The Covenant Breaks: When AI's Promise Meets Hong Kong's Ledger

There is a particular silence that follows a sharp decline. It is not the quiet of peace, but the hush of a covenant broken. Over the past seven days, the tickers for Zhipu AI and MiniMax, two of China's most storied AI model companies, bled more than 11% on the Hong Kong exchange. The numbers flashed on screens, but the true signal was deeper—a whispered reassessment of what these companies are actually worth. My code was the covenant, not just the contract. Yet here, in the cold light of a Hong Kong trading session, the market was asking a question that no whitepaper had ever answered: What does an unredeemed promise cost? We have entered a season where the narrative of intelligence itself is being audited, and the auditors are not forgiving. To understand this drop, we must first understand the terrain. For the better part of three years, the Chinese AI model market was a theater of the grand. The so-called "Four Little Dragons"—Zhipu, MiniMax, Moonshot AI, and Baichuan—were not just companies but banners. They represented a national ambition, a counterweight to the American giants. The capital that flowed into them was not just money; it was faith, encoded as term sheets. Zhipu, with its Tsinghua lineage and the GLM series, positioned itself as the intellectual, the sovereign servant of enterprise and state. MiniMax, with its consumer-facing products like Talkie, sought to be the emotional, the one who made the machine feel like a companion. The market, in its wisdom, initially rewarded this. The valuations were set in the private markets, where stories are still more potent than revenue. But Hong Kong is a different beast. It is a ledger that does not dream. It is a place where the liquidity is real, and the patience is not. When these companies chose to list there, they submitted their high-minded narratives to a rigorous, often cruel, test. And in the last week, the test has not been kind. The core insight here is a shift, a paradigm conversion that is more profound than a simple sell-off. We are witnessing the completion of a migration: from the story-driven era to the performance-driven era. In the private markets of 2023 and 2024, an AI company's value was a function of its potential—the sheer size of the addressable market for synthetic intelligence, the strategic imperative to own the foundational model. Metrics like revenue, gross margin, and customer retention were secondary, almost trivial. But on the Hong Kong exchange, the code is the law, and the law demands earnings. The investors there, particularly the northbound flow from the mainland, are not tolerant of the "burn for growth" ethos. They have seen too many stories fail. Look at the precedent. SenseTime, the first AI stock in Hong Kong, has seen its value evaporate by over 70% from its peak. Horizon Robotics, despite a strong listing in 2024, has struggled to find a bid. The Hong Kong market is not a garden for unprofitable technology; it is a high-stakes poker table where only the bottom line wins. When Zhipu and MiniMax arrived, they were not just bringing their models; they were bringing a valuation anchor from the private markets that was often 2x to 3x higher than what the public market was willing to underwrite. The drop we are seeing is not a crash; it is a convergence. The price is moving towards the value. This is where the contrarian angle emerges. The narrative says that this decline is a sign of the AI winter. That it is a failure of the Chinese model or a loss of confidence. But I argue that this is a moment of clarity, a brutal but necessary purification. We are not seeing the end of the AI story; we are seeing the end of the story-only phase. In the long history of decentralized systems, we have seen that the initial influx of speculative capital always leads to a correction, and the correction is what forces the real builders to dig deeper. This is not a bear market; it is the bear market that weeds out the tourists. We are currently in a season of doubt, where the market is testing the very foundation of the two companies. The challenge is whether they have a moat, or just a hypothesis. For Zhipu, the question is whether its enterprise and government contracts are sticky enough to justify its valuation. Is the API call volume growing, or is it just being used for pilot projects? For MiniMax, the question is whether its social products can overcome the industry-wide problem of user retention. Are the subscriptions coming back, or is it a novelty that fades? The silence in the bear market is often where the truth is found. In the silence of the bear, we heard the truth. The truth here is that the floor is not a number; it is a foundation of real users and real margins. For those of us who are looking at this from the blockchain perspective, this is a familiar rhythm. The crypto market is full of projects that promised a decentralized future but could not build a working product. We are now seeing the same pattern in the AI market. The first era was about the technology, the second about the story, and the third is about the story. What is the opportunity then? The opportunity is not in the decline, but in the aftermath. Once the market has fully priced in the risk, the value will be left for those who are patient. If Zhipu and MiniMax can pivot, can prove that they are not just a model but a business, then the current levels might be the base of a new cycle. I have seen this in the crypto markets—the projects that survive the bear market are the ones that are building for the long term, not the ones that are just pumping the token price. The same principle applies here. The key is to watch the fundamental metrics, not the red candles. My code was the covenant, not just the contract. The market is a covenant of its own, a promise between the company and the investor. The investor promises capital, and the company promises returns. When the returns do not materialize, the covenant is broken, and the price falls. But the covenant can be renewed. It is a process of the sacred duty to build, to code, and to deliver. The current bloodbath is not the end; it is the cleansing. The question is not if they will recover, but who will be standing when the silence is over. Will they have the resilience to build in the noise, to find the signal? I believe the honest builders will. They always do. The market is a harsh teacher, but it is a teacher. The lessons learned in the last week will not be forgotten. The question is not whether the AI companies are dead, but whether they are reborn. Every broken token taught me how to hold value, and the broken price will teach them how to build. The future is not about the narrative; it is about the code. The code is the new covenant. The only thing that will save the market is the code that works. The rest is just noise.