Demis Hassabis was shifting away from DeepMind CEO duties for a year
The CEO of Google’s AI unit was increasingly unsatisfied with the role of tech executive, preferring a more scientific position, said two people familiar with his thinking.
In 2014, Google DeepMind CEO Demis Hassabis, who co-founded the company in 2010 and sold it to Google, decided to step away from his day-to-day responsibilities, shifting them to Koray Kavukcuoglu, the company's chief AI architect. This move, which coincided with the departure of chief scientist Jeff Dean, was at least a year in the making, according to sources familiar with Hassabis's thinking.
Hassabis, a Nobel Prize winner, was reportedly becoming unsatisfied with his role as a tech executive, as opposed to a visionary scientist. He has been more animated in recent times discussing Isomorphic Labs, Google's biotech spinout, where he focuses on using AI to solve scientific puzzles and prevent catastrophic harm to humanity.
The management change was met with a 4% drop in Google's stock price and raised concerns about the company's position in the fast-paced AI race. Despite reporting a 24% revenue increase to $120 billion, Google's cloud revenue grew by 82%, its capital expenditures are at record highs, causing its free cash flow to turn negative. However, executives believe Hassabis's departure could help accelerate AI development.
While Hassabis was the face of DeepMind's AI efforts, he wasn't focusing on what made the company stand out in the AI race. Kavukcuoglu, who joined DeepMind in its early days, now oversees the highly anticipated Gemini 4 model. This leadership change also brings up the question of whether Google can still attract top AI talent, as the company has recently lost another Nobel laureate, John Jumper, to Anthropic.
Despite these challenges, Google remains well-positioned to win the consumer AI race due to its compute capabilities, existing consumer base, and control of the world's most popular mobile operating system.
Written by urgent.news from Semafor's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.