Google pivots to cloud compute as top AI leaders exit
- Google lost key leaders Jeff Dean and Demis Hassabis amid delayed frontier model releases.
- Alphabet is directing $200 billion toward cloud infrastructure rather than speculative frontier models.
- Internal research talent continues fleeing corporate bureaucracy to launch independent startups or join competitors.
Google stopped trying to win the frontier AI war.
The retreat became explicit when DeepMind CEO Demis Hassabis stepped aside to become Alphabet's chairman and chief scientist, handing daily operations to Koray Kavukcuoglu. Simultaneously, chief scientist Jeff Dean resigned after 27 years to launch an independent automated research startup called Discovery Loop. The executive shakeup knocked 4% off Alphabet's stock, signaling Wall Street's anxiety over a leaderless AI division.
The double departure caps months of accelerating brain drain across Google's research labs. Last month, AlphaFold co-creator John Jumper left for Anthropic, following Gemini technical lead Noam Shazeer's jump to OpenAI. As commercial competitors captured lucrative enterprise workflows with coding agents, DeepMind struggled to ship Gemini 3.5 Pro on schedule, leaving internal teams disheartened.
Corporate politics worsened the operational drag inside DeepMind. Former researcher Susan Zhang reported an environment where timid managers avoided hard choices, forcing subordinate teams into political squabbles over project directions. Meanwhile, the lab prioritized academic prestige projects over the commercial reasoning models enterprise customers demanded.
The institutional decay stems directly from corporate capital strategy. On All-In, David Friedberg argued that Google's executive exodus reflects a rational financial pivot rather than simple low morale. Alphabet committed $200 billion to infrastructure CapEx this year, chasing predictable tax write-offs and guaranteed cloud rental yields instead of spending tens of billions on speculative frontier models.
Selling raw compute offers high returns with minimal downside while open-source alternatives narrow the capability gap. As Brad Gerstner highlighted on All-In, Google Cloud now faces an internal conflict: selling server capacity to Anthropic generates steady cash, but starves Google's in-house builders of necessary GPUs. David Sacks noted that frontier AI has effectively consolidated into an OpenAI and Anthropic duopoly.
Google chose to become the utility powering its rivals.