The delay wasn't a single slip—it was a cascade of broken promises:
The model has now missed at least three self-imposed deadlines: the June I/O window, the June 30 specific date, and the July 17 target .
Internal dysfunction contributed alongside the technical failures. Bloomberg and subsequent coverage describe bureaucracy inside Google, with "multiple layers of stakeholders and overlapping AI coding initiatives" slowing decision-making . The decision to rebuild from scratch—rather than ship an imperfect version—reflects a crisis of internal confidence about what the DeepMind team could deliver on the original timeline .
Interestingly, the U.S. government was not a factor. A June 2 executive order from the Trump administration set up a voluntary pre-release review framework for frontier AI models . OpenAI's GPT-5.6 and Anthropic's Claude Fable 5 were both held up by this review process. Gemini 3.5 Pro was never subjected to government restrictions—it was delayed purely by its own performance issues. This created an ironic situation where Google was the only major frontier lab free to ship without government interference, but couldn't ship because the model wasn't ready .
Internal sources directly express concern that while Google struggles with coding benchmarks, OpenAI's GPT-5.6 and Anthropic's Claude Fable 5 have already reached users with strong performance . The full architectural rebuild was motivated in part by this pressure, as frontier AI competition shifted from incremental upgrades to major architectural leaps .
Adding to the pain, two of Google's most celebrated researchers departed for OpenAI and Anthropic during this period . A separate report noted that five top Google AI researchers left in a single week in late June .
Alphabet shares fell sharply on July 16, 2026, following the Bloomberg report:
Note: The share price decline was amplified by new EU regulatory actions announced the same day, so the Gemini delay was not the sole factor .