According to news website Axios, six current and former Google DeepMind employees revealed that low employee morale is one of the reasons for the delay in the release of Google's flagship model Gemini 3.5 Pro. Bloomberg previously reported that the model has been months behind schedule, and the recent series of small models released by Google have received mixed market reactions. Even Meta AI's head, Wang Tao, sarcastically said on X, "Gemini? I've never heard of it."

Top talent continues to leave, military agreements become a centrifugal force

Google is currently facing serious talent loss. The co-lead of Gemini, Noam Shazeer, has moved to OpenAI, and Nobel Prize-winning chemist John Jumper has joined Anthropic. Several employees resigned due to Google's military cooperation agreement with the U.S. Department of Defense in April, which was frequently mentioned during their exit interviews. Alex Turner, a former research scientist who left due to the military contract, stated that CEO Hassabis appears internally much less frequently than competitors' CEOs, Altman and Amodei. One employee openly said, "We are falling behind," and the Gemini series has failed to rank in the top ten in the OpenRouter large language model rankings.

AI investment leads to negative free cash flow, Google denies morale issues

Google's total spending this year is expected to reach $19 billion, and its free cash flow has turned negative, with growing concerns about the returns on AI capital expenditures. Google denies that employee morale issues have caused the model to lag, stating that the AI talent attrition rate this year is lower than the same period last year, with over 90% of those offered positions accepting them. A Google spokesperson said they are satisfied with the Flash series of models and the future roadmap. However, the report pointed out that while Google can afford to be temporarily overtaken by competitors, it cannot afford talent loss and investor impatience.