Shares of Alphabet fell 4.4% on Thursday, closing at $354.46 and erasing approximately $200 billion in market capitalization after Bloomberg reported that Gemini 3.5 Pro, Google's flagship next-generation AI model, has fallen months behind its original release schedule. The setback stems from the model's failure to meet Google's internal performance benchmarks, with coding capability identified as the primary shortfall. Engineers, researchers, and managers inside the company are reportedly frustrated, with some concerned Google is at risk of ceding its competitive position in the high-stakes large language model race to OpenAI and Anthropic.
The original launch window for Gemini 3.5 Pro was June 2026. Google had announced the model at its annual I/O developer conference in May, saying at the time that it was being used internally and would be available externally the following month. An updated training run incorporating new data was launched in late June in an attempt to improve coding performance, but sources told Bloomberg the early results were disappointing. Some engineers have floated the idea of scrapping earlier base models entirely and rebuilding from scratch, though no final decision has been made on that approach. Google's problems extend beyond coding, with token efficiency, agentic task handling, and long-horizon reasoning all flagged as areas needing improvement.
The timing of the delay is particularly damaging from a competitive standpoint. OpenAI has shipped the GPT-5.6 Sol, Terra, and Luna family of models, which has been described as substantially more token-efficient on agentic coding tasks. Meta has released its latest Llama generation. Anthropic, meanwhile, has launched Claude Fable 5 and Mythos 5, the latter following security clearances. Each of those releases has taken market share in developer and enterprise deployments at the precise moment Google is unable to respond with its most capable model.
An Alphabet spokesperson told CNBC in a statement that the company is "shipping quickly across a wide range of models while keeping them highly cost-effective for customers," and added that Gemini 3.5 Pro is currently being tested with partners. Google also reiterated that it is "productively engaged" with the US government. Analysts noted the statement stopped short of confirming any revised timeline, which the market interpreted as a signal that a near-term launch is not imminent.
The stock's decline came despite Alphabet CFO Anat Ashkenazi raising the company's 2026 capital expenditure guidance to a range of $180 billion to $190 billion as recently as the most recent earnings call, underscoring the scale of investment the company is making in AI infrastructure even as its flagship model shipment slips. For context, Thursday's single-day market cap decline was approximately 5.6 times the $35.7 billion in capital expenditure the company reported in its most recent quarter.
Alphabet is scheduled to report second-quarter 2026 results on Wednesday, July 22, seven trading days from now. Investors will be closely watching the Cloud segment for evidence that AI infrastructure revenue is keeping pace with spending, and management will face direct questions about the revised Gemini 3.5 Pro timeline. Some analysts see the pre-earnings selloff as an opportunity, particularly given that Warren Buffett's Berkshire Hathaway has been reported to be building a position in GOOGL shares, but others caution that the technical picture has deteriorated meaningfully, with the stock now below both its 20-day and 50-day exponential moving averages.