Gemini 3.5 Pro flunked its own tests and Google hit pause — wiping 200 billion dollars in a day
Google delayed Gemini 3.5 Pro by months after the model fell short on coding and complex reasoning. Alphabet shares dropped about 4%, erasing roughly 200 billion dollars in market value.
Gemini 3.5 Pro was supposed to arrive in June — and Google had already pushed the launch to July promising fine-tuning. Now Google admits its most powerful AI model is months behind schedule — and Wall Street was not amused. Alphabet shares fell about 4% in a single day, erasing roughly 200 billion dollars in market value.
Why did Google delay Gemini 3.5 Pro?
Because the model is not living up to the company’s own promises. In internal testing, its ability to write, understand and debug code — exactly the ground where OpenAI and Meta have been flexing hardest — came in short of expectations. Google did try to rescue the timeline: in late June it retrained the model with more programming-heavy data. The results still disappointed, and the company went back into development, effectively restarting parts of the training process without committing to a new date. The model was unveiled in May at I/O, the annual conference whose announcements the company publishes on Google’s official blog.
What does this mean for the AI race?
That the technical lead has changed hands faster than Google would like to admit. The stumble lands in a market where investors increasingly demand proof over promises — the same logic that has pushed Apple to the top of the world’s most valuable companies on the back of concrete results.
For everyday users the takeaway is simpler: today’s Gemini stays where it is, and the promised leap slips past the summer — at best. In the AI race, announcing is easy; training is the expensive part.
By Oliver Grant
Image: Photographer: Lukasz Kobus (European Commission) / Wikimedia Commons (CC BY 4.0)
