The honeymoon period for China’s artificial intelligence 'tigers' on the public markets has come to a bruising end. MiniMax, one of the nation’s most prominent AI unicorns, saw its market value crater by 27% in just two days following the expiry of its initial lock-up period. The sell-off wiped out over 200 billion HKD in market capitalization, signaling a harsh reality check for investors who previously drove the company’s valuation to dizzying heights.
At the heart of the rout is a massive liquidity shock. On July 9, MiniMax saw roughly 146 million shares—representing 63% of its total equity—become eligible for trading, effectively expanding its public float tenfold overnight. While strategic backers like Alibaba and miHoYo pledged to hold their positions, a silent cohort of financial heavyweights, including Hillhouse and Tencent, left the market guessing, triggering a stampede toward the exits.
This collapse stands in stark contrast to the performance of Zhipu AI, MiniMax's primary domestic rival. Despite facing its own lock-up expiry, Zhipu’s stock surged as state-backed cornerstone investors signaled long-term commitment. The diverging fortunes of these two giants have widened their valuation gap to a staggering 800 billion HKD, reflecting a market that is becoming increasingly discerning about shareholder structure and corporate governance.
MiniMax’s troubles are not merely technical; they are also strategic. The release of its M3 model in June was marred by a botched pricing transition that saw costs for some developers spike by over 250% without prior warning. This move alienated the developer community and prompted major investment banks, including Citigroup and JPMorgan, to slash their price targets, citing concerns over user retention and the company’s ability to defend its pricing power.
In a bid to reclaim the narrative, MiniMax is doubling down on a high-stakes technical gamble. The company is reportedly developing a new flagship model, codenamed M3 Pro, which boasts an unprecedented 2.7 trillion parameters. This would make it the largest large language model (LLM) in China to date, as the company seeks to use brute-force scaling and a proprietary 'MiniMax Sparse Attention' architecture to leapfrog competitors.
However, the financial fundamentals remain precarious. In 2025, MiniMax reported a net loss of $1.87 billion against a modest revenue of $79 million, highlighting the astronomical burn rate required to sustain the current AI arms race. While Goldman Sachs remains optimistic about the company's path to a $1 billion annual recurring revenue goal by 2026, the immediate future depends on whether M3 Pro can deliver tangible utility or if it will simply be another expensive experiment in a tightening capital market.