Alphabet's stock falls as Gemini delays suggest Google is struggling to keep up in the AI race

By William Gavin and Christine Ji

Meanwhile, several rivals have launched their latest and greatest AI models

Google engineers and AI researchers are reportedly frustrated by the company's delays.

Alphabet shares fell more than 4% on Thursday, reflecting concerns that the Google parent company is falling further behind in the artificial-intelligence race.

Google's Gemini 3.5 Pro, its most powerful AI model to date, is months behind schedule because of the company's work to try and boost its performance, according to Bloomberg. In early June, CEO Sundar Pichai said the model was expected to be launched later that month, but it still has not yet been released.

The delay has frustrated Google's engineers and AI researchers, Bloomberg reported. Some of them are said to be concerned that the company risks ceding ground to rivals Anthropic and OpenAI, among others. Part of the issue for Google is that the company needs to integrate AI across a massive portfolio, according to the report.

Google in June also updated its training data for Gemini in an attempt to improve its code-writing capabilities, but the results have been disappointing, the Bloomberg report said.

Alphabet shares (GOOG) (GOOGL) fell more than 4% on Thursday, as the delay struck a blow to the company's reputation as a full-stack AI winner. Driven by Gemini's aggressive growth in 2025, Google had shed its reputation as an AI laggard; investors lauded the its comprehensive AI offerings, from custom tensor processing units to its various distribution surfaces. The company looked positioned to deploy its AI capabilities across its entire product portfolio.

Thursday's report built on earlier worries of Gemini falling behind, as the lab experienced several high-profile departures last month.

Alphabet's stock drop also contributed to a broader selloff across the tech sector. Fears of AI overspending have weighed heavily on tech names in recent weeks, as the market looks for a tangible return on investment. Investors are questioning the durability of a business model where the leading AI model shifts every few months, Bryan Wong, portfolio manager at Osterweis Capital Management, told MarketWatch.

"We're shipping quickly across a wide range of models while keeping them highly cost-effective for customers," a Google spokesperson said in a statement. "We're currently testing 3.5 Pro, an upgraded Flash model, and other models with partners, and we're productively engaged with the U.S. government."

Anthropic was temporarily forced to disable access to its latest models, Fable 5 and Mythos 5, last month after the U.S. government flagged them as a security threat. OpenAI in June voluntarily limited access to its GPT-5.6 family of new models - Sol, Terra and Luna - to "trusted partners" in an effort to appease the government. Those models were made available for wider use last week.

They're not the only rivals Google has to worry about. Meta Platforms (META) last week released Muse Spark 1.1, its first model to tout near-frontier agentic coding capabilities, while SpaceX (SPCX) launched Grok 4.5, its first model "specifically" trained for coding and agents.

And Beijing-based Moonshot AI on Thursday became the latest lab to publish a new model, its Kimi-K3, that it said was built for "long-horizon agentic coding." According to Arena.ai, which measures and compares the performance of leading AI models, the Kimi model has surpassed Anthropic's Fable 5 in some respects.

"This may be the single biggest release of the year, and marks the moment that [open-source software] Chinese models have surpassed U.S. models," said Arena CEO Anastasios Nikolas Angelopoulos in a post on X.

-William Gavin -Christine Ji

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

07-16-26 1732ET