Hong Kong Stock Market's AI Giants Hit a Braking Shock: Spending Money Faster than Earning It, Market Losing Confidence
Hello everyone, I'm your financial journalist. Today, we're talking about an event that can be described as a "earthquake" in the Hong Kong stock market on September 14th.
If you follow tech news, you might know that AI (Artificial Intelligence) has been particularly hot lately, especially with Chinese companies like Zhipu and MiniMax, which were highly touted before. But in the past couple of days, the wind has turned. The stock prices of Zhipu and MiniMax have plummeted, dragging down the entire AI sector in Hong Kong stock market as well.
What exactly is going on behind this? Is the AI industry in trouble, or have these companies exposed their weaknesses? Don't worry, let's break it down in simple terms.
---
Summary of the Key Points: A Crisis of Confidence Triggered by the Speed of Spending and Changes in the Industry
In short, the decline in the Hong Kong AI sector wasn't caused by a single factor; it was the result of both internal issues and external signals.
1. Internal Problems (Zhipu, MiniMax): Although these companies have seen rapid revenue growth, they are spending money much faster than they are earning it. Zhipu, which went public just a few months ago, has almost spent all the funds it raised, and the rate of expenditure has even accelerated tenfold. The market is worried: when will they start to generate their own revenue?
2. External Signals (OpenAI, Anthropic): The leading AI companies in the U.S. have suddenly stated that they need to slow down, emphasizing safety and alignment rather than blindly pursuing performance improvements. This has shaken the foundation of AI stock valuations. Previously, investors bought AI stocks with the expectation that models would become stronger and stronger. Now that the leaders are saying they need to slow down, do those previous expectations need to be re-evaluated?
The result is that investors are re-examining these high-valued, high-loss AI companies, and funds are flowing out, leading to a sharp drop in their stock prices.
---
In-Depth Analysis: Understanding the Adjustment from Five Dimensions
1. Where Does the Money Come From, and Where Does It Go? – The Concerns About a Vicious Cycle of Financing and Spending
Let's look at Zhipu's actions. On September 13th, Zhipu announced plans to raise another $5 billion (about HK$39.2 billion). The amount is significant, but what's more alarming to investors is how quickly the money is being spent:
- Previous Pace: Zhipu went public in January 2026 and raised about HK$4.9 billion. At that time, it seemed like there would be enough money for a long time.
- Current Pace: By July, that HK$4.9 billion was almost gone. Then, in July, it issued additional shares to raise HK$31.3 billion, and within 50 days, HK$10.9 billion was spent, averaging out to 220 million per day.
- Dramatic Acceleration: The monthly spending increased from 600 million during the IPO phase to 6.5 billion after July. In just two months, the spending rate nearly doubled!
In Simple Terms: It's like a startup that receives its first investment and says it will last a year, but it runs out in half a year and quickly asks for more funding. Investors ask, "Why are you spending so fast?" The company replies, "We need to buy better graphics cards to train smarter models." While the reason sounds reasonable, the logic of the capital market is: if you can't prove that this money will turn into profits or at least stable revenue, you're just filling a bottomless pit. Zhipu's new round of financing, although intended for developing the next generation of models, shows that its spending rate is accelerating, while its ability to generate revenue has not kept up. This pattern of relying on financing to survive is more tolerable in a bull market, but once the market sentiment changes, investors will ask when the next round of funding will come. If it can't be secured, the company may face a crisis.
2. Is the Profit Enough to Cover Expenses? – The Profit Trap Behind High Revenue Growth
Many people saw the news that Zhipu's revenue grew by 399% in the first half of the year, and MiniMax by 283%, and thought, "Wow, the business is booming!" But as investors, we need to look at more than just revenue; we also need to look at gross profit and net profit:
- Zhipu's Situation:
- Revenue increased by 954 million yuan, which is impressive.
- However, the gross profit margin dropped from 50% to 26.4%. This means that for every 100 yuan in sales, it used to earn 50 yuan; now it only earns 26 yuan.
- More importantly, research and development (R&D) expenses accounted for 2.2 times its revenue. In other words, for every 1 yuan earned, 2.2 yuan is spent on R&D.
- MiniMax's Situation:
- Revenue increased by 283% to 117 million dollars.
- R&D expenses were 297 million dollars, 2.5 times its revenue.
- The gross profit margin was only 17.9%, resulting in a net loss of 293 million dollars after adjustments.
In Simple Terms: It's like opening a restaurant that sees more customers (revenue growth), but the cost of each dish (computing power, electricity, R&D) is increasing faster than the price of the dishes. You might have been able to survive with low margins or subsidies before, but now that the gross profit margin has dropped significantly, it means your products are not becoming cheaper or more efficient due to the expansion. Instead, you have to invest more money in training models and purchasing computing power to stay competitive.
Core Conflict: The AI industry is highly capital-intensive, similar to building a highway, which requires a large initial investment. Once completed, the marginal cost is low. However, these companies seem to be in a situation where they need to keep investing just to stay ahead. Revenue is increasing, but profits are declining, and the decline is getting worse, which is what the capital market fears the most.
3. Why Did the U.S. Leaders' Announcement Shock the Hong Kong Market? – The Collapse of Valuation Logic
Another major factor in the decline was the statements from leaders in the U.S., such as OpenAI and Anthropic, suggesting they needed to slow down and focus on safety and alignment:
- Previous Narrative: AI stocks were expensive because people believed the "Moore's Law" also applied to AI—models would become larger, smarter, and faster. Investors bought stocks based on the expectation that models would outperform their competitors.
- Current Narrative: The leading companies are saying, "We need to slow down because going too fast can be uncontrollable, and we need time to focus on security measures."
In Simple Terms: It's like a racing competition where everyone was betting on who would go the fastest, so the fastest teams got the highest valuations. Now, the champion team says, "We need to slow down because the track is too dangerous, and we need to build safety barriers first." This puts other teams (like Zhipu and MiniMax) in an awkward position:
1. Expectations Not Met: If the industry slows down, the high valuations based on rapid iteration no longer hold.
2. Changing Competitive Landscape: If the progress of leading models slows, companies that rely on the most advanced models to attract customers may lose their advantage.
3. Shift in Capital Flow: Investors might think it's better to invest in companies that are already making money or have more practical applications rather than continuing to fund cutting-edge labs.
Key Point: Anthropic is about to go public and has already turned a profit, which sends a strong signal to the market: AI can be profitable without endless spending. In contrast, companies in Hong Kong that are still losing huge amounts of money seem less attractive.
4. The Market Is No Longer "Tolerant": From "Storytelling" to "Looking at Financial Reports"
In the past two years, the capital market has been very tolerant of AI companies. As long as their stories sounded good and their technology sounded impressive, even with large losses, their stock prices could rise. But this tolerance is quickly fading:
- Anthropic's Example: Anthropic's revenue in the second quarter exceeded 1.15 billion dollars, and it turned a profit for the first time, preparing for an IPO. This sets a new standard: AI companies can be profitable.
- The Contrast: When one AI company becomes profitable and goes public, another one is still spending 220 million dollars per month and sees its gross profit margin decline. Naturally, investors' preferences shift.
In Simple Terms: Previously, investors viewed AI companies as potential stocks with great future potential. Now, they are evaluating them like mature companies, asking questions like: What's your cash flow like? Can you maintain a stable gross profit margin? Can your revenue cover your expenses? This shift in evaluation means the era of pure concept speculation is over. Investors no longer want to pay high prices for future possibilities; they want to see current certainty. For Zhipu and MiniMax, this means they need to prove their commercialization capabilities quickly, or their valuations could be re-evaluated, leading to significant stock price adjustments.
5. Medium- and Long-Term Impacts: Who Will Be Hurt, and Who Will Benefit?
This adjustment is not just about short-term stock price fluctuations; it may also indicate a turning point in the AI industry:
- Short-Term Pressures:
- Large Model Manufacturers: Companies like Zhipu and MiniMax will face pressure on their valuations, and it may become harder for them to raise funds.
- AI Hardware/Computing Power Stocks: If the pace of model iteration slows down, the demand for computing power may decrease, affecting related stocks (such as chips and optical communications).
- Medium- and Long-Term Benefits:
- AI Application Companies: If the pace of model iteration slows, companies that focus on specific industries (such as healthcare, law, finance) will have more room to thrive because business customers value stability, accuracy, and cost-effectiveness.
- Domestic Computing Power: According to Zhejiang Securities, global token usage is still growing rapidly, indicating strong demand for computing power. This means that although the development of leading models may slow down, the number of people and businesses using AI is still increasing, which is good news for domestic computing power infrastructure.
In Simple Terms: This adjustment is the market squeezing out bubbles. Those hurt are pure R&D companies that rely on technical stories to maintain their valuations but lack actual profitability. Those that benefit are companies that can apply AI technology to solve real problems and control costs. The big trend of AI remains unchanged: the demand for AI is still growing, but the logic of how to make money has changed—now it's about which applications are more stable and cost-effective.
---
Conclusion
The sharp decline in the Hong Kong AI sector on September 14th was not just a simple market fluctuation; it was a reshaping of the industry's logic.
It tells us that:
1. Spending Money Is Not a Guarantee: The capital market will no longer support endless R&D; efficiency and profitability are now the new focus.
2. Balance Between Safety and Speed: The call for slower development by industry leaders reminds us that AI must pursue stability as well as speed.
3. Applications Are Key: In the future, companies that can integrate AI into their businesses and generate stable cash flows will be the ones that succeed.
For investors, this means that when investing in AI-related stocks, they should no longer rely solely on concepts and growth rates but also on **gross profit margins, cash flows, and commercialization capabilities. The era of AI's rapid growth may have temporarily come to an end, but the era of focused, practical applications has just begun.