Check the order flow. Over the past 30 days, short interest against Zhipu AI and MiniMax has hit an all-time high. Not a blip. Not a hedge. A record pile-up of bearish positions against two of China's most prominent AI startups. The narrative on the street is simple: the AI price war is eating margins, and these two are the next casualties. I don't trade narratives. I trade the logs. And the logs here tell a story that goes deeper than a few hedge funds betting on a crash.
Let me be clear about what we're looking at. Zhipu AI, the Beijing-based lab behind the GLM series, and MiniMax, the Shanghai outfit known for its conversational models and C-end applications, are both facing unprecedented bearish sentiment. The catalyst is the ongoing price war in China's large language model market — a brutal, margin-destroying contest where API costs have collapsed by over 90% in some tiers since early 2024. The market's verdict? These companies can't survive the squeeze. The shorts are betting on a death spiral: falling revenue, rising costs, and a valuation correction that will make the current funding rounds look like charity.
But here's the thing about short positions: they're a lagging indicator of sentiment, not a leading indicator of fundamentals. What the market is actually telling us is that investors have stopped believing in the 'AI miracle' narrative and started demanding something far more uncomfortable: unit economics. And that's where the real analysis begins.
The Price War Is a Feature, Not a Bug
Let's break down the market structure. China's LLM space is a battlefield with three tiers. Tier one: the hyperscalers — Baidu, Alibaba, Tencent, ByteDance. These players can subsidize API costs indefinitely because AI is a loss leader for their cloud businesses. Tier two: the well-funded independents — Zhipu, MiniMax, Moonshot, DeepSeek. These companies have raised billions, but they don't have a cloud business to cross-subsidize their AI spend. Tier three: the long tail of smaller labs that are already dead but haven't stopped breathing yet.
The shorts are betting that tier two gets squeezed between tier one's pricing power and tier three's desperation. And on the surface, the logic holds. Zhipu's GLM-4 and MiniMax's abab6.5 are competitive models, but they're not so far ahead that they can command a premium. When ByteDance drops the price of its Doubao API by 80%, Zhipu has to follow or lose market share. That's the trap. That's the price war.
But here's what the shorts are missing: the price war is not a sign of weakness. It's a sign of commoditization. And commoditization is the natural endgame for any technology that reaches scale. The question isn't whether prices will keep falling — they will. The question is who can survive the fall and emerge with a moat that isn't based on price.
Reading the On-Chain Signals — Or the Lack Thereof
Now, I'm a blockchain guy. I watch the blockchain, not the ticker. So when I look at Zhipu and MiniMax, I'm looking for the crypto equivalent of their situation: a protocol with strong fundamentals but a token price that's being hammered by short-term sentiment. The parallel is uncanny.
In crypto, we see this all the time. A project with real usage, real revenue, and a real community gets shorted into oblivion because the market is fixated on a single metric — usually the token price. Then, when the shorts are exhausted, the price snaps back. The same dynamic is playing out here, but with a twist: Zhipu and MiniMax aren't tokens. They're equity. And equity has a different risk profile.
Let me give you a concrete example from my own playbook. In 2021, I was tracking a DeFi protocol that had a similar setup — strong fundamentals, but a token that was being shorted relentlessly. The shorts were convinced the protocol's yield model was unsustainable. They were right about the model, but wrong about the timeline. The protocol survived long enough to pivot, and the shorts got liquidated. The lesson? Shorts are often right about the problem but wrong about the timing.
That's what I see here. The shorts are right that Zhipu and MiniMax face a margin squeeze. They're wrong that this is a death sentence. Here's why.
The Hidden Variables the Market Is Ignoring
First, let's talk about the enterprise channel. Zhipu has been quietly building a government and enterprise business that's largely invisible to the public markets. In China, state-owned enterprises and government agencies are mandated to use domestic AI solutions. Zhipu's GLM series is one of the few models that meets the compliance requirements. This is a captive market with sticky contracts and high switching costs. It's not the kind of business that gets disrupted by a price war because the buyers aren't price-sensitive — they're compliance-sensitive.
Second, MiniMax has a consumer angle that's underappreciated. Their C-end applications, particularly in the short-video and interactive entertainment space, generate real usage data. That data is a moat. Every interaction trains their models, improves their recommendations, and deepens their understanding of user behavior. In the AI world, data is the ultimate barrier to entry. The shorts are pricing MiniMax as a pure API play, but they're actually a data company with an API on top.
Third, and this is the contrarian angle that most analysts are missing: the price war is actually a consolidation tool. When prices collapse, the weak players die. The strong players — the ones with cash reserves, enterprise contracts, and data moats — absorb the market share. Zhipu and MiniMax are in the strong category. They've raised enough capital to survive a two-year price war. Their competitors haven't. The shorts are betting on a bloodbath, but they're betting on the wrong side of the carnage.
The Real Risk Isn't the Price War — It's the Regulatory Overhang
Let me pivot to the risk that nobody's talking about. The SEC's approach to crypto has been regulation-by-enforcement — deliberately withholding clear rules while punishing individual actors. China's approach to AI is different, but the effect is similar: uncertainty. The Chinese government has been tightening its grip on AI models, requiring approvals for public releases and imposing content restrictions. This regulatory overhang is a bigger threat to Zhipu and MiniMax than any price war.
Why? Because regulation creates compliance costs. And compliance costs are fixed costs. In a price war, fixed costs become a death sentence. If Zhipu has to spend millions on compliance infrastructure while simultaneously cutting API prices, that's a double squeeze. The shorts might not be explicitly pricing this in, but it's the hidden variable that could turn a manageable margin squeeze into a full-blown crisis.
I've seen this play out in crypto. Smart contracts don't care about your feelings, but they do care about gas costs. When gas prices spike, marginal protocols die. The same logic applies here: when compliance costs spike, marginal AI companies die. Zhipu and MiniMax are not marginal — yet. But if the regulatory burden increases faster than their revenue, they could become marginal very quickly.
The Death Spiral Scenario — And Why It's Unlikely
Let me walk through the bear case, because it's not entirely wrong. The death spiral goes like this: price war → revenue decline → valuation markdown → difficulty raising new capital → forced cost-cutting → talent exodus → product quality decline → further revenue decline. It's a vicious cycle, and it's killed more than a few crypto projects.
The key variable is the funding environment. Zhipu and MiniMax have both raised significant rounds — Zhipu's valuation was reportedly around $2 billion in its last round, and MiniMax has raised over $1 billion. But if the shorts are right and the market turns against them, the next round will be a down round. That's when the spiral accelerates.
Here's why I think the spiral doesn't happen: the Chinese government has a strategic interest in maintaining a competitive domestic AI ecosystem. They're not going to let their top two independent AI labs collapse. There's a reason the state has been funneling capital into these companies through various vehicles. It's not just about technology — it's about national security and geopolitical positioning. The government will step in if the market turns too hostile. That's the backstop that the shorts are ignoring.
The Takeaway: Position for the Snap-Back, Not the Collapse
So where does this leave us? The record short positions are a signal, but they're a signal of sentiment, not of fundamentals. The market is anxious about the price war, and that anxiety is understandable. But anxiety is not analysis. The fundamentals — enterprise contracts, data moats, government backing — are still intact. The shorts are betting on a collapse that the structural dynamics don't support.
Code is law, but human greed is the bug. The greed here is on the short side — the belief that a price war is a death sentence rather than a consolidation tool. The smart money is watching the fundamentals, not the ticker. And the fundamentals say these companies survive.
My advice? Don't chase the short narrative. Watch the funding rounds. Watch the enterprise contract announcements. Watch the regulatory signals. If Zhipu or MiniMax announce a major government contract in the next quarter, the shorts are going to get squeezed. If they announce a down round, the shorts are right. The data will tell you which way it goes. I don't trade on hope. I trade on logs. And the logs are still pointing to survival.
The next 90 days will be decisive. The shorts have made their move. Now we watch to see if the fundamentals hold. I'm not betting against the fundamentals. Not yet.