← 返回 Avalaches

2026年2月22日,Bloomberg报导了美国与中国股市在 AI 相关行为上的明显分化:美国投资者因担忧颠覆而进行「AI scare trade」,抛售软体与财富管理类股票;而中国投资者主要把 AI 定价为渗透率与效率提升的叙事。包括 Charu Chanana 与 Gary Tan 在内的策略师认为,这一差距反映了市场结构差异,尤其是中国更为隔离的本土 AI 生态,外国大型语言模型进入受限,降低了对本地龙头的直接外部竞争压力,也使焦点从利润池被侵蚀转向增长获取。

最强涨幅集中在新上市的中国纯 AI 标的:自 2026年1月在香港上市以来,Zhipu 上涨 524%,MiniMax 上涨 488%,且两者仅在 2月 就都翻倍以上;其他 AI 相关上市也大涨,包括 Shanghai Biren Technology 较 1月2日 上市水位高出超过 80%,以及 Montage Technology 较 2月9日 起始水位高出超过 98%。产品与基准信号进一步强化了动能,包括 Zhipu 的 GLM-5 在 Artificial Analysis 登上全球开源第一;此外还有板块催化因素,如 ByteDance 新影片应用带动媒体股上涨,以及市场预期 DeepSeek 的后续发布持续支撑投机需求。

估值支撑也来自全球私募市场的参考点:据报导,OpenAI 正接近一轮超过 $100 billion 的融资,估值超过 $850 billion;Anthropic 则以 $380 billion 估值融资 $30 billion,这有助于为中国同业更高倍数提供合理性。卖方信心也在上升,Morgan Stanley、Jefferies 与 UBS 均对 MiniMax 给出等同买入观点;Morgan Stanley 预计其到 2027年 的营收约 $700 million(约 $0.7 billion),意味著 2 years 内最高可达 10x 增长。核心风险在于可持续性:若实际盈利增长落后于当前估值扩张,AI 龙头与更广泛市场都可能面临重定价,尤其是在投资者可能低估跨行业颠覆风险的情况下。

ee97d3fe24f4.png


5a153678d028.png



On February 22, 2026, Bloomberg reported a sharp divergence between US and China equity behavior around AI: US investors are running an "AI scare trade" by selling software and wealth-management names on disruption fears, while China investors are pricing AI primarily as a penetration and efficiency story. Strategists including Charu Chanana and Gary Tan argue this gap reflects structural market differences, especially China’s more insulated domestic AI ecosystem where foreign large language models have limited access, reducing direct external competitive pressure on local incumbents and shifting focus toward growth capture rather than profit-pool erosion.

The strongest gains concentrated in newly listed Chinese pure-play AI names: since January 2026 Hong Kong debuts, Zhipu rose 524% and MiniMax gained 488%, with both more than doubling in February alone; other AI-linked listings also surged, including Shanghai Biren Technology at over 80% above its January 2 listing level and Montage Technology at over 98% above its February 9 start. Momentum was reinforced by product and benchmark signals, including Zhipu’s GLM-5 taking the top global open-source spot on Artificial Analysis, plus sector catalysts such as ByteDance’s new video app rallying media shares, and expected releases from DeepSeek sustaining speculative demand.

Valuation support is also coming from global private-market reference points: OpenAI is reportedly nearing a funding round above $100 billion at a valuation above $850 billion, while Anthropic raised $30 billion at a $380 billion valuation, helping justify higher multiples for China peers. Sell-side conviction has risen, with Morgan Stanley, Jefferies, and UBS issuing buy-equivalent views on MiniMax; Morgan Stanley projects about $700 million (about $0.7 billion) revenue by 2027, implying up to 10x growth over 2 years. The core caveat is sustainability: if realized earnings growth lags current valuation expansion, AI leaders and the broader market could face repricing, especially as investors may be underestimating cross-sector disruption risks.
2026-02-23 (Monday) · 040de5bdefb53d2106ce4b53d5e62bd8036faa1f