Several current and former Google DeepMind staff members have disclosed that Google's AI research division has delayed the launch of its premier model, Gemini 3.5 Pro, in part due to a slump in employee morale. The model is already months behind its intended release date, and the Gemini series has not made it into the top ten rankings for large language models. Google's recently launched smaller models have met with mixed responses in the market and have even drawn ridicule from the head of Meta AI.

Google is currently experiencing a significant loss of top talent. Notably, the co-lead of the Gemini project has already left the company, and several employees have resigned in protest over Google's military collaboration agreement with the U.S. Department of Defense. Some staff members feel that the company is lagging behind its competitors and have noted that CEO Hassabis is seen less frequently within the company compared to the CEOs of rival firms.

Google's overall expenditures are projected to hit $190 billion this year, with free cash flow dipping into negative territory. This has sparked concerns among investors regarding the return on investment for AI-related capital expenditures. Google, however, refutes claims that morale problems are responsible for the model's delay, stating that AI talent attrition in the first half of the year was actually lower than during the same period last year. The company also expressed confidence in the Flash series models and its future plans.

Nevertheless, reports suggest that while Google can tolerate being temporarily outpaced by competitors, it cannot afford to lose key talent or to test the patience of its investors any further.