Google DeepMind is hemorrhaging momentum in the artificial intelligence race, and the problem runs deeper than missed deadlines. Conversations with current and former employees paint a picture of a lab struggling with internal fractures: delayed model releases, departing top researchers, and widespread frustration over the company's military partnership.
The warning signs are everywhere. Gemini 3.5 Pro, billed as Google's most powerful model, is months behind schedule. When the company did ship a batch of smaller, cheaper models this week, competitors didn't hesitate to take shots. Meta's Alexander Wang mocked the releases on social media, asking simply "Gemini who?" Meanwhile, Google's cash flow just turned negative, driven largely by the company's staggering 190 billion dollar AI spending commitment this year.
Investors are watching closely. The company beat earnings estimates on Wednesday, but search revenue fell short of Wall Street expectations, raising fresh questions about whether the massive AI investments will ever pay off. Cloud revenue surged 82 percent, but that's not the same as winning the AI arms race.
The talent drain is real and costly. Noam Shazeer, a Gemini co-lead and influential researcher, jumped to OpenAI. John Jumper, a Nobel Prize winner in chemistry, left for Anthropic. Multiple departures have been tied to Google's April agreement with the Pentagon allowing the military to use its AI technology. Employees fear retaliation for speaking publicly about the deal, but anonymously they say it comes up constantly in exit interviews.
One DeepMind employee summed up the mood bluntly: "We're behind." Another described the moral conflict over military contracts as a "constant battle" that has led to "emotional burnout." Alex Turner, a former research scientist at the lab, resigned specifically over the Pentagon deal. He also criticized CEO Demis Hassabis for lacking the visible internal engagement that OpenAI's Sam Altman and Anthropic's Dario Amodei bring to their respective companies.
Google has pushed back hard. The company disputes the claim that morale problems are driving model delays or that the Pentagon contract is fueling departures. It notes that attrition rates for AI talent in the first half of this year are lower than last year, and that over 90 percent of candidates offered AI roles accept them.
There's one counterargument worth considering. The smaller models Google released this week may not be the fastest or smartest, but they're cheaper to run. In an era when enterprises are reeling from AI costs, there's real market demand for efficiency over raw power. Google is betting that delivering more value per dollar spent will matter as much as holding the performance crown.
The AI landscape has always been volatile. Leads shift fast. OpenAI appears primed to leapfrog Google with a new model it's briefing lawmakers about next week. Google held the high ground just a year ago. The company has the cash and resources to buy back into the race.
What it cannot survive is sustained talent loss and the erosion of employee confidence. Once researchers start heading for the exits and morale collapses, recruiting the next generation becomes exponentially harder.
Author James Rodriguez: "Google's problem isn't that competitors landed a punch, it's that the company's own team has lost faith."
Comments