A Hong Kong listed Chinese AI model developer has shed $39 billion since March, leaving founder Yan Junjie down $9 billion, as a market read hands the lead to rival Zhipu and a lockup looms.
Yan Junjie has lost roughly $9 billion in paper wealth in three months as the Hong Kong-listed Chinese AI model developer MiniMax has shed about $39 billion in market value since its March peak. New models still ship on schedule. The defining question, and the one Wednesday's lockup expiration will stress-test, is whether the selloff reflects a genuine competitive shift to rival Zhipu or simply a crowded trade unraveling under its own weight.
MiniMax shares touched a record near HK$1,330 (about $170) in March, more than double the company's January Hong Kong IPO price. They have since plunged more than 70%, dragging the founder's net worth from a peak of roughly $12.6 billion to about $3.3 billion. The stock is still up modestly year-to-date, a reminder of how steep the round-trip has been.
The founder's $12.6 billion peak was largely paper, tied to the post-IPO surge; the descent to $3.3 billion is the same paper, marked down. Pre-IPO investors and early backers, who rode the same surge, are now the ones whose lockup expiration gives them the right to convert that paper into cash. The lockup print is the moment institutional paper becomes institutional cash, while the prices retail holders paid set the new reference.
The competitive story is murkier. MiniMax has rolled out updates with measured gains on coding benchmarks and agentic financial-analysis tasks. In the market's read, the Chinese AI model leadership baton has passed to Zhipu, the other high-profile Hong Kong-listed AI lab, according to Shanghai-based analyst Charlie Chai quoted in coverage of the rout. That reordering is harder to measure than a share-price chart, since model leadership is contested in benchmarks and product adoption that don't show up in tape.
Wednesday's lockup expiration is the more concrete catalyst. Pre-IPO investors and early backers become free to sell, expanding the float at exactly the moment supply is least welcome. The IPO raised about $619 million in January, and the post-IPO float is concentrated, so a small number of sellers can move the price disproportionately. The mechanics are well-rehearsed in Hong Kong listings, but they cut harder when the underlying thesis has already rolled over.
The supply event lands on a company constrained on the demand side, too. The Hong Kong stock exchange prospectus shows Alibaba holds roughly a 13.66% indirect stake in MiniMax, a relationship that doubles as a cloud-compute channel and a cap on how aggressively the smaller lab can source scarce AI training compute from third parties. For a company whose product is GPU-hours, that tradeoff sits at the center of the competitive question. The prospectus detail, not a third-party digest, frames Wednesday's print as a test of both float mechanics and structural compute access.
The read-through extends beyond MiniMax. Hong Kong has emerged as the listing venue of choice for Chinese AI model labs, with Zhipu and a queue of smaller developers following the same playbook of pricing access to dollar-denominated capital against a backdrop of US chip-export controls. A messy lockup print would lengthen the discount on subsequent Chinese AI listings; a clean one would compress it. The float event is a pricing test for the broader Chinese AI listing complex, not just one ticker.
The model's competitive position is harder to read on a weekly chart. A stable print would shift the dominant story to the model race, while a messy one would extend the overhang and reset the cost of capital for the next Chinese AI listing waiting in line.