Google's AI Empire Is Growing. So Is the Brain Drain.

Alphabet's cloud division posted 82% revenue growth in its most recent quarter, and Alphabet CEO Sundar Pichai reported that 90% of Fortune 100 companies are using Gemini Enterprise. Yet at the same moment that Wall Street was digesting those figures, Google's AI organisation was absorbing a significant shakeup: chief scientist Jeff Dean announced his departure after 27 years, and Google DeepMind CEO Demis Hassabis stepped down from daily management of the division to become its chairman. Alphabet shares, which had already slipped following the earnings report on capital expenditure concerns, fell further on the news.

Dean is leaving alongside prominent Google researchers Sanjay Ghemawat, Oriol Vinyals, and Quoc Le to found Discovery Loop, a public benefit corporation backed by Google whose mission, Dean wrote on X, is to 'automate machine learning, science, and engineering to accelerate discoveries and progress.' The departures extend a pattern that has stripped Google of some of its most consequential technical minds. Noam Shazeer, one of the authors of the landmark 2017 paper 'Attention Is All You Need' — which provided the foundation for generative AI — left for OpenAI in June, less than two years after Google paid nearly USD 3 billion to bring him back through an acquihire. Nobel laureate John Jumper subsequently left DeepMind for Anthropic. All eight authors of the 2017 transformer paper have now left Google.

Why it matters: the tension at the heart of Google's predicament has consequences well beyond personnel announcements. The company's homegrown tensor processing units, or TPUs, sit at the centre of a resource allocation battle that is fuelling researcher frustration. Some researchers have grown frustrated watching Google Cloud sell TPUs to outside customers — including Anthropic, whose models compete directly with Gemini — while their own ambitious projects go under-resourced, according to people familiar with the matter who asked not to be named due to confidentiality. Every TPU assigned to training a model, fulfilling a cloud contract, or serving a consumer product reflects a choice among competing priorities, and, as Dan Niles, founder of Niles Investment Management and a Google shareholder, noted, 'somebody's always going to be unhappy in that situation.' For the broader AI industry, Google's struggle to retain frontier researchers while monetising their work at scale is a case study in the structural difficulty of housing a basic-science culture inside a multi-trillion-dollar commercial enterprise.

Gil Luria, an analyst at D. A. Davidson who recommends holding Alphabet stock, offered a pointed diagnosis of why researchers keep leaving. 'They're not interested in commercialising AI,' Luria said. 'They're interested in being part of history, and so they look at Anthropic, OpenAI or another startup as being the place where they can pursue history.' Google's bureaucracy compounds the pull of smaller rivals, with layers of approval required to move research into products. That friction makes labs such as OpenAI and Anthropic — and even younger startups — more appealing, especially for researchers who prefer lab work to balance sheets.

Pichai has sought to reassure investors and researchers alike that DeepMind's compute needs remain the company's top priority. Asked about TPU allocation in July, he said Google's 'first priority' is securing the compute needed to compete at the frontier in AGI development, calling that work 'the foundation for everything we do.' He added that Google increasingly places TPUs directly in third-party data centres to help satisfy external demand without cannibalising internal capacity. Whether that assurance is sufficient to slow the talent drain remains an open question. Tomasz Tunguz, founder of Theory Ventures, argued that for most commercial applications, the compute arms race may already be beside the point. 'Many of the models that are reasonable are good enough,' Tunguz said, echoing a view shared by Niles, who put it more bluntly: 'You don't need a Ferrari for this stuff. A Ford will work for 90% of the use cases.' For the scientists trying to produce the next transformer-scale breakthrough, however, good enough has never been the goal.

With Dean's exit, Koray Kavukcuoglu — DeepMind's technology chief and Alphabet's chief AI architect — will take over daily management of DeepMind and lead development of the next Gemini model. According to a person close to the DeepMind team, Kavukcuoglu had already been absorbing a broader set of responsibilities from Hassabis over the past year, including directing model development and presenting major Gemini releases. Hassabis will take on the newly created role of chief scientist at Alphabet, focusing on longer-term research and the societal implications of artificial general intelligence, and plans to devote more time to Isomorphic Labs, the AI drug-discovery company that grew out of DeepMind. Google Cloud CEO Thomas Kurian, who has led the division since 2019 and transformed it into the company's fastest-growing business, is expected to play an increasingly prominent role alongside the reorganised DeepMind leadership as the company works to bring its research and commercial operations closer together — a shift already symbolised by the joint appearance of Hassabis and Kurian at the World Economic Forum in Davos, which one person familiar with Google Cloud's operations described as striking given how separately the two organisations have historically operated.