Google's flagship AI model is months behind on coding performance, and investors are not waiting for a fix
Alphabet shares fell more than 4% on Thursday after Google pushed back the release of Gemini 3.5 Pro by several months, wiping roughly $200 billion in market capitalization in a single session. Google engineers need more time before the model can meet internal benchmarks against rivals on coding tasks, according to multiple reports.
Investors have priced Alphabet's AI ambitions into its valuation for years, and a delay to its most capable model lands as OpenAI and Anthropic compete for the same enterprise and developer customers Google is trying to hold.
Coding capability has become a primary battleground in frontier AI. Developers choose models based on how well they perform on programming tasks, and enterprise customers increasingly evaluate AI tools by how well they handle software automation. A model that underperforms on code misses the use case driving a large share of commercial AI adoption right now.
Google has not disputed the delay. The company has said it is working with the US government on model testing and broader AI safety frameworks, a statement that addresses regulatory posture rather than the timeline.
Get the latest model rankings, product launches, and evaluation insights delivered to your inbox.
The delay arrives alongside a separate pressure. The EU has ordered Google to open Search and Android data to competitors, compounding investor anxiety even though it operates on a different timeline and through different mechanisms than the Gemini product issue.
The AI buildout that major labs are running, including compute, talent, and infrastructure, is expensive at a scale that makes delays materially costly. Every month Gemini 3.5 Pro stays out of market is a month competitors can strengthen their foothold with developers and enterprise buyers.
Delayed model releases are not unusual in this industry, and Google has shipped capable models before under competitive pressure. The open question is whether the coding gap is a narrow, solvable problem or a sign of something structural in how Gemini 3.5 Pro was designed. The available reporting does not resolve that, and Google has not offered a revised launch date.
The next checkpoint will be an official announcement from Google on a new release timeline, or a competitor release that forces Google's hand before its engineers are ready.