Google Gemini 3.5 Pro delayed months, shares drop 5% on report

Alphabet's most powerful AI model is months behind schedule as internal benchmarks show it trailing Anthropic and OpenAI on coding tasks.

Google's Gemini 3.5 Pro, its flagship AI model, has fallen months behind schedule as engineers struggle to improve coding capabilities, according to a Bloomberg report that sent Alphabet shares down 5% on Thursday.

"The delay has been a source of frustration for Google engineers, AI researchers and managers, many of whom are concerned the company risks losing an edge in the market," 10 current and former employees told Bloomberg's Julia Love and Davey Alba.

The company has multiple layers of stakeholders involved in preparing models for release, working to weave AI across a vast product portfolio including search, maps and YouTube, which can cause delays, the people said. Rivals Anthropic and OpenAI have produced models that exceed Gemini's capabilities in key areas, according to the employees.

The setback threatens Google's position in a market where Alphabet has committed more than $80 billion in AI investments, including a landmark deal with Berkshire Hathaway. Alphabet shares closed down 5% on the news, erasing billions in market value.

The coding deficiency is particularly concerning because it strikes at the heart of enterprise AI adoption. Software developers represent one of the largest paying customer segments for large language models, and coding benchmarks such as HumanEval and SWE-bench have become standard measures of model capability. Google's inability to match competitors on these metrics could slow enterprise adoption of its cloud AI services, a business that competes directly with Microsoft-backed OpenAI and Amazon's Bedrock platform.

Internal Pressure Mounts as Competition Intensifies

The delay reflects a broader challenge inside Google: balancing the speed of AI development with the safety and integration requirements of a company whose products reach billions of users. While startups like OpenAI and Anthropic can ship models with fewer internal approvals, Google must ensure its AI works reliably across search, advertising, YouTube and cloud — each with its own engineering team, safety standards and product timeline.

Anthropic's Claude 4 and OpenAI's GPT-5 have both set new benchmarks in coding and reasoning over the past six months, according to published evaluation scores. Google's Gemini 2.0, the predecessor to the delayed 3.5 Pro, trails on several key metrics, putting pressure on the company to deliver a model that can reclaim the technical lead.

What's at Stake for Investors

Alphabet trades at about 22 times forward earnings, a discount to Microsoft's 30 times but a premium to many legacy tech peers. The Gemini delay raises questions about whether Google can monetize its AI investments as quickly as rivals. Microsoft has already integrated GPT-5 into its Azure OpenAI Service, GitHub Copilot and Microsoft 365 suite, generating measurable revenue from AI features. Google's ability to do the same depends on shipping competitive models.

The company has not disclosed a revised release date for Gemini 3.5 Pro. Analysts expect an update at Google's next major developer event, though the timeline remains uncertain.

This article is for informational purposes only and does not constitute investment advice.