DeepSeek blew up markets year ago. Why hasn't it done so since?

• DeepSeek's January R1 model challenged assumptions about AI development costs and China's competitive position, directly threatening semiconductor and hyperscaler revenue narratives with its efficient, low-cost approach
• Limited access to advanced computing power due to U.S. chip export controls has constrained DeepSeek's ability to release new frontier models, with its R2 model delayed from May over difficulties training on Huawei chips
• Western AI labs responded with major releases including OpenAI's GPT-5, Anthropic's Claude Opus 4.5, and Google's Gemini 3, easing fears of sudden commoditization while AI infrastructure spending accelerated in 2025

Summary: DeepSeek's Diminished Market Impact

Key Event: In January 2025, Chinese AI startup DeepSeek triggered a major market selloff when it released its R1 reasoning model, causing Nvidia and Broadcom to plunge 17% each, and Microsoft to drop 7% in a single day. Nvidia lost $600 billion in market capitalization.

Why the Initial Shock: DeepSeek's V3 and R1 models claimed to match leading AI systems from OpenAI and Google while using less powerful chips and achieving training costs at a fraction of competitors' expenses. This challenged assumptions about U.S. AI dominance and raised concerns about reduced demand for AI infrastructure, directly threatening semiconductor and hyperscaler companies.

Subsequent Releases: Since January, DeepSeek released seven model updates, none generating comparable market reactions. Analysts attribute this to several factors:

• Incremental vs. Revolutionary: Recent releases were updates to existing V3 and R1 models rather than breakthrough new models
• Market Reassurance: U.S. companies demonstrated continued leadership with OpenAI's GPT-5 (August), Anthropic's Claude Opus 4.5, and Google's Gemini 3 (November)
• No Spending Slowdown: AI infrastructure spending accelerated through 2025, disproving fears of reduced demand

DeepSeek's Constraints: The company faces significant compute limitations due to U.S. chip export restrictions. Its R2 model, originally planned for May, was delayed due to challenges training on domestic Huawei chips. DeepSeek acknowledged in research papers its compute resource limitations compared to frontier models.

Market Recovery: Affected companies fully recovered, with Nvidia becoming the first to reach a $4 trillion valuation in October.

Future Outlook: Wedbush analyst Dan Ives warns more "DeepSeek moments" are likely, suggesting potential future market disruptions.