Zhipu's HK$1 Trillion Valuation Casts an Existential Shadow Over China's Internet Giants
A loss-making AI startup founded less than seven years ago has surpassed the market capitalizations of several of China's most profitable internet companies, forcing investors to reprice the structural durability of an entire generation of business models.
On June 22, 2026, Zhipu AI crossed the HK$1 trillion (approximately US$128 billion) market capitalization threshold on the Hong Kong exchange — a milestone that arrived not with fanfare from the company's own financials, but as a verdict on the incumbents surrounding it. On the same trading day, Meituan was valued at HK$444.5 billion, JD.com at HK$302.9 billion, Kuaishou at HK$196.9 billion, and Trip.com Group at HK$229.3 billion. Zhipu, a company generating just over RMB 700 million (approximately US$97 million) in annual revenue while still burning cash, had leapfrogged all of them.
The market's message was neither subtle nor accidental. Capital is not pricing Zhipu's present-day cash flows — it is pricing the probability that large language models will systematically cannibalize the most profitable revenue pools that China's internet economy has spent a decade constructing.
Valuation Gap Exposes Structural Discount on Legacy Moats
The arithmetic is jarring by any conventional metric. Tencent posted net profit exceeding RMB 200 billion (approximately US$27.8 billion) last year, yet trades at a price-to-earnings ratio below 15x. Alibaba Group is valued at roughly HK$1.97 trillion — a figure that analysts note sits at or below the sum of its Ant Group equity stake and its core e-commerce business, implying the market assigns near-zero terminal value to its operating franchise. Meanwhile, Zhipu — with no profits, limited revenue, and a product roadmap still under construction — commands HK$1 trillion.This is not irrational exuberance in isolation. It is a relative pricing signal: the market is applying a structural discount to internet incumbents' existing assets while assigning an option premium to AI-native challengers. The discount reflects a specific thesis — that the three widest moats in Chinese internet history are being outflanked, not frontally attacked.
Tencent's WeChat ecosystem monetizes through advertising tied to user attention on Moments and Official Accounts. ByteDance's Douyin extracts value through behavioral recommendation advertising. Alibaba's Taobao and Tmall platforms profit from information asymmetry embedded in the search-compare-purchase funnel. All three monetization architectures share a common vulnerability: they depend on users actively navigating platforms. An AI agent that completes tasks on behalf of users — booking flights, drafting reports, selecting and purchasing products — compresses those navigation layers into backend API calls, stripping away the advertising surface entirely.
Three Giants, Three Distinct Structural Vulnerabilities Emerge
Tencent's core identity as a "connector" faces obsolescence pressure. WeChat's strategic value has never been advertising per se — it has been the connective tissue linking people, content, and services. Mini Programs and Official Accounts are monetization derivatives of that connectivity. The shift from "connection" to "agency" — where users instruct an AI assistant rather than open individual applications — inserts a new intermediary between Tencent and its users. The stronger that intermediary becomes, the more peripheral WeChat's role in the daily task-completion stack. For a company whose mission is connectivity, a world in which connectivity is commoditized by AI agents poses a question that transcends quarterly earnings: what does Tencent become when its founding purpose loses scarcity value?
Alibaba is running two divergent growth curves simultaneously. Alibaba Cloud stands to benefit directly from large model proliferation — more Qwen API calls means more compute demand, and cloud infrastructure revenue has a clear upward trajectory in an AI-intensive economy. But Alibaba's advertising revenue from Taobao and Tmall faces the inverse dynamic: as AI assistants perform product selection and price comparison on behalf of consumers, the information asymmetry that justified merchant advertising spend erodes. These two curves are not correlated. The critical strategic question is whether Alibaba Cloud can scale fast enough to replace e-commerce advertising revenue before that revenue line hits its ceiling — a race against the company's own business model.
ByteDance's dilemma is the most concealed of the three. Douyin's recommendation engine is built on behavioral data granularity — watch time, scroll depth, interaction patterns — accumulated across billions of user sessions. Large language models are shifting recommendation logic from behavioral pattern matching to semantic understanding: a user saying "I'm in a bad mood today" conveys more actionable signal than a thousand passive scroll behaviors. If semantic understanding renders behavioral data walls structurally irrelevant, ByteDance's most defensible asset — its proprietary behavioral data corpus — loses its competitive premium overnight. ByteDance's Doubao series has performed well in benchmark evaluations, but internal deployment remains constrained to an "assist recommendation" role, insulated from the core commercial logic. That constraint is not a failure of ambition; it is a function of corporate DNA. A company built on behavioral matching cannot easily transplant its operating philosophy into semantic intelligence without dismantling the engine that generates the majority of its revenue.
"No Baggage" Becomes Zhipu's Most Valuable Asset
The common thread across all three incumbents is the innovator's dilemma operating at compressed speed. Kodak invented the digital camera and locked it away. Nokia built early touchscreen prototypes and protected the physical keyboard. Each technology transition in commercial history has forced incumbents into the same painful oscillation between protecting existing profit pools and embracing the new paradigm. What distinguishes the current AI transition is the velocity of compression: from GPT-3 to trillion-parameter frontier models took under three years. Kodak had roughly two decades between the invention of digital photography and its 2012 bankruptcy filing. The window for managed transition is narrowing at a rate that legacy organizational structures were not designed to accommodate.
Zhipu's trillion-dollar valuation is, in this context, a function of absence rather than presence. The company carries no advertising revenue to protect, no e-commerce platform interests to preserve, no recommendation algorithm moat to defend. Its entire resource allocation — model capability advancement, developer ecosystem construction, API commercialization — points in a single direction without internal conflict. This is the OpenAI (OpenAI) and Anthropic (Anthropic) playbook applied to the Chinese market: establish a model capability lead, attract developers, build an ecosystem, and let commercial architecture emerge from adoption rather than imposing it top-down.
Whether that playbook translates in China remains unresolved. To sustain a HK$1 trillion valuation on RMB 700 million in annual revenue, Zhipu will need to demonstrate at minimum three things over the next two to three years: sustained model performance within the global first tier; rapid scaling of API and enterprise service revenue; and a credible path from high R&D expenditure to improved gross margins and recurring revenue. Absent those proof points, the current valuation is a forward option, not a discounted cash flow.
Capital Markets Signal a Paradigm Repricing, Not a Bubble
The "greater fool" framing — that Zhipu's valuation is purely speculative momentum — misses the more consequential signal. Even if Zhipu's specific valuation proves unsustainable, the relative pricing between AI-native challengers and internet incumbents encodes a structural judgment: the next generation of dominant business models will not organically emerge from existing internet profit pools. They require companies unencumbered by legacy revenue dependencies, building on new architectural foundations.
China's internet giants are not standing still. Tencent, Alibaba, and ByteDance have each committed substantial capital to large model development. But capital commitment and strategic freedom are not the same variable. The incumbents' AI initiatives operate within organizational ecosystems where the most profitable existing business lines function as implicit veto players over any initiative that threatens their revenue. That constraint does not disappear with investment announcements.
The HK$1 trillion valuation assigned to Zhipu on June 22, 2026, is less a celebration of one company's prospects than a collective market verdict on the durability of an era. The coordinates that defined value creation in Chinese internet for the past fifteen years are losing their calibration. A new map is being drawn, and the companies holding the pen are not the ones who drew the last one.
Related Coverage:
Zhipu AI Surges 1,900% as GLM-5.2 Challenges Closed-Source Frontier