Google Consolidates AI Leadership in California as Race With OpenAI and Anthropic Intensifies

Alphabet Inc.'s Google is pulling the center of gravity for its artificial intelligence efforts back to its Mountain View, California, headquarters, as it tries to regain momentum against Anthropic and OpenAI in the race to build the world’s most advanced AI models.
The restructuring has reshaped the top ranks of Google DeepMind, the unit created in 2023 when Google merged its Brain and DeepMind labs. Koray Kavukcuoglu, a longtime DeepMind researcher who relocated to the Bay Area within the past year, has taken day-to-day control of Google's sprawling AI research and operations. Demis Hassabis, DeepMind's co-founder and CEO, has stepped back from operational duties to serve as chairman of Google DeepMind and chief scientist at Alphabet. Sebastian Borgeaud, who leads a critical AI coding effort, has also moved from London to California, while a number of other London-based staff have left the company entirely.
The centralization is an attempt to fix a structural problem that has dogged Google since it combined Google Brain and DeepMind under one banner. The labs shared a name but remained split across continents, with Brain based in Mountain View and DeepMind rooted in London. That arrangement complicated decision-making and frustrated researchers on both sides, according to people familiar with the matter.
Kavukcuoglu is formally inheriting responsibility at a precarious moment. Google is months behind schedule on shipping the most powerful version of its flagship Gemini model, and the company is widely seen as an also-ran in coding automation, one of AI's most commercially valuable early markets. The pressure is only heightened by the billions Google is pouring into the fight: the company has said it expects to spend as much as $205 billion this year on capital expenditures to support its AI push.
Employees' frustration with that competitive position has triggered a steady outflow of senior talent. Some have left to found startups, including Google engineering luminary Jeff Dean, while others have joined Anthropic or OpenAI, lured by the chance to work at the frontier and the prospect of an upcoming initial public offering.
The departures have hit DeepMind's London office especially hard. David Silver, a distinguished researcher there, left earlier this year to start Ineffable Intelligence, a startup that quickly raised $1.1 billion in seed funding. Simon Bouton, DeepMind's chief experience officer and a 13-year veteran overseeing operations, is also preparing to leave for CuspAI, a London-based startup, according to a person familiar with the matter. Bouton did not respond to a request for comment.
Google insists London still matters. The company says it remains a crucial hub for AI talent and recently opened a new office there. Hassabis, who will stay in London, is expected to focus on AI vision and strategy rather than the daily scramble to ship products.
The shuffle is also a reminder of Silicon Valley's gravitational pull in an industry where proximity to decision-makers can matter as much as research breakthroughs. For London, which has long positioned itself as Europe's AI hub, losing senior executives to California is a familiar source of anxiety.
The net effect, after a year of movement, is that the leaders most responsible for Gemini — the model at the center of Google's AI strategy — are now working from the same place. That was not true when the AI boom began. Sundar Pichai, Google's CEO, and Sergey Brin, its co-founder, have been frequent presences on the floor dedicated to Gemini employees, according to one Google employee. Hassabis, by contrast, was rarely seen there. Over time, oversight of Gemini effectively shifted to Kavukcuoglu.
Brin seemed to acknowledge as much at an AI conference in June, when asked how he divided responsibilities with Kavukcuoglu and Hassabis. Brin described himself as deeply involved with Gemini and said he works closely with Kavukcuoglu.
“I would say, I poke and prod him and the team about, ‘Hey, are you really doing that?’ Sometimes a little bit disruptive, not going to lie,” Brin said. “But Koray actually organizes the groups and has them deliver stuff.”
Kavukcuoglu has long been a key figure inside DeepMind, though a less familiar face to the outside world than Hassabis. He founded DeepMind's deep learning team and helped produce several of the lab's most important research breakthroughs. Last summer, he was promoted to Google's chief AI architect. Yann LeCun, the NYU professor and deep learning pioneer who mentored Kavukcuoglu as a PhD student, called him “a great engineer” who honed strong research management skills during his rise at DeepMind. LeCun said Kavukcuoglu's early work helped persuade the field that AI could fundamentally change how machines process language.
Kavukcuoglu now faces the hard part: showing that Google can still out-build the startups nipping at its heels, while dealing with the loss of some of its most prominent AI leaders. He is the only remaining co-lead of Gemini still at the company.
