Odaily Planet Daily reports that after the release of Kimi K3, the U.S. semiconductor sector declined amid market concerns over a "DeepSeek Moment 2.0." However, UBS, Nomura, Bank of America Securities, and Citigroup all believe that Kimi K3 has not diminished AI computing demand; rather, it may further drive the expansion of AI infrastructure.
The institution stated that Kimi K3 features 2.8 trillion parameters, a 1 million token context window, and supports always-on inference, native multimodality, and MoE architecture. Its large-scale parameters and long-context capabilities will increase KV cache occupancy and elevate demand for HBM, server DDR5, enterprise-grade SSDs, cloud infrastructure, and high-speed interconnects.
Citigroup refers to this trend as "another Jevons paradox," where increased model efficiency and lower costs may lead to more use cases and greater token consumption. UBS notes that open-source models typically require more memory and storage due to longer context windows; Citigroup believes that large-scale deployment of Kimi K3 may require super-nodes composed of more than 64 GPUs. Nomura argues that global competition in large models will continue to drive investment from leading labs and cloud platforms, benefiting the AI infrastructure supply chain.
However, Bank of America Securities cautions that if the rate of improvement in model efficiency continues to outpace workload growth, and actual usage does not expand accordingly, AI infrastructure may still face a potential downturn.