On March 20, the global AI large model sector witnessed a major breakthrough: Xiaomi’s MiMo-V2-Pro large model ranked first on the weekly list of the internationally authoritative OpenRouter platform, achieving a performance leap with its trillion-parameter Mixture of Experts (MoE) architecture, marking that domestic large models have officially joined the world’s top tier. It is reported that the model leverages Xiaomi’s full-scenario ecosystem advantages, outperforming mainstream overseas models in core indicators such as multi-turn dialogue, logical reasoning and multimodal understanding. Xiaomi has invested over 16 billion yuan in the AI field to date, continuously deepening R&D to consolidate core competitiveness.
The AI computing power competition among global tech giants is also heating up simultaneously. Meta signed a five-year strategic cooperation deal worth 60 billion US dollars with AMD, focusing on AI computing power deployment, with the first phase of computing power to be delivered in the second half of 2026 to fully support large model training and commercialization. Huawei also launched a full-stack storage solution for the AI era, addressing the computing power bottleneck of large models and improving the domestic AI computing power ecosystem. At the same time, Norway’s sovereign wealth fund issued a warning, pointing to bubble risks in the current AI sector. If market sentiment cools, related assets may face a downside of around 35%, sparking rational reflection on AI valuations.
Insiders pointed out that the AI industry has shifted from technological competition to commercialization, and enterprises with computing power advantages and scenario landing capabilities hold greater long-term value. In the short term, the AI sector faces intensified volatility and bubble risks that require vigilance; in the medium to long term, AI remains the core main line of the technology industry, with investment opportunities hidden in the three links of computing power, algorithms and applications. Investors are advised to focus on leading enterprises with strong performance delivery capabilities, avoid pure speculative thematic targets, and rationally allocate in the AI industry track.
