LONDON–The research lab that produced AlphaGo and the models behind Google’s Gemini has spent the past year navigating something it rarely experienced at its peak: internal strain. When Google announced that Demis Hassabis would step down as chief executive of DeepMind to become chief scientist, the official narrative was one of orderly transition. A new report from Fortune, published Aug. 10, describes a more complicated reality: sagging morale, a steady outflow of researchers, and delays in the development of new models.
The departures are the most visible symptom. DeepMind has lost a succession of senior researchers over the past two years, including leaders of major research programs, and the report describes the attrition as a slow drain that has reshaped the lab’s ranks. Some left for startups, some for competitors, and some for academic positions, but the pattern has been consistent enough that industry recruiters now treat DeepMind alumni as a pipeline.
The morale problem has multiple sources. The report describes frustration with the pace of model development, with the lab’s role inside a larger corporate structure, and with the growing pressure to produce commercial results. DeepMind was founded as a research-first organization, and its researchers have watched as the center of gravity in AI shifted toward product cycles and deployment deadlines.
The delays are the part with the most consequence. New models from DeepMind’s teams have slipped from internal schedules, according to the report, citing people familiar with the lab’s operations, and the delays have affected the Gemini roadmap that Google has staked its AI future on. In a race where months separate leaders from followers, slipped timelines carry an outsized cost.
Google’s response to the internal news has been to emphasize continuity. The company has said the leadership transition was planned, that Hassabis remains deeply involved as chief scientist, and that DeepMind’s research priorities are unchanged. It has also pointed to the scale of its AI investment and the breadth of its research portfolio as evidence of its commitment.
The reality on the ground is more mixed. The report describes a lab where the research culture that made DeepMind famous has been diluted by the demands of shipping products, and where the autonomy researchers once enjoyed has been constrained by the needs of Google’s broader organization. The tension between research ambition and corporate reality is not new, but the stakes have risen as AI has become Google’s most important competitive front.
The talent outflow is a competitive problem for Google in a specific way. AI research is a field where individual researchers carry enormous value, and the companies that attract the best minds tend to produce the best models. Every senior researcher who leaves DeepMind takes not just expertise but also the ability to attract more talent, and the report describes recruitment efforts aimed at backfilling the losses.
The industry context sharpens the challenge. OpenAI, Anthropic, and a growing list of well-funded startups are competing for the same researchers, offering equity packages that Google’s compensation structure struggles to match. The report describes specific cases in which researchers left for opportunities they considered more aligned with their research interests, and the pattern suggests that money is not the only factor driving the exodus.
The Gemini roadmap is the near-term casualty risk. Google has positioned Gemini as the model family that will power its search, cloud, and consumer products, and delays in its development ripple through all of those businesses. The report’s account of slipped schedules, if accurate, means Google’s AI products will face the market with a model that is behind where the company had planned.
Hassabis’s new role is itself a signal. The founder of DeepMind built the lab around his vision of AI research, and his move to chief scientist suggests a desire to return to the research that made his name, even as he keeps a hand in the lab’s direction. The succession, with a new chief executive drawn from inside the organization, is designed to preserve continuity, but the report questions whether any single person can replace the gravitational pull Hassabis had on the lab’s culture.
Google’s AI position is not in imminent danger. The company remains one of the few organizations in the world with the compute, data, and talent to compete at the frontier, and its research output across AI remains enormous. But the frontier is crowded, and the report’s account of DeepMind’s internal strain suggests that Google’s advantage is thinner than its public posture implies.
For the industry, the report is a reminder of how much of AI’s progress rests on a small number of institutions and the individuals within them. The lab that defined the modern era of AI research is navigating the transition from a research institution to a product organization, and how it manages that transition will shape Google’s AI trajectory for years. The report’s details are disputed by Google, which called the account incomplete, but the underlying picture, of a lab under strain at the moment of its greatest importance, is one the company has not fully rebutted.
