Google’s next flagship AI model is reportedly running behind schedule, and the timing could hardly be more awkward. Gemini 3.5 Pro was expected to be Google’s strongest answer to the latest models from OpenAI, Anthropic, xAI, Meta and China’s fast-moving AI labs. Instead, reports say the model has been delayed by months as Google tries to improve its performance, especially in coding.
The report, first carried by Bloomberg and picked up by several market outlets, says Google has been working to improve Gemini 3.5 Pro after internal results fell short of expectations. Alphabet shares fell more than 4 percent after the report, showing how tightly investor confidence in Big Tech is now tied to whether a company looks competitive at the frontier of AI.
This is not only about one model missing a timeline. Gemini sits inside Search, Android, Workspace, Cloud, developer tools and Google’s broader AI strategy. When a flagship Gemini release appears to slip, it raises questions about product momentum across the whole Google ecosystem.
Coding has become one of the most important battlegrounds in AI because it is both measurable and commercially valuable. A model that can write, debug, refactor and understand large codebases can save companies time, power developer agents and justify premium enterprise pricing. That is why Anthropic, OpenAI, Google and others now treat coding capability as a public proof point.
If Gemini 3.5 Pro is struggling to meet Google’s internal coding goals, the issue goes beyond demos. Developer trust is hard to win back once engineers decide another model is more reliable. This is especially important as AI coding agents become central to enterprise workflows and software teams start standardising around specific model providers.
The pressure is even sharper because Google is shipping other AI products quickly. Its move to turn NotebookLM into Gemini Notebook with a secure cloud computer shows how aggressively the company wants Gemini inside work tools. But those products need the underlying model layer to keep improving.
Google has more AI surfaces than almost any rival. It must defend Search from answer engines, Android from rival assistants, Workspace from Microsoft and OpenAI, Cloud from Amazon and Microsoft, YouTube from creator AI tools, and developer mindshare from Anthropic and xAI. That breadth is an advantage, but it also raises the cost of falling behind in any one area.
A recent European Union decision requiring Google to open Android and some Search data to rival AI services adds another layer of pressure. The fight over Android access for rival AI assistants means Gemini cannot rely only on default placement. It has to be good enough that users and developers choose it.
The company still has enormous strengths: talent, infrastructure, DeepMind research, distribution, cloud customers and the ability to put AI directly into products used by billions. But in the current AI race, perception moves quickly. A delayed flagship model can make the market wonder whether the company is executing fast enough.
The Gemini delay story also says something broader about the AI boom. Bigger models are not automatically better on schedule. Frontier development is messy, expensive and uncertain. Coding, reasoning, tool use, safety, latency and cost all have to improve together, and a model can look strong in one area while missing expectations in another.
That is why investors are becoming more demanding. Big Tech companies are spending heavily on data centres, chips and talent. The market now wants proof that the spending turns into better products, stronger revenue and durable competitive advantage.
Google will almost certainly recover from one delayed release if Gemini 3.5 Pro eventually ships strong. But the episode shows how narrow the margin has become. In AI, being early is useful, being powerful is essential, and being perceived as slow can now move a company’s stock.