Nvidia’s shares saw a significant decline during early U.S. trading hours, dropping more than 7% at one point to $169.55—a retreat of over 20% from last month’s all-time high. Key ecosystem partners including Super Micro Computer, Oracle, and CoreWeave also faced selling pressure, with CoreWeave falling 3.7%. This chain reaction is being interpreted by the market as a response to breakthroughs in Google’s TPU chip technology. Notably, Nvidia has long dominated the data center infrastructure space, but the emergence of TPU is reshaping the fundamental logic of AI computing power.
As Alphabet, Google’s parent company, surged past a $4 trillion market capitalization, attention around its in-house TPU chip continues to intensify. The Gemini 3 large language model, released last week and powered by TPU’s computing advantages, earned praise from industry authorities as “surpassing GPT” in performance benchmarks. This marks a shift in AI chip competition from pure hardware performance to a comprehensive battle over computing architectures and model training efficiency. Industry observers note that TPU’s heterogeneous computing architecture may offer superior energy efficiency for large model training.
Nvidia’s dominance in the AI chip sector is now facing unprecedented challenges. While its A100 and other products remain dominant in data centers, TPU has achieved performance breakthroughs in specific scenarios through customized architecture. This divergence in technological paths reflects a broader trend toward diversification in AI computing infrastructure—encompassing both Nvidia’s general-purpose GPU solutions and dedicated chips like TPU. Market analysts believe this competition will accelerate the iterative upgrade of AI infrastructure.
The AI industry is currently in a phase of explosive demand for computing power, with large model training consuming exponentially increasing resources. The arrival of TPU not only reshapes the chip supplier landscape but also pushes AI infrastructure toward heterogeneous computing. This technological transformation is redefining the AI industry ecosystem, prompting companies to reassess their computing power investment strategies and technology roadmap choices.