Google has reorganised its DeepMind artificial intelligence division in a sweeping shake-up that removes Demis Hassabis from day-to-day chief executive duties and concentrates power in Mountain View, California, as the search giant scrambles to close the gap with OpenAI and Anthropic in the race toward artificial general intelligence.
Hassabis, who shared the 2024 Nobel Prize in Chemistry for his work on protein-structure prediction, will step back from operational leadership to become chairman of DeepMind and chief scientist of parent company Alphabet. The move, disclosed this week, shifts executive control toward Google's headquarters in California while retaining London as a research hub under new senior leadership including Koray Kavukcuoglu.
The restructuring follows a wave of researcher departures from DeepMind that had raised questions about the division's ability to retain talent amid intensifying competition. By elevating Hassabis to a purely scientific role, Google is effectively acknowledging that his strengths lie in research breakthroughs rather than the organisational demands of running a commercial AI laboratory at scale.
Industry analysts described the reorganisation as a defensive manoeuvre. OpenAI and Anthropic have captured much of the public imagination and venture capital flowing into frontier AI, while Google's Gemini family of models has struggled to establish a clear market leadership position despite substantial technical capabilities. The shift to Mountain View is widely interpreted as an effort to accelerate product development by placing DeepMind closer to Google's consumer-facing engineering teams.
Hassabis's new mandate centres on artificial general intelligence, the still-theoretical threshold at which machines can match or exceed human cognitive abilities across virtually any task. In public remarks accompanying the announcement, he framed the role as a return to his foundational interests, suggesting that the administrative burdens of running DeepMind had distracted from pure research.
The timing is hardly accidental. Anthropic disclosed this week that it is building custom AI chips, hiring semiconductor engineers at salaries reaching $485,000 in a bid to reduce dependence on Nvidia hardware. OpenAI continues to expand its GPT-5.6 model family while planning its developer conference for late September in San Francisco. Google's competitors are moving aggressively on multiple fronts simultaneously.
Google DeepMind has also pushed forward with applied research, including the WeatherNext model for cyclone forecasting and ongoing refinements to Gemini agent tools. Yet these achievements have generated less public traction than the headline-grabbing products from rival labs, feeding internal pressure to demonstrate that Google's decade-long investment in AI research can translate into market-defining consumer and enterprise products.
The personnel changes extend beyond Hassabis. Multiple senior researchers have departed DeepMind in recent months for competing labs or startup ventures, a pattern that mirrors the broader talent migration reshaping the AI industry. Google's response has been to consolidate authority under a smaller leadership team with tighter lines of reporting to Alphabet's senior executives.
For Alphabet shareholders, the reorganisation carries both promise and risk. Concentrating AI leadership in Mountain View could streamline decision-making and accelerate product launches, but it also risks alienating the London-based research talent that built DeepMind's reputation. The dual-hub structure, while politically necessary given British regulatory scrutiny of the division, may prove operationally cumbersome.
The broader question is whether any organisational restructuring can address Google's fundamental challenge in the AI era: converting research supremacy into product dominance. DeepMind's scientific achievements are unmatched, yet the commercial landscape is increasingly defined by user-facing applications rather than academic citations. Hassabis's retreat to the laboratory may be exactly what the science requires, but whether it serves Google's business ambitions remains an open question.
Image source: i.ibb.co