth by improving human health. That means pushing on drug discovery and therapeutics as a clear, measurable way to show what advanced AI can do. Expect him to spend more time shaping research agendas across Alphabet while staying hands-on at Isomorphic Labs.

There is also change at the top of DeepMind’s technical roster. Jeff Dean, one of Google’s earliest engineers and a long-time senior research leader, has decided to leave. He’s founded a new startup called Discovery Loop, and Google is an early investor. In addition to funding, Google will provide cloud infrastructure. Discovery Loop aims to build AI systems that can iterate and improve themselves with minimal human intervention, automating complex experimentation cycles across machine learning and scientific research.

It is a striking moment. Key figures who have shaped modern AI are reconfiguring their roles. Some move to new ventures. Others refocus on strategic influence. The net effect is less a collapse and more a redistribution of talent and ambition across Alphabet and the broader ecosystem.

One more operational note: parts of the AlphaFold project have been reorganized, with staff reassigned or reduced in recent weeks. AlphaFold is widely celebrated for transforming how scientists predict three-dimensional protein structures, and its tools have become central to research into neurodegenerative diseases such as Alzheimer’s and Parkinson’s. The team’s reallocation suggests Alphabet is balancing investments between immediate scientific tools and longer-term, high-risk efforts in general AI and automated discovery.

So what should observers read into this? First, DeepMind is not being shut down. It is being re-positioned. Hassabis will continue to drive the vision, but in a role designed to influence the whole Alphabet portfolio. Second, Google is backing spinouts and startups that can accelerate scientific progress by combining automated experiments with scalable AI. Third, the departure of senior researchers to found new companies reflects a broader phase in AI: teams that proved breakthroughs are now experimenting with new organizational forms to turn those breakthroughs into reliable, real-world impact.

For the field of AI and for anyone watching the race to apply machine learning to medicine, this is a moment to expect both rapid technical iteration and strategic hedging. Will AI-driven drug discovery deliver the headline cures investors and patients hope for? That remains to be seen. But with leaders repositioning themselves around health, autonomy, and infrastructure, the next chapters will be written where research meets disciplined productization.