By Jiahui Huang
MiniMax's shares fell sharply after the Chinese artificial-intelligence startup unveiled plans for a mainland listing and released a new model, following a strong run-up in the stock since going public early in the year.
The shares fell 16% to 708.00 Hong Kong dollars, equivalent to US$90.34, on Monday, the largest one-day loss since listing in January.
The Shanghai-based company said Sunday in a stock exchange filing that it plans to issue yuan-denominated shares and has engaged advisers for a proposed listing on Shanghai's STAR market, a Nasdaq-like board with a focus on technology companies.
MiniMax signed a listing-guidance agreement with Citic Securities and completed the required filing with the Shanghai bureau of the China Securities Regulatory Commission, according to a CSRC statement on Saturday.
The move underscores how China's leading AI companies are tapping domestic capital markets to fund their heavy computing and research spending in the country's rapidly evolving AI race.
The fundraising push comes as MiniMax's business has expanded rapidly. The startup's annual recurring revenue doubled between February and April and could top $1 billion by the end of the year, Jefferies analysts said in a note.
Investor enthusiasm for the company has grown. Hang Seng Indexes said last month that MiniMax would be added to the Hang Seng Tech Index, effective June 8.
MiniMax's move follows a similar path taken by rival AI startup Knowledge Atlas Technology, known as Zhipu AI, which in February hired Guotai Haitong Securities and China International Capital Corp. as advisers for a proposed STAR Market listing.
Investors are increasingly channeling capital into China's AI sector, betting that advances by domestic developers could narrow the technology gap with U.S. rivals while benefiting from significantly lower deployment costs.
Separately, MiniMax on Monday released its highly anticipated M3 model. The company claims the model has achieved frontier-level performance on specialized tasks such as coding and agentic work. The launch highlights MiniMax's efforts to tackle one of the industry's biggest challenges--high computing costs associated with running increasingly sophisticated AI systems.
Citi analysts said in a note that M3's proprietary architecture allows it to handle much longer contexts more efficiently, potentially lowering deployment costs for enterprise users.
While some investors may have hoped for a more significant breakthrough in reasoning capabilities, Citi said the new model could help shift competition in the AI sector toward cost efficiency, as companies seek ways to deploy increasingly powerful systems at lower cost.
Write to Jiahui Huang at jiahui.huang@wsj.com