What happened
Employees at Google’s DeepMind told reporters that low morale helped slow the Gemini model rollout. Some left after a Pentagon contract and others complained of burnout and mixed priorities.
The company pushed smaller, cheaper models while its flagship Gemini 3.5 Pro slipped months behind schedule. Google says attrition and hiring stay healthy, but insiders describe the picture differently.
Who wins here
Executives hoping to show progress keep short-term gains from small models. Competing labs gain face time when Google misses big model milestones.
Investors and cloud customers win if Google turns costs into usable products fast. Workers lose leverage when management keeps shifting priorities and deals with the Pentagon raise trust problems.
How the play works
This is a staffing and priority play. Leadership sets priorities, then uses hiring, pay, and deals to chase them. If leaders push military contracts, some staff quit or protest, and focus splits.
That split slows big projects. Small model releases look like progress, but they don’t replace a delayed flagship model. The company’s big spending keeps pressure on quick wins.
Why it matters
If morale and contracts slow model progress, Google risks losing technical edge. Falling behind can mean worse products for people and higher costs for customers.
Military ties also change what engineers will work on. That can shift what Google builds for everyone. Public trust and how Google shares AI power are both at stake.
What to watch next
Watch hiring and attrition numbers for AI teams in public filings. Look for exec statements and team reorganizations tied to Gemini timelines.
Also watch how the Pentagon deal is managed. If exits and walkouts keep happening, expect more delays and public pushback.