Silence hangs between the spreadsheet and the truth. A leaked estimate from Menlo Ventures partner Deedy Das claims five Chinese AI startups—Zhipu, DeepSeek, Kling, Moonshot, MiniMax—collectively generated $2.6 billion in revenue last year. The numbers circulate like a hot rumor through WeChat groups and Telegram channels. Investors nod, relief flickering in their eyes. The Chinese AI narrative, they whisper, has crossed into commercial reality.
But I map the silence between the code and the chaos. And what I see is not a victory lap—it is a carefully constructed narrative bridge, built to carry capital across a chasm of unverified assumptions. This is not a revenue story. It is a story about stories.
The five companies sit atop China’s large model landscape: Zhipu (the state-backed giant), DeepSeek (the open-source cost killer), Kling (Kuaishou’s video generation arm), Moonshot (the long-context contender with Kimi), and MiniMax (the multimodal social play). Their business models diverge wildly. Zhipu leans on government contracts and enterprise deployments. DeepSeek relies on an ultra-low pricing API and an open-source community that rivals Meta’s Llama. Kling is embedded inside a video platform. Moonshot and MiniMax chase consumer attention with chat and entertainment.
The $2.6 billion figure, if accurate, would represent a remarkable leap from zero to scale in under two years. But accuracy is the first casualty of narrative warfare. Das’s estimate is not audited. It is not derived from public filings. It is a back-of-the-envelope projection from a venture capitalist whose firm, Menlo Ventures, has a vested interest in the AI funding narrative. The source is “Dongcha Beating”—an obscure outlet with no disclosure of methodology. The article itself reads like a press release dressed as journalism.
Here is the core insight the spreadsheets do not capture: Revenue is not the same as value. In crypto, we learned this lesson during the ICO summer. Token sale numbers looked staggering until the underlying projects burned through treasury. The same dynamic applies to these AI companies.
Consider DeepSeek’s $500 million revenue. To achieve that with API pricing at roughly 1/20th of OpenAI’s rates, the company must process an astronomical volume of tokens. At current inference costs for dense models, every dollar of revenue may cost $0.80 to $1.20 in compute alone. That is not a business. That is a compute subsidy disguised as growth. The narrative sells it as market share capture. The silence reveals a unit economics nightmare.
Zhipu’s $1 billion is more opaque. Government contracts often bundle AI with hardware, data center services, and consulting. The pure software revenue may be half that number. And government deals are lumpy—one lost bid can slash revenue by 30% overnight. The narrative of stable enterprise cash flow is a story of dependencies, not durable demand.
Kling’s $500 million faces a similar problem: most of it likely comes from internal transfer pricing inside Kuaishou, not external customer payments. When your parent company is both customer and distributor, revenue numbers are legible only within the group’s accounting fiction.
The contrarian angle is uncomfortable for the optimists: The $2.6 billion narrative is precisely the kind of narrative that precedes a correction. In my years mapping DeFi sentiment during the Summer of 2020, I watched identical patterns unfold. Projects would announce total value locked (TVL) figures that looked impressive until you realized 60% was wash trading or token incentives. The market celebrated the number, ignored the quality, and paid for it later.
Here, the structural weakness is the dependence on subsidized pricing and government tailwinds. When the subsidy ends—either because venture capital becomes scarce or because policy shifts—the revenue numbers will compress. The narrative will pivot from growth to survival. And the companies that survive will not be the ones with the highest top-line revenue, but those with the highest gross margin and most retentive customer base.
The narrative is the only immutable ledger. Right now, that ledger shows a credit entry for Chinese AI commercialization. But I see a debit column full of hidden costs: the cost of compute, the cost of customer churn, the cost of competing against Alibaba’s free model tier, the cost of regulatory uncertainty. The net balance is far redder than the headlines suggest.
What does this mean for crypto? The signal for us is not in the AI numbers themselves, but in the pattern of narrative construction. The same forces are at play in crypto AI tokens—projects touting user counts, transaction volumes, and total value. The narrative of AI-crypto convergence is being built on similar shaky foundations. Autonomous agent economies, decentralized compute networks, AI marketplace tokens—many are still in the pre-revenue stage, yet their market caps already price in billions.
Truth hides in the bear market’s quiet shadows. The bear market for AI hype may be closer than we think. When the Chinese AI revenue story unravels—as it inevitably will under the weight of unit economics—the spillover will affect the crypto AI sector. Not because the two are technically linked, but because narratives are contagious. The same investor enthusiasm that inflated Chinese AI revenue estimates is propelling crypto AI valuations. The same skepticism that will puncture one will puncture the other.
The takeaway is not despair. It is discernment. As narrative hunters, our job is to separate the signal of sustainable adoption from the noise of constructed hype. The $2.6 billion story is a construct. But within it lies a real signal: Chinese companies are beginning to deploy AI into production. The question is at what real cost, and with what real stickiness.
In the wild west, stories are the only compass. But a compass that points to a mirage leads only to thirst. I hunt for the story that the data cannot speak—the story of unit economics, of customer retention, of marginal revenue over compute cost. That is the narrative that will determine which AI companies survive the coming winter, and which crypto AI projects follow their fate.