The drop was larger than Alphabet's entire projected 2026 capital spending budget of $180-190 billion, and it happened in hours on the back of a single report. Bloomberg, citing current and former employees, said Gemini 3.5 Pro's coding ability — the area where rivals OpenAI and Anthropic have pulled ahead — fell short of internal targets, forcing Google to retrain on more programming data late last month rather than ship on schedule.
The timeline matters: Google showed the model internally at I/O in May, told the market it would roll out in June, and June came and went with no release. Some coverage ties the slip to a wave of senior DeepMind departures — including Transformer co-inventor Noam Shazeer and Nobel laureate John Jumper moving to OpenAI and Anthropic — though the sequencing suggests the delay and the talent exits are compounding rather than a single cause. Google has not published a model card, pricing, or benchmark scores; every leaked spec, including a rumored 2-million-token context window, remains unconfirmed.
Wall Street's reaction is the real story: a single report about an unreleased model moved one of the world's largest companies by hundreds of billions of dollars in a single session, underlining how tightly investors are now pricing frontier-model execution risk into Big Tech valuations.
