Morgan Stanley analysts forecast a significant boost for Chinese artificial intelligence stocks, anticipating inflows of over $1 billion into the underperforming Hang Seng Tech Index. This surge is attributed to the upcoming inclusion of two prominent Chinese AI companies, Knowledge Atlas Technology and MiniMax, into the index on June 8th. Despite a year-to-date tumble of over 11% for the index, these AI players have seen substantial growth since their January IPOs.
Knowledge Atlas Technology, operating Zhipu AI, and MiniMax, known for its generative AI models, are expected to drive between $1.25 billion and $1.75 billion in passive fund flows. Morgan Stanley has raised price targets for both, reflecting their strong performance and potential. Zhipu AI's models are noted for their coding prowess, while MiniMax offers a broader range of AI capabilities, from text to audio generation, and is a cost-effective alternative to U.S. counterparts.
The report highlights a trend of increasing costs for accessing Chinese AI models, which have risen to 17% of U.S. model costs, up from 5% a year ago. Morgan Stanley projects that these leading Chinese AI models could each generate at least $1 billion in revenue this year, potentially doubling that figure next year. Knowledge Atlas and MiniMax are pioneers as publicly listed Chinese AI model companies, with competitors like Moonshot and StepFun remaining private.
"We believe AI and [large language model] names will become a much bigger driver of Hong Kong equity markets, reshaping index composition, performance, liquidity, and fund flows," the analysts stated. They also noted strong regulatory support, with tech comprising 40% of Hong Kong IPO fundraising and 43% of the pipeline year-to-date, reinforcing AI's durable impact on the market. Even the index's largest constituents, Tencent and Alibaba, have experienced double-digit declines this year, though Alibaba is identified by Morgan Stanley as a top pick due to its AI integration across its business, from cloud computing to AI models.
