We didn't need another red candle to tell us something was wrong. We needed someone to explain why the market suddenly decided that two of China's most prominent AI startups were worth 11% and 10% less than they were the day before. On August 24th, Hong Kong-listed AI concept stocks took a beating—Zhipu dropped over 11%, and MINIMAX fell more than 10%. The headlines screamed. The data flowed. The explanations were... absent.
This is the story of what happens when market psychology collides with technical fundamentals, and why the blockchain community—of all people—should be paying attention.
The Context: When Valuations Meet Reality
Let's be precise about what we're actually discussing here. Zhipu (the team behind the GLM series of models) and MINIMAX (known for their abab series) are two of China's "Four Little Dragons" in the large language model space. Both have achieved national-level compliance certifications from the Cyberspace Administration of China. Both have raised significant capital—Zhipu's valuation exceeded ¥20 billion after its early 2024 funding round, and MINIMAX crossed the $1 billion threshold. Both were, until this week, the darlings of the AI narrative.
The market's sudden recalibration is not about technology—it's about belief.
Here's the uncomfortable truth that nobody in the blockchain space should find surprising: when a market is built on narrative rather than cash flows, sentiment becomes the only true pricing mechanism. I've spent years auditing smart contracts and watching decentralized protocols face exactly this phenomenon. The patterns are eerily familiar.
The Core: What Actually Happened
Based on my experience analyzing market dislocations in the crypto space, the situation can be broken down into four interconnected drivers.
First, valuation pressure. When Zhipu was valued at 200 billion RMB and MINIMAX at over a billion dollars, the market was pricing in not just current performance, but a decade of unbridled growth. The revenue multiples were astronomical. When investors look at these numbers against actual API revenue, the gap creates cognitive dissonance. And cognitive dissonance in markets creates sell orders.
Second, competition dynamics. The AI landscape in China has become a price war that rivals anything we've seen in commodity markets. Companies like Baidu, Alibaba, and ByteDance have slashed API prices by up to 90% to capture market share. For startups like Zhipu and MINIMAX, this creates a margin compression scenario that challenges even the most optimistic unit economics. Meanwhile, DeepSeek's open-source V3/R1 models have captured massive community mindshare, and Moonshot's Kimi has dominated the long-text niche. Zhipu and MINIMAX sit in the awkward middle—not differentiated enough to command premiums, not large enough to survive a prolonged price war.
Third, market sentiment. The broader global AI sector has experienced volatility throughout 2024. Investors are becoming impatient with the "AI investment returns will come eventually" thesis. Hong Kong's liquidity constraints amplify this impatience—when sell pressure hits a high-valuation, pre-profit company on a thin order book, the price impact is magnified.
4. Funding realities. The implications extend beyond stock prices. A correction at the public market level reverberates into private fundraising. When Zhipu and MINIMAX next raise capital, they will be negotiating against their recent public market performance. This creates a self-reinforcing cycle of valuation compression.
The Contrarian Angle: What the Market Gets Wrong
Here's where I need to push back against the herd. The market's negative reaction tells you more about market structure than about the companies' technology.
Let me draw from my experience auditing smart contracts and protocol governance. Time and time again, I've seen the market price a protocol at a fraction of its eventual value during a bear cycle, then watch it recover to a premium when the fundamentals become undeniable. The same pattern plays out in AI markets.
The AI industry is still in its infrastructure phase. We're at the equivalent of blockchain's 2017 era—building roads and railways, not yet delivering goods. The demand for AI capabilities is real and growing—code generation, intelligent customer service, content creation. The actual usage data shows growing API adoption, not declining. What's declining is the market's patience.
The core insight: market corrections in technology sectors are frequently re-ratings of timing, not final verdicts on the technology. This is true in blockchain, and it's true in AI. If Zhipu and MINIMAX can demonstrate viable unit economics—if they can show that their API revenue is growing, that their enterprise clients are retaining, that their vertical-specific solutions are gaining traction—then today's price is a narrative opportunity, not a fundamental judgment.
There's also a deeper issue: the AI infrastructure story is entangled with compute constraints. Zhipu and MINIMAX are dependent on GPU clusters that are expensive and, due to export controls, constrained. The cost of compute is one of the largest components of their operating expenses. In a market where capital is becoming scarcer, this creates a real operational risk.
The Blockchain Parallel: Why We Should Care
I've spent the last several years building and observing decentralized systems, and I see the same structural issues in AI that we saw in crypto during the "DeFi Summer" of 2020. The hype cycle creates valuations that are impossible to justify by current metrics. The market corrects when the gap between narrative and reality becomes unsustainable. The projects that survive are not necessarily the ones with the best technology—they're the ones that can demonstrate pragmatic, sustainable use.
Zhipu and MINIMAX are facing their "DeFi Summer" moment. The question is whether they can pivot from being narrative-driven to being value-driven.
What Comes Next
In the coming weeks, I'll be watching for the following signals: whether Zhipu and MINIMAX announce new model releases or major enterprise partnerships that could serve as a catalyst; whether the Hong Kong AI sector as a whole recovers or whether this is a broader market rotation; and whether there's any public disclosure of API utilization data or revenue figures.
The AI sector is in a phase where infrastructure is being built, but the market is asking for revenue. This is the same phase the blockchain industry went through, and I've seen how it plays out: the strong build, the weak fade, and the investors who keep a clear head in the middle of a sell-off are the ones who capture the best risk/reward.
We didn't need another market drop to tell us that AI is real. But we do need to remind ourselves that price and value are not the same thing—and in both AI and crypto, we are still in the early innings.
The correction is a signal, not an obituary. For those of us who've been through a few market cycles, the playbook is clear: look at the fundamentals, ignore the noise, and build for the long term. The truth doesn't get priced in overnight—it gets priced in over time.