The departures began with a resignation letter, then another, then three more, and by the end of seven days Google had lost five senior researchers from its most prestigious AI laboratory. The researchers, who held leadership roles at DeepMind, have joined Anthropic, OpenAI and other rivals, according to people familiar with the moves, and their exits have forced Google to push back the release of Gemini 3.5 Pro to July.
The timing could not be worse for Google. The company has been fighting to hold its position in artificial intelligence against a group of younger, more aggressive competitors, and its Gemini models are the public face of that fight. Losing the people who build those models, in a cluster that a company spokesman described as unusually concentrated, raises questions about continuity that no press release can answer.
Alphabet’s stock has absorbed the news poorly. The company’s shares have come under pressure since the departures became known, and one widely circulated estimate, from the technology publication Memeburn, put the market value lost in connection with the talent drain at 270 billion dollars. The figure is rough and almost certainly overstated, but the anxiety behind it is real: investors are asking whether Google can keep winning the AI race without its best people.
The defections are part of a broader pattern. DeepMind, founded in London in 2010 and acquired by Google four years later, has long been considered the world’s best pure research lab in AI, the place where breakthroughs like AlphaGo were born. But its researchers are in demand, rival companies are paying extraordinary sums, and the flow out of the lab has accelerated as the competition for AI talent has intensified.
The exits are not the first. DeepMind has lost a steady stream of researchers over the years, and Google has usually responded by promoting internal talent and moving on. What is different this time is the concentration of the departures, five senior people in a single week, and the visible effect on a product schedule. Gemini 3.5 Pro, which was expected sooner, is now set for July.
Google has tried to counter the drain with money. The company has raised compensation for AI researchers, handed out stock packages that rival the offers from startups, and given top scientists more control over their projects. But money has limits against the appeal of building a new company’s models from scratch, or the chance to work on problems that Google’s bureaucracy makes difficult.
The deeper problem may be organizational. DeepMind operates inside Alphabet, a company of more than 180,000 people, and its researchers have complained, in private and occasionally in public, about the gap between research and product, about approval processes and about the size of the organization around them. Rivals offer smaller teams, faster decisions and equity in companies that could be worth hundreds of billions.
Anthropic has been the biggest beneficiary. The company, founded by former OpenAI executives, has built a research culture that attracts DeepMind defectors, and it has hired several of the departing researchers. OpenAI, despite its own churn, has also taken some. The flow of talent from Google to its rivals is now visible enough that hiring announcements from Anthropic are routinely scanned for DeepMind names.
Google has been here before. The company’s AI history is full of founders and leaders leaving, from the earliest days of DeepMind to more recent exits, and Google has repeatedly rebuilt its teams. What has changed is the market’s tolerance: in the era of AI competition, talent is treated as a strategic asset, and visible departures are read as evidence of deeper problems, whether or not they are.
The market’s reaction is worth taking seriously even if the 270 billion dollar figure is not. Alphabet trades at a valuation that assumes Google will remain a leader in AI, and anything that calls that assumption into question moves the stock. Google has responded with promises of a “Gemini era” and with renewed investment in its AI teams, but the departures suggest the problem is cultural as much as financial.
For the researchers who left, the calculation was simple: they can name their price, and they chose environments where their work reaches products faster. For Google, the challenge is to make DeepMind a place where people want to stay, which is harder than matching a salary. The next Gemini release, due in July, will show whether the lab can still deliver with a thinner bench. The departing researchers themselves have said little publicly, and people close to them describe the moves as career decisions rather than protests. But the destinations matter. Two of the five joined Anthropic, one joined OpenAI, and the others are heading to smaller AI labs that offer founding-level equity. The pattern, senior researchers leaving a giant for challengers, is one Google has seen in every technology wave, and it has not found an answer to it yet.
The week that emptied five senior chairs is a test Google did not choose, but it is one it must pass.
