Zhipu AI and MiniMax's double-digit drop signals the end of the narrative era โ and the beginning of a brutal fundamentals test.
The Hook: A 11% Haircut in a Single Session
The data is unambiguous. Zhipu AI and MiniMax, two of China's most prominent AI large-model startups, saw their Hong Kong-listed shares plunge more than 11% in a single trading session. This wasn't a subtle drift. This was a coordinated repricing event.
When a stock drops double digits in one day, the market isn't debating fundamentals. It's executing an exit. The question every serious investor should be asking is not "why did it drop" but "what was the market previously pricing that it no longer believes?"
Over the past 12 months, I've watched Hong Kong's AI basket transition from momentum-driven speculation to something far more unforgiving. The shift isn't subtle. It's structural.
Context: Two Unicorns, One Unforgiving Market
Zhipu AI, backed by Tsinghua University's intellectual pedigree, operates the GLM series of large language models. Its commercialization strategy centers on B2B API calls, private deployments, and government-enterprise partnerships. MiniMax, by contrast, bet on consumer-facing products โ Talkie and Hailuo AI โ built around subscription and advertising revenue.
Both companies represent the second tier of China's AI "Big Four" (the others being Moonshot AI and Baichuan Intelligent). Both chose Hong Kong as their listing venue. Both are now facing the same uncomfortable reality: Hong Kong investors don't extend the same valuation grace period that U.S. markets give to unprofitable AI companies.
The contrast with the U.S. market is instructive. OpenAI's revenue trajectory has been rewarded with increasingly aggressive valuations. Meanwhile, Hong Kong-listed AI names face a structural skepticism rooted in years of disappointment โ SenseTime, the city's first AI flagship, has seen its market cap erode over 70% from its 2021 peak.
This isn't a company problem. It's a venue problem.
Core: The Valuation Paradigm Shift
Here's what the price action is actually telling us.
First, the primary market overshot. Between 2023 and 2024, AI model companies raised capital at valuations based on narrative โ "technology leadership" and "total addressable market" projections that had no basis in audited financials. Zhipu's valuation reportedly reached ยฅ20 billion in 2024. That was a story multiple, not an earnings multiple.
Second, the secondary market is now enforcing a different standard. Investors are demanding proof of revenue growth, gross margins, and customer retention. When a company cannot provide that proof, the market doesn't wait. It reprices immediately.
Third, the SPAC problem. While the specific listing mechanism for Zhipu and MiniMax requires confirmation, SPAC-listed technology companies historically face a 50%+ average decline within 6-12 months of listing. The structure itself creates valuation air pockets โ early investors seek liquidity, lock-up periods expire, and the market absorbs the supply shock.
I've seen this pattern before. In my 2022 post-mortem of the Terra collapse, I noted that incentive structures โ not technology โ determine survival. The same logic applies here. When a company's incentive structure is built around "raising at the next round's valuation" rather than "generating sustainable cash flow," the correction is not a matter of if, but when.
The signal for the broader market is clear: AI valuations in China's primary market have entered a systematic de-rating cycle. Every subsequent fundraising round for unlisted AI companies โ Moonshot, Baichuan, and others โ will face more demanding terms.
Contrarian: The Retail vs. Smart Money Divergence
The mainstream interpretation of this decline is straightforward: "AI stocks are falling because AI is overhyped." That's lazy analysis.
The more accurate read involves understanding who's selling and why. The sharp, concentrated nature of the drop suggests institutional positioning rather than retail panic. Retail investors typically sell gradually, driven by sentiment. Institutions execute in size, driven by mandates and risk limits.
Consider the possibility that this drop is not a rejection of AI as a technology, but a rejection of specific valuation assumptions for specific companies at a specific venue. The technology itself remains transformative. The business models remain unproven. These are different statements.
There's also the structural angle. Hong Kong's AI narrative has historically followed a predictable pattern: hype cycle, valuation peak, correction, stagnation. SenseTime set the template. Horizon Robotics, listed in 2024, followed a similar trajectory. The market's AI enthusiasm in 2023-2024 has now entered its de-leveraging phase.
The hidden risk beneath the surface: liquidity. Hong Kong's equity market liquidity for small-to-mid-cap tech names is notoriously thin. When institutional investors decide to exit, the lack of buy-side depth amplifies the decline. Liquidity dries up faster than hope.
Takeaway: The New Rules of Engagement
For investors watching this space, three signals warrant attention over the next 3-12 months:
First, monitor trading volume. If the decline continues on increasing volume, the correction hasn't finished. If volume contracts and prices stabilize, a base may be forming.
Second, track the financial disclosures. Zhipu and MiniMax's next quarterly reports โ specifically revenue growth, gross margin trajectory, and cash runway โ will determine whether this is a valuation reset or a structural decline.
Third, watch the primary market ripple. If unlisted AI companies begin delaying funding rounds or accepting down-rounds, the systemic repricing is confirmed.
The era of story-driven valuation for Chinese AI is over. What replaces it will be uglier, more demanding, and ultimately healthier. Companies that can demonstrate real adoption, real revenue, and real retention will survive. Those that cannot will follow the path of every other narrative-driven asset in market history.
The question isn't whether AI is overvalued. The question is which companies deserve to trade at a premium โ and which were never worth the price in the first place.
The market just delivered its verdict on the latter. Now we wait to see which companies prove the market wrong.
