Hook
On a humid Tuesday morning in late July, I watched something remarkable unfold across my trading terminals. MiniMax and Zhipu AI, two of China's most prominent pure-play AI model companies listed in Hong Kong, were bleeding value at a pace that even seasoned crypto traders would find dizzying. Kimi K3's release had just hit the market, and the response wasn't the typical "buy the hype" enthusiasm—it was a brutal 24% and 18% decline for Zhipu and MiniMax respectively.
But here's what really caught my attention: the short interest ratio on MiniMax had hit 20%. Twenty percent. That's not a market wobble. That's a coordinated assault.
As someone who's spent years teaching communities to audit risk in decentralized systems, I've learned to recognize the difference between healthy skepticism and organized short attacks. This wasn't just a market correction. This was Wall Street's playbook being deployed against China's AI champions, and the implications ripple far beyond two Hong Kong-listed companies.
Context: The Tale of Two AIs
ZhipuAI, once valued as a "little giant" of Chinese AI, has seen its stock climb over 800% from IPO price even after halving from its peak. MiniMax, riding the "Short Video" wave with its Talkie and Hailuo platforms, found itself in the crosshairs of investors who suddenly wanted profits, not promise.
But here's the uncomfortable truth neither company's public relations team wants to admit: the market's shift from "story-driven" to "data-driven" evaluation is not a Chinese phenomenon. It's a global one. The same forces that caused Ethereum to lose its narrative grip in 2022, that forced NFT platforms to confront their lack of utility, are now knocking on the doors of China's most celebrated AI laboratories.
The trigger? July brought Kimi K3—the latest model from rival Moonshot AI—which Jefferies analysts noted performs similarly to Zhipu's GLM-5.3 but at 19% lower cost per task. On paper, Zhipu should have celebrated this validation of its engineering efficiency. Instead, the market read it as confirmation that the AI industry has entered a price war, and both Zhipu and MiniMax lack the pricing power to survive it.
Core: The Decentralization of AI Value
Let me tell you what I've observed from my position running a crypto education platform. The market's shifting treatment of these AI companies echoes something I've witnessed in decentralized systems for years—the distinction between fundamental value and speculative value is blurring.
The Bear Case: A Framework of Accountability
Hedgeye's bearish stance on Zhipu AI centers on a valid observation: price wars are crushing their ability to raise prices and maintain margins. When Kimi K3 enters the market at lower cost per task, and DeepSeek's open-source models offer comparable performance at a fraction of the cost, the AI model layer is becoming commoditized.
In my years analyzing DeFi protocols, I've seen this pattern repeatedly. When Uniswap faced competition from every new automated market maker, the network effect of liquidity saved it. But when foundational technology becomes a commodity, the winners aren't those with the best tech—they're those with the strongest distribution, brand, and customer lock-in.
For MiniMax and Zhipu, the tech is good. But the distribution is weak. The brand is still forming. And the lock-in is nonexistent when switching costs are measured in API endpoints, not years of accumulated trust.
The Red Flag That Matters Most: Locked Up
On July 7th, the IPO lockup period ended for both companies. ZhipuAI released 25.68 million shares, and MiniMax released 150 million shares. At current prices, that's roughly $11.5 billion in unlock pressure—a tidal wave of shares that early investors have been waiting years to sell.
Based on my experience auditing financial markets, I can tell you this is the most significant signal. In crypto, we call this "unlock dump" and it's the single most predictable driver of token price decline. The market knows these shares are coming, so the smart money positions in advance, creating a self-fulfilling prophecy.
The "buy the rumor, sell the news" effect is amplified when the "news" is locked-in shares.
But here's the nuance most analysis misses: Southbound capital—mainland Chinese investors—has been steadily buying. Zhipu's southern capital holding is around 12%, and MiniMax's is 8.1%. This is significant because it indicates that a segment of investors sees value in the price that Western and Hong Kong investors are selling at. It's a classic "value trap" versus "contrarian opportunity" dynamic.
The Contrarian: The Unseen Value in the Blood
Every short seller, every bearish analyst, focuses on the numbers—the decline in prices, the unlock pressure, the price war. But what they're missing is the story of the community that's being built.
Let me be clear: I've never been a fan of the "AI will save us all" narrative. I've seen too many technology cycles where the promise exceeds the delivery. But I've also seen what happens when a technology becomes deeply woven into the fabric of daily life.
In 2017, when I was building my first educational modules for blockchain, everyone thought it was a bubble. The "experts" were saying the same things about Bitcoin that they're now saying about AI models: it's overvalued, it's a "greater fool" story, it doesn't have real use cases. But what I learned—and what the shorts miss—is that technology's adoption curve doesn't follow the financial models.
The core insight that gets lost in the noise is this: the AI model market is not a winner-take-all game. It's a multi-layered ecosystem where different models serve different purposes.
When I built my education platform, I didn't need the most powerful AI model. I needed a model that was affordable, reliable, and could be integrated into my workflow without breaking the bank. That's the market that Zhipu and MiniMax are serving.
Zhipu's 19% cost advantage over Kimi K3 is not a weakness. It's a strategic position. In a market where every Chinese internet company is racing to build AI, the ability to deliver comparable performance at lower cost is a moat.
The short sellers are treating these companies as if they were in a race to the bottom. But what if they're not in a race to the bottom? What if they're building a foundation for the next wave of AI applications?
I've seen this play out in crypto. When everyone was shorting Bitcoin in 2022, the people who were building Lightning Network infrastructure and creating real use cases were unaffected. They knew that the price was a lagging indicator of the value being created.
The Core: Technology and Values
Let me get into the technical details. Both Zhipu and MiniMax are focusing on different aspects of the AI stack.
Zhipu's GLM-5.3 is a general-purpose model that competes directly with Kimi K3. The Jefferies note highlights that it performs similarly but costs less. This is not a negative—it's a positioning for the mass market. While the open-source community is focused on "benchmark-topping" models, Zhipu is targeting the enterprise and SME segment that wants cost-effective AI.
MiniMax, on the other hand, has a slightly different focus. It's leveraging its position in the "Short Video" market to create AI-powered content generation tools. The models are designed for creating video content, and this is a high-growth area. The market is undervaluing this because it's not a "general AI" play, but a specialized AI play.
The key difference between these two companies and the "pure play" AI companies like OpenAI or Anthropic is their vertical integration.
Zhipu has its own cloud infrastructure and is building out enterprise solutions. MiniMax has its own consumer apps and is building out content creation tools. They're not just selling APIs; they're building full-stack AI solutions.
This is exactly what happened in the blockchain space. The projects that succeeded were not the ones that just built a protocol; they were the ones that built an ecosystem. The ones that created a user experience, not just a technology.
The market is treating these companies as if they are just API providers with no moat. But they're building ecosystems. And ecosystems are sticky.
The Contrarian Angle: The Bear Case is Too Simple
Let me play devil's advocate with the bear case.
The short sellers are saying, "These companies can't make money because they're in a price war." But they're ignoring the fact that the price war is creating a massive new market. When the cost of AI drops, more businesses can afford to adopt it. The demand for AI is elastic, not fixed.
We've seen this in the storage industry, the cloud industry, and the mobile industry. When prices fall, the market expands. The companies that are positioned to capture that expansion are the ones that are the cheapest. Zhipu and MiniMax are positioned to be the cheapest.
The shorts are also ignoring the fact that the Southbound capital is buying. This suggests that mainland Chinese investors see value that Western investors don't. There could be information asymmetry here—perhaps the mainland investors have a better understanding of the government contracts and opportunities.
Finally, the shorts are ignoring the possibility of consolidation. Both Zhipu and MiniMax could be acquired by a larger tech giant. The stock price is down, which makes them attractive acquisition targets. This is a classic scenario where the shorts are right about the current quarter but wrong about the longer-term trend.
The Takeaway: What I've Learned
My community is not a user base; it is a shared soul. I've watched the same thing happen in AI. When the price drops, the community's resilience is tested. But if the foundation is strong, they survive.
What I see here is not a collapse but a decoupling of the AI industry. The first wave of AI companies that went public is facing the same reality check that crypto faced in 2018 and 2022: "The market will not pay for technology alone."
The real signal here is the "risk-first" approach. As an educator, I've always taught my students to think about the risk before the reward. The risk for these AI companies is real. The price war is real. The lockup pressure is real.
But the opportunity is also real. The AI industry is still in its early stages. The companies that can build the most compelling ecosystems, that can offer the best value, will win in the long run.
The shorts are betting that these companies are just like every other hot money play: a story that fades when the money runs out. But I've seen what happens when technology becomes a real infrastructure. It takes time, but it eventually becomes the foundation of everything else.
We build not for the token, but for the tribe. These AI companies are building for the tribe of developers, creators, and enterprises who need AI to be accessible, affordable, and practical.
The next few months will be the ultimate test. The earnings reports in August will tell us if the market is right to be bearish. But as I always say, "The market is a story, but the technology is a process." The process is still unfolding.
The Bigger Picture: The AI Market Is Not a Zero-Sum Game
Let me zoom out for a moment. The shorts are treating this as if Zhipu and MiniMax are in a fight with DeepSeek and Moonshot AI. But the market is much bigger than that.
The AI market is expanding globally. Every enterprise in China, every startup, every government agency is trying to figure out how to integrate AI. The pie is growing, and these companies are competing for a slice of that growing pie.
The real risk is not competition; it's the slowing of the adoption curve. If AI adoption slows, then all of these companies will suffer. But I don't see a slowdown. I see an acceleration.
The shorts are fighting the trend. They're betting on the fear of the future, not on the reality of the present. The reality is that AI is becoming the new electricity. And the companies that provide it at the lowest cost and best quality will be the utilities of the future.
Conclusion: The Forest and the Trees
The traditional financial analysts are looking at the trees—the margins, the revenue growth, the price-to-sales ratios. They're missing the forest—the fundamental shift in how value is created and captured in the AI era.
In my years of observing the decentralized ecosystem, I've learned that the most important metric is not the price of the token or the stock, but the growth of the network. The network is growing for Zhipu and MiniMax. The user base is expanding. The technology is improving.
I'm not saying this will be a smooth ride. The next few months will be volatile. The earnings reports will be messy. The unlock pressure will continue to sell.
But if I've learned anything from watching the rise and fall of crypto cycles, it's that the fundamentals eventually win. The companies that have the technology, the distribution, and the vision will survive the winter.
The shorts are making a bet on fear. I'm making a bet on the future.
What to Watch Next
The immediate focus should be on the August earnings reports. MiniMax's earnings (August 26) and Zhipu's (August 31) will be a critical test. I'll be looking at three key metrics: revenue growth, gross margins, and customer acquisition costs.
If these companies can show that they're not just burning cash but are building a sustainable business model, the shorts will be squeezed. If they show that the price war is destroying their margins, the bearish case will be confirmed.
Beyond the earnings, I'll be watching the next model release cycle. If Zhipu can leapfrog Kimi K3 with its next model, and if MiniMax can create a killer app that demonstrates the value of its technology, the narrative can shift quickly.
The shorts are on the right side of the current trend, but the trend is a micro-trend. The macro-trend is the AI revolution. And that revolution is just getting started.