The market has lost faith in the story. Over the past week, MiniMax and Zhipu AI, two of China's most prominent AI large-model companies, have become the most shorted stocks on the Hong Kong exchange. The numbers are staggering: MiniMax's short interest has hit a record 20%, while Zhipu AI's share price has collapsed more than 50% from its peak. This isn't just a market correction—it's a systemic rejection of the "pure AI model company" business model.
The narrative that drove these companies to astronomical valuations is dead. The market is no longer paying for technical leadership alone. The transition from "story-driven" to "data-driven" pricing is happening in real-time. Investors are asking a brutal question: Can these companies ever become profitable? The answer, based on the current data, is a resounding "not anytime soon."
The core issue is the market's shifting sentiment. The unlock of IPO lock-up periods has flooded the market with supply, creating a overhang of approximately $11.5 billion worth of shares. This forced sellers, combined with a brutal price war in the AI model space, has created a perfect storm. The stock is not pricing in the future; it's pricing in the pain.
I've been tracking this convergence of AI and crypto markets, and the pattern is hauntingly familiar. It's the same shift we saw with liquidity mining: when the incentives stop, the real users vanish. Here, when the narrative stops, the capital disappears. The question is whether these companies have any real substance beneath the hype.
The data from Jefferies is revealing. Zhipu's GLM-5.3 model delivers performance comparable to MiniMax's Kimi K3, but at a 19% lower cost per task. This engineering efficiency is a genuine technical advantage. Yet, the market's reaction is cold. The stock didn't surge on this news; it continued to fall. Why? Because in the current environment, "cheaper and similar" isn't enough. It's not a competitive edge; it's just table stakes. The market is waiting for a breakthrough, not an incremental improvement.
Meanwhile, the bearish thesis from firms like Hedgeye is brutal. They argue Zhipu's gross margins are structurally capped by price wars. They see MiniMax as "neither the smartest nor the cheapest," a company stuck in an awkward middle ground. I find this analysis to be dangerously accurate. When you are a mid-tier player in a race against well-funded giants like DeepSeek and Alibaba, your technical advantage is a temporary reprieve, not a moat.
The market is now pricing for a world where these companies cannot compete on technology, so they must compete on price. And price competition in this capital-intensive industry is a race to the bottom. The Southbound capital, which is steadily buying the dip, seems to be betting on a long-term sectoral growth story. But they might be catching a falling knife.
The data points are undeniable. From the record short interest to the post-IPO bloodbath, the market is signaling a fundamental shift in how AI companies are valued. The old metric of "technology superiority" is now being replaced by "time to profitability." The market wants to see a clear path to earnings, not just a demo video.
Let me be blunt: I've analyzed the unlock schedules and the cash burn rates. The math is ugly. At current price points, the stock is pricing in a near-zero probability of profitability in the next 12 months. The next earnings report in August will be the moment of truth. If it shows revenue growth is not enough to cover the massive operational costs, the valuation will continue to be crushed. The short thesis is not just about the numbers; it's about the speed of the market's realization.
The contrarian angle here is the whisper about the untold story. The market is focusing on the negative, but I see a potential pivot. What if the market is underestimating the potential for these companies to become acquisition targets? With valuations down more than 80% from peak, they are becoming attractive to tech giants looking to acquire AI talent and technology at a discount. Or, a more radical shift: what if the short-sellers are too focused on the "model" and missing the "agentic" AI application layer? The real value might be in the application, not the model. But that's a speculative bet.
This is a classic case of the market punishing the "story" and rewarding the "business." The market is not wrong; it's just early. The correction is painful, but it's a necessary purge. It's a clear signal that the AI market is maturing from a speculative phase into a monetization phase. The speed of the market's move is the pulse of the new reality. It's a lesson we've seen in DeFi, and now it's being replayed in AI. The market doesn't care about the best technology; it cares about the most sustainable business model.
Regulation doesn't need to catch up here. The market is the ultimate regulator. It's sorting out the winners from the losers. The question is: who's left standing when the dust settles? From chaos to clarity: we are tracking the summer of the shakeout. The exchange leads see the wave before it breaks. The next watch is the earnings call. Will the management present a viable path to profitability, or will they double down on a vision that the market no longer buys? That is the question, and the answer is coming soon.