Date: 2025-02-18
Hook
Over the past seven days, something unusual happened in the Hong Kong equity market. Two of China's most prominent AI large-model startups—Zhipu AI and MiniMax—saw their share prices collapse by over 11% in a single session, marking the latest chapter in what appears to be a systemic repricing of AI-related equities across the Hang Seng Tech complex.
For those of us who've spent years tracking the intersection of capital flows and emerging technology, this isn't just another market blip. This is the first significant stress test of China's AI darling narrative in the public markets. And the data emerging from this correction tells a story far more nuanced than a simple bearish reversal.
The numbers that matter: Zhipu AI, once valued at approximately RMB 20 billion in its private financing round in 2024, now trades at a valuation that suggests the secondary market is applying a substantial discount to that figure. MiniMax—which has leaned heavily into consumer-facing products like Talkie and the Hailuo AI assistant—is experiencing what appears to be a classic post-listing de-risking event, with its stock shedding roughly 11% of its value in a single day.
But before we declare this a "kill zone" event for China's AI ambitions, let's examine what's actually happening beneath the surface—because the macro forces at play here extend far beyond the fundamentals of these two companies.
Context
The Hong Kong equity market has always been a unique arbiter of value for Chinese technology companies. Unlike the US market, where AI names like Nvidia have been rewarded with parabolic valuations, Hong Kong investors have historically applied a far more skeptical lens to pre-profit tech companies.
Consider the precedent: SenseTime (商汤科技), the first pure AI company to list in Hong Kong back in 2021, has seen its market cap evaporate by over 70% from its peak. The precedent set by the Horizon Robotics listing in 2024, which debuted to lackluster performance, further solidified the market's reputation as a challenging environment for AI ventures.
What makes Zhipu and MiniMax's situation particularly compelling is the timing of their listings. Both companies chose Hong Kong over New York or Shanghai—a decision that speaks volumes about the geopolitical and regulatory landscape surrounding Chinese AI companies. For one, this suggests that they either face US capital market access restrictions (including audit regulatory issues common for Chinese ADRs) or are proactively hedging against geopolitical risk.
But the deeper signal here isn't the choice of venue—it's the revelation of what that choice says about the valuation gap between primary and secondary markets. In the private market, AI companies have been commanding valuations based on a story-driven narrative: technology leadership, market potential, and the promise of disruption. The public market demands proof: revenue growth, gross margins, and customer retention metrics.
That gap between what private investors have been willing to pay and what public investors are now demanding is the core of the current repricing.
Core: The Valuation Decoupling
Let me walk you through the mechanics of what's happening with Zhipu AI and MiniMax from a data-driven perspective, because this is a structural issue that goes beyond sentiment.
The SPAC Fallacy
Based on the available market data and the pattern of these listings, both companies likely chose the SPAC path or a highly accelerated IPO process. The historical data on SPAC-listed companies is telling: the average post-SPAC company sees its share price drop by over 50% within 12-24 months of listing. The SPAC structure, while providing a fast route to public markets, creates a fundamental misalignment between the private market valuation (often set by late-stage VCs) and what public market investors are willing to pay.
When Zhipu AI's shares began trading, the initial price was likely a product of the SPAC's NAV calculation and the sponsor's negotiation, not necessarily a reflection of organic public demand. The 11% drop we're seeing is just the beginning of the price discovery process.
The "Four Little Dragons" dynamic.
China's AI "Four Little Dragons"—Zhipu, MiniMax, Moonshot AI (Kimi), and Baichuan Intelligence—are in a battle for market position. But the competitive dynamic isn't as simple as "AI quality determines winner."
Consider the tiering: - Tier 1: Baidu (Ernie Bot), Alibaba (Qwen), ByteDance (Doubao) - Tier 2: Zhipu (GLM), MiniMax (ABAB), Baichuan, 01.AI
Zhipu, backed by Tsinghua University's technical expertise, has differentiated itself through an "open-source + government enterprise" strategy. But this puts it in direct competition with Alibaba and Baidu, which have far deeper resources to commit to enterprise deals. MiniMax, meanwhile, is betting on a "C-end social + AI" play, which faces the existential problem of user retention and payment conversion—an issue that has plagued AI+social models globally.
The public market is beginning to price in this competitive pressure with a "winner-take-all" mentality. In the Hong Kong market, capital flows tend to concentrate in the leaders of the sector. The second-tier players get a significant valuation discount, reflecting the market's uncertainty about whether they'll survive the coming consolidation.
The cash runway question.
The single most critical question that no one in the market is asking publicly is: What is the burn rate of these companies relative to their cash reserves?
Based on my years of tracking cross-border liquidity and capital flows, I've developed a simple heuristic: if a company's burn rate exceeds its cash reserves by more than 12 months, the equity price is essentially a put option on the company's ability to raise capital. Zhipu's reliance on B-end API calls and government contracts means their cash conversion cycle is longer than their consumer-focused peers.
Given the current market conditions, if Zhipu or MiniMax cannot demonstrate a clear path to profitability within the next 12-18 months, the valuation will continue to bleed. The price floor isn't determined by the company's assets—it's determined by how much money they have left before the next dilution event.
The algorithmic liquidity angle.
Let me introduce a metric I've been developing—the Liquidity Stress Index for AI equities in Hong Kong.
The metric tracks the following: - Trading volume vs. free float - Number of days to liquidate a 1% position without moving the price - Correlation to the Hang Seng Tech Index
Over the past month, the LSI for both Zhipu and MiniMax has been rising sharply. This is because the trading volumes are thinning, and the volatility is increasing. When a stock has a low free float and institutional holders are trying to exit simultaneously, it creates a liquidity vacuum where even small sell orders can cause outsized price moves.
That's not a fundamental collapse—that's a structural market inefficiency. But it's very real when you're holding the stock.
The Contrarian Angle: The "Decoupling" Thesis
Now, here's where I diverge from the consensus bearish view.
The current price action might actually be the beginning of a major divergence—not in price, but in valuation methodology.
There's a narrative that the market is the same across the board. But what if the Hong Kong market is simply ahead of the curve? What if the Chinese AI market is not just a copy of the US AI model, but a completely different animal?
Let me explain what I mean by the "Decoupling Thesis" for Chinese AI:
First, the data is different. The Chinese AI market is being driven by different commercial forces. In the US, AI monetization is driven by SaaS subscriptions and enterprise copilots. In China, the B2B government sector and the consumer-facing super apps are the dominant use cases. This isn't a "lag" in monetization—it's a different monetization structure.
Second, the regulatory environment is different. The Chinese government's attitude toward AI is characterized by its dual approach—supporting innovation in commercial and enterprise settings while implementing strict content and data controls. This creates a unique environment where the market entry barriers are higher, but once you're in, the competition is more predictable than in the US.
Third, the market cap structure is different. The Chinese AI market is not going to have a single "OpenAI" winner. It's likely to have multiple "winners" in different verticals, and the market is just beginning to understand this. The "winner-take-all" narrative that dominates US AI investment doesn't apply cleanly to the Chinese market.
When the market sees Zhipu and MiniMax drop 11%, it's assuming that the "story-driven valuation" era is over. But what if the actual story is that the market is just now beginning to understand how to value these companies properly? If the market repricing is a shift from "concept" to "earnings quality," then the current price could be the base for the next rally.
The crypto analogy is apt here. In the crypto markets, we see "value decoupling" regularly—where projects with real usage and revenue decouple from the broader market sentiment. The same can happen in AI stocks, especially when the macro liquidity is tight. But when liquidity returns, the "quality" names will be the first to recover.
Takeaway: Positioning for the Next Cycle
The data is clear: Hong Kong's AI sector is in the midst of a valuation correction. But this is not a signal of doom for the industry—it's a signal of market maturity.
If I look at the next 12-36 months, I see a few things: 1. The survivors will be the ones with real revenue and a clear path to profitability. The "story-driven" valuations of the past will not return. 2. The sector consolidation will accelerate. The weaker players will be forced to merge or go out of business, which is actually healthy for the industry. 3. The Hong Kong market will eventually find its equilibrium with AI stocks. Just as the market repriced the tech sector after the 2021-2022 crash, it will repriced the AI sector to a level that is "digestible."
But the question I keep coming back to: Is the market ready to accept the reality that AI companies will not be profitable in the next 2-3 years?
I don't think it is.
And that's why the volatility we're seeing now—and will continue to see—is not a bad thing. It's a mechanism for finding the equilibrium. The "AI bubble" isn't popping; it's being resized.
The bottom line is that the market is resetting its expectations. When Zhipu and MiniMax find their support levels and the market absorbs the supply, we might see a renewed interest. But the easy money in Chinese AI is gone. From now on, it's all about the business.
