DeepMind Morale Issues Blamed for Gemini 3.5 Pro Delays
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google DeepMind is suffering from poor morale that is contributing to delayed model releases, per six current and former employees who spoke to Axios anonymously, with employees split between citing the Pentagon deal and feeling behind competitors as the root cause.
- Gemini 3.5 Pro, Google's most powerful model in development, is months behind schedule per Bloomberg, and none of Gemini's models rank in the top 10 most-used on OpenRouter's LLM leaderboard.
- Noam Shazeer, a Gemini co-lead, left for OpenAI, and Nobel Prize in Chemistry winner John Jumper departed for Anthropic, while multiple researchers publicly resigned over Google's April deal allowing the Pentagon to use its technology.
- Google's free cash flow turned negative as it spends $190 billion on AI this year, though the company posted 82% cloud revenue growth and search revenue came in slightly below Wall Street expectations.
- Google disputes the morale narrative, stating that AI talent attrition rates in the first half of this year are lower than the same period last year and that more than 90% of those offered AI roles accept the position.
- Former DeepMind research scientist Alex Turner told Axios that CEO Demis Hassabis lacks the consistent internal presence that competitors like Sam Altman and Dario Amodei maintain with their employees, describing a 'constant battle' causing 'emotional burnout' for those morally opposed to the military contract.
- DeepMind didn't prioritize agentic coding while racing to defend search from ChatGPT, per a former model training employee, even as Axios notes that leapfrogging is the norm — Google led the race just last year before Anthropic's Mythos and now OpenAI took turns at the front.
- Google released a series of smaller, more efficient models this week to mixed reactions, with Meta's Alexander Wang posting 'Gemini who?' on X, though the company argues that delivering cost-efficient models meets enterprise demand even if they aren't at the leading edge.
Why it matters: Google is burning $190 billion on AI this year while its free cash flow turns negative, making its Gemini roadmap critical to keeping investors patient. The departure of marquee researchers like Noam Shazeer to OpenAI and Nobel winner John Jumper to Anthropic, combined with the Pentagon deal driving resignations, gives competitors a window while cloud revenue's 82% growth shows the infrastructure business is still carrying the load.
