Google's AI Brain Drain, SpaceX's Huge Quarter, Airtable's 90% Collapse, US Data Fuels China AI

Google's AI Shake-Up and the Future of AI Models

Google recently experienced a significant shake-up in its AI division, with Demis Hassabis moving to Chair of DeepMind and Chief Scientist at Google. Reports suggest this move could be interpreted as either a promotion or a demotion, with some sources indicating low morale within Google's AI teams. Several top researchers, including Gemini's co-lead, have left Google for competing AI labs, and Jeff Dean, a legendary AI engineer, is departing to start a new company called Discovery Loop, focused on deep scientific breakthroughs in AI. This news led to a 4% drop in Google's stock, representing a $200 billion loss in market cap.

David Friedberg questioned whether this exodus is a sign of creative destruction or if the allure of unlimited capital and opportunity in AI startups is simply too strong for Google's established talent. He suggested that Google's commitment to deploying $200 billion in CapEx this year for AI infrastructure and data center build-outs, which offers significant tax advantages and high profit returns, might be diverting capital away from cutting-edge model development. Friedberg argued that while building advanced frontier models requires substantial capital, the question is whether these models can deliver the same profit returns as investing in compute infrastructure, especially when open-source models are rapidly catching up.

Brad Gerstner agreed with Friedberg, highlighting that a similar trend is occurring at Microsoft, where they are seeing over a 30% return on invested capital in "tokens as a service" (infrastructure). He emphasized that while infrastructure businesses are highly profitable, scientists focused on super-intelligence and frontier models face a "channel conflict" within Google. Google Cloud wants to rent out compute to external companies like Anthropic, while internal teams building frontier models also need that compute to compete. Gerstner believes this conflict is being resolved in favor of Google becoming more of an infrastructure company.

Jason Calacanis, however, believes Google will remain the number one AI company due to its massive consumer reach. He pointed out that Google has five products with over 3 billion monthly active users each (Android, Search, Gmail, Chrome, YouTube), all of which are becoming AI-first products. He also noted that Gemini had over 950 million monthly active users in Q2, tripling year-over-year. Calacanis argued that for 95% of tasks, open-source and non-frontier models are "good enough," making them a viable alternative to expensive, cutting-edge models. He believes Google will dominate consumer AI usage.