Hong Kong-listed shares of Chinese AI giants Zhipu AI and MiniMax fell more than 11% in a single session, and the market is only beginning to price the gap between private hype and public reality.
The selloff wasn't isolated. Sector-wide pressure hit AI concept stocks across the exchange. Zhipu and MiniMax, both considered part of China's "Four Little Dragons" of large language models, bore the brunt. The question no one wants to answer directly: how much of their valuation was ever real?
I've spent the last seven years auditing the edges of projects that promised the world and delivered a hashed mess. I've seen order matching engines with integer overflows that could drain liquidity pools. I've traced $8 billion in missing funds through unrelated wallet addresses. I've analyzed Ponzi structures hiding behind 19% APY yield farms. The market believes what it wants to believe, but the code—and the balance sheet—does not lie.
Let's pull the ledger on this one.
The Context: A Market That Tolerates Narrative, Not Burn Rates
Zhipu AI, with its GLM series of models, represents China's elite academic lineage. MiniMax, with consumer products like Talkie and Hailuo AI, represents the C-end battle for adoption. Both chose Hong Kong over the U.S. capital markets. That's not a neutral choice. It's a signal.
Chinese AI companies face increasing scrutiny in the U.S. due to audit and geopolitical constraints. Hong Kong becomes the only viable escape hatch. But a listing venue change does not change the underlying economics. The two companies are still at the "high investment, low return" stage of AI development. And Hong Kong investors, unlike their American counterparts, are historically less tolerant of story-driven valuations.
The market dynamics are clear. The sector has been riding the AI narrative since late 2023, but by 2025, the patience has thinned. The recent drop signals that investors are demanding a different kind of proof: not just technological leadership, but revenue growth, gross margins, and customer retention.
Core: The Forensic Analysis of What Just Happened
Let's break down the mechanics of this decline from a systems perspective.
The SPAC Valuation Trap
Zhipu and MiniMax likely went public via a SPAC structure. This is the fastest route to public markets for Chinese AI companies, but the costs are extreme. SPAC issuances often carry embedded valuations that outpace reality, because the promoters and the target company have a shared interest in announcing a high value.
Historical data shows SPAC-listed companies average a drop of more than 50% within six to twelve months post-merger. The SPAC valuation process is a negotiated number, not a market-clearing one. When the public market sees the actual financials, the mark-to-market adjustment is brutal.
I've seen this pattern before in crypto. Projects list with a high initial circulating supply and a "team allocation" that gets sold off as soon as the lock-up expires. The market absorbs the narrative, but the data behind it is thin. Once the floor drops, the protocol and its token reveal their true weight: nothing.
The same pattern is repeating in AI stocks. The market is taking the narrative out to the back alley.
The Valuation Anchoring Problem
The previous market value of Zhipu was around 20 billion RMB at the private stage. That's a huge number for a company that likely has limited disclosed revenue. The public market, which sees the actual burn rate and the lack of clear customer stickiness, is now telling the private market it was wrong.
This is a classic case of primary market (VC/PE) pricing being anchored to the secondary market's expectations.
When a company goes public, the information asymmetry closes. The public market has access to the same materials but with an added layer of caution. The result is a "correction" where the price drops from the IPO level to a level that reflects actual risk-adjusted cash flows.
The Hong Kong Market's Structural Dislike for AI
Hong Kong investors are not like their Nasdaq counterparts. They have a deep memory of the 1997-2000 dot-com crash. They also have more exposure to Chinese tech regulation cycles. The negative sentiment is not just about AI fundamentals; it's about the perceived fragility of Chinese tech companies under government scrutiny.
Compare this to SenseTime, the first AI stock in Hong Kong. Since its listing, it has lost over 70% of its market value. The market is already baked with the sense that AI unicorns will face a significant amount of volatility. This is not a speculative asset class anymore; it's a way to create liquidity for early investors before the bubble completely deflates.
The Liquidity Drain
The financial statement of the market is simple: liquidity is a premium. The AI sector requires a constant inflow of capital to sustain its growth. When the market sees a stock drop by more than 10% in one day, the signal to other market participants is clear: "The tide is going out." This forces more selling, especially from leveraged funds and those who use AI as a hedge against inflation.
The drop is not a "buy the dip" moment. It's a systematic risk event.
The Chain Reaction
This drop is not just about Zhipu and MiniMax. It's a signal to the entire Chinese AI ecosystem. The "four little dragons" of AI (Zhipu, MiniMax, Moonshot AI, and Baichuan Intelligence) are all facing the same challenge. If the second-tier AI players are being marked down by the market, the primary market will be forced to follow. This will make it harder for the remaining players to raise capital. It will also affect the upstream GPU cloud service providers and downstream application developers.
The market is starting to remember that complexity is often a disguise for theft. Or in this case, a disguise for the lack of sustainable value.
The Contrarian Angle: What the Bulls Got Right
I'm not here to declare AI dead. The market is not always right in the short term, and the sell-off is not necessarily the end of the story.
The core value of the technology is still real.
Zhipu's GLM models are competitive. MiniMax's product, especially the C-end products, have a real user base. The fundamental technology is not a Ponzi scheme. There is actual, tangible technology here, and the market will not completely ignore that.
This might be a case of the market overcorrecting.
The market is currently pricing in a total lack of trust in the AI sector's profitability. But the market is also ignoring the massive potential in the Chinese AI ecosystem. The government is pushing for AI adoption. The market is demanding a 10% - 20% drop in valuation for these companies, but the market has not yet priced in the full value of the potential of the Chinese government's subsidies and support.
The market is creating a "Price of Admission" for these companies.
The market is simply saying: "Prove to me that you are not a Ponzi scheme." The Ponzi schemes leave trails in the data. The market is now demanding to see that the trails are not just a marketing budget.
The fall is not the final ledger. The fall is the opening of a new block. The market is forcing these companies to prove their commercial value. This is the necessary pain for the industry to mature.
The Takeaway: The Audit, Not the Verdict
The market has spoken. The immediate verdict is: the valuation was a Ponzi scheme of narrative, not a Ponzi scheme of actual value. The remaining question is whether these companies have the real substance to recover.
The block chain remembers what humans forget. The same is true of the stock market. The market will forget the initial narrative, but it will not forget the data. If Zhipu and MiniMax can demonstrate strong revenue growth, customer retention, and a clear path to profitability, the market will eventually re-rate them. If they can't, they will be stuck in the same low-priced zone as SenseTime.
The industry is entering the "verification phase" of AI. The code does not lie; intent does. And the intent is to turn a technology into a business. If the intent is clear, the price will follow. If the intent is just to sell narrative, the market will not be patient.
Verify the hash, trust no one. The public market has just verified the hash, and the number is not matching the narrative. The smart investor will wait for the next block. The next block is the quarterly earnings report.
The current price is a signal. The revenue is the confirmation. Do not act on the signal alone. Wait for the confirmation.
The silence is the only honest ledger. The silence is what happens between the IPO and the earnings call. Watch the silence. The data will speak.
