Southern Finance 21st Century Economic Report Journalist Yuan Sijie Hong Kong Report
In early spring 2026, the Hong Kong stock market's AI sector is becoming the focus of global capital with a reconstruction of valuation logic. As of the midday on March 24, the Hang Seng TECH Index reported 4779.52 points, up 1.42%. Notably, the Hang Seng Hong Kong Stock Connect Artificial Intelligence Theme Index, which reflects the overall performance of the AI industry, has risen about 15% this year, indicating that funds are accelerating their accumulation in the AI sector.
Amid this wave, the domestic AI large model &#34;Gemini&#34; in the Hong Kong stock market has performed particularly well. As of the midday on March 24, MiniMax's latest stock price was HKD 999, up 9%, with a total market capitalization of approximately HKD 313.322 billion; while Zhipu's latest stock price reached HKD 627, up 6.27%, with a total market capitalization of HKD 279.544 billion. Since their listing, both companies' stock prices have increased by over 400% this year.
Behind the market's enthusiasm, a clear valuation theme is emerging: as AI agents like OpenClaw ignite the &#34;application landing&#34; wave, the Token consumption from single tasks is growing exponentially. As the smallest unit of information processed by AI large models, Tokens are becoming the core pricing unit for measuring the depth of AI business implementation and commercialization potential.
In this competitive landscape reshaped by the wave of AI large models, Hong Kong stock investors and analysts are also seeking a new &#34;ruler&#34; to measure value.
Reshaping the Competitive Landscape
The change in valuation anchors for the Hong Kong stock AI sector began with shifts in the competitive landscape of the Hong Kong stock market.
Market analysis suggests that the current Hang Seng TECH Index still centers around traditional internet and software companies, but in this round of AI revolution, large models are becoming a new traffic entry point and value center, continuously eroding the user time and commercial space of traditional platforms and software.
In January 2026, MiniMax and Zhipu became among the first global companies to develop and list large models. After several-fold increases in stock prices, the market capitalization of the domestic AI large model &#34;Gemini&#34; has reached parity with Baidu and is on par with JD.com.
&#34;The listings of technology companies like MiniMax and Zhipu are expected to allow investors to directly participate in the development of AI in China, while also freeing them from the limitations of passively bearing exposure to traditional businesses like e-commerce and food delivery when investing in the eight major tech giants,&#34; said Chen Mingkang, a senior equity strategist at Bloomberg Industry Research. He noted that the large language models developed by these two companies rank among the top globally, and their products demonstrate strength in performance and cost-effectiveness.
Chen Mingkang believes that the several-fold increase in stock prices of the domestic AI large model &#34;Gemini&#34; since their listing may reflect the market's optimistic expectations for early adoption prospects and the demand for high-performance and cost-effective models.
The market expects that both MiniMax and Zhipu will see year-on-year revenue growth exceeding 150%, far surpassing the single-digit high growth levels of China's eight major tech giants and the low double-digit growth levels of the seven major tech giants in the United States.
If the story of the new AI forces is &#34;from 0 to 1,&#34; then the traditional internet giants are performing a leap &#34;from 1 to 10.&#34;
Since the beginning of this year, the capital expenditure plans of established tech giants like Tencent, Alibaba, and Baidu have frequently attracted market attention. To compete for AI entry points, these cash-rich players are investing heavily in computing infrastructure and model development
&#34;Old players have encountered new opportunities, increasing capital expenditure, which must indicate optimism about the development of the sector. Using new business models to complement existing ones for geometric growth is a model that traditional internet giants are either proactively or necessarily developing,&#34; pointed out Wang Xinjie, Chief Investment Strategist of Standard Chartered China Wealth Solutions. In the current industry business models, various application scenarios are bound to emerge. For internet giants, increasing capital expenditure and then quickly monetizing is the current direction of development.
Chen Mingkang predicts that cloud service giants control the full-stack AI infrastructure and may launch service bundling strategies, forming advantages in pricing and profit margins, which may further intensify industry competition in the future.
With AI agents like OpenClaw igniting the wave of &#34;application landing,&#34; internet giants are accelerating their shift from &#34;model competition&#34; to &#34;scenario positioning.&#34; Tencent quickly launched the all-scenario AI agent WorkBuddy compatible with OpenClaw skills; Alibaba Cloud launched exclusive image services; JD Cloud is piloting in retail, logistics, and other business lines—these old players with vast user ecosystems are attempting to open new growth spaces through OpenClaw.
This collective embrace of OpenClaw quickly generated feedback in the capital market. Tencent became one of the biggest winners; on March 10, the day after WorkBuddy's launch, the stock price surged at the opening, reaching a peak of HKD 556 during the day, with an overall increase of 7.27%, marking the largest single-day increase in nearly three months, and the trading volume exceeded HKD 30 billion, completely igniting market expectations for its AI commercialization; Alibaba, which owns Alibaba Cloud, did not experience a single-day surge, but its historical highest growth rate due to the surge in Token consumption from its refined MaaS business led to a steady recovery of its stock price in mid-March during the tech stock rebound, confirming the market's recognition of its Token business layout; after the news of JD Group's pilot in retail and logistics scenarios was released, its stock price also broke free from the previous sideways trend, rising for five consecutive trading days from March 12 to 16, with the Hong Kong stock price gradually recovering from HKD 108.6 to HKD 111.5.
As OpenClaw pushes AI from &#34;Q&amp;A interaction&#34; to &#34;task execution,&#34; the Token consumption from single tasks shows exponential growth. Tokens are no longer just a technical measure of computing power but have become the core pricing unit for measuring the depth of AI business landing and commercialization potential, which is the key logical starting point for the migration of AI valuation anchors to Tokens.
&#34;Hard Currency&#34; in the Digital Economy Era
In March 2026, at the NVIDIA GTC conference, the company's founder and CEO Jensen Huang proposed a new concept to reshape industry perception—&#34;Token Factory.&#34; In his view, data centers are undergoing a fundamental role transformation: no longer the &#34;electronic warehouses&#34; used for storing files and data in the past, but rather an intelligent production line operating day and night. This factory inputs electricity and data, producing Tokens.
As Jensen Huang stated, Tokens have become the &#34;hard currency&#34; of the digital economy era, and their generation efficiency will directly determine the survival capability and revenue curve of technology companies
What is a Token?
In simple terms, a Token is the smallest unit of measurement for each interaction between the user and the model: every word and punctuation mark you input to the AI, as well as every character the model replies to you, will be broken down and consume a certain number of Tokens.
For investors, Tokens have become a focal point because they unify the revenue and cost sides of AI companies into the same quantifiable scale for the first time.
&#34;For the AI sector, traditional PE valuations are actually not very effective. For this type of potentially high-growth industry, investors mainly focus on its future potential profitability,&#34; Wang Xinjie told reporters.
Wang Xinjie pointed out that in the current layout phase, a large amount of capital expenditure may make it difficult for these companies to achieve immediate profitability. Therefore, in the AI industry, current user traffic and Token consumption are what investors are more concerned about, as this can enhance the probability of future profitability.
For example, when analysts issue valuation reports for AI companies, they are no longer just focusing on traditional income statements but are trying to estimate the company's share of Token consumption in the global large model subscription and API market, as well as the gross profit that can be achieved per Token.
Behind this shift in valuation logic is the market's re-understanding of the essence of AI companies' business models: the growth in Token consumption indicates an increase in user activity and deeper business penetration, while the improvement in the monetization capability of each Token represents that the business model is maturing.
Specifically, native large model companies win high valuations based on their technological foundation because investors believe their Token consumption is likely to achieve exponential growth; while traditional platform giants with vast scenarios are endowed with greater monetization imagination because existing businesses like e-commerce recommendations, food delivery scheduling, and content distribution naturally have channels to convert Token usage into actual revenue.
Yang Delong, chief economist of Qianhai Kaiyuan Fund, pointed out to reporters that Hong Kong stocks in the AI era pay more attention to new indicators like customer traffic. This change in valuation logic will also have a significant impact on the valuation pricing of Hong Kong tech stocks, as investors will place greater importance on tech companies' advantages in Tokens and big data.
When Tokens Become the New Unit of Valuation
In this trend, Tokens are becoming a new valuation anchor for AI companies. Song Weiwei, manager of the China Europe CSI Robotics Index Fund, pointed out to reporters that OpenClaw represents a paradigm shift from &#34;dialogue&#34; to &#34;execution,&#34; and Token consumption is the first principle of the entire investment chain.
The business model of pure AI companies highly relies on Token consumption, and the underlying logic of the directions likely to benefit is the transmission of this multiplier effect.
The commercial data of Hong Kong's AI &#34;Twin Stars&#34; is the best evidence: MiniMax, driven by explosive growth in its M2 series text models, saw its daily Token consumption surge more than sixfold compared to December 2025, pushing its full-year revenue for 2025 to skyrocket by 158.9% year-on-year to $79.04 million, with gross margin jumping from 12.2% to 25.4%, demonstrating exceptional commercialization efficiency
Zhipu is relying on the rapid commercialization of its GLM-5 flagship model API services, with token consumption continuously rising—by November 2025, the daily average token consumption had reached 4.2 trillion, driving revenue to achieve exponential growth in 2025. Its MaaS platform has gathered over 3 million enterprises and application developers, with API prices cumulatively increasing by 83% in the first quarter of 2026, showing a clear trend of both volume and price growth.
The common characteristic of these two companies is that token usage has become a direct revenue indicator, with performance realization driving the transition from &#34;concept-driven&#34; to &#34;revenue-driven&#34; valuations.
Duan Nairong, a senior investment strategist at Royal Bank of Canada Wealth Management, told reporters that most &#34;new AI forces&#34; are currently not profitable, and the market is more focused on their dynamic growth potential.
Duan believes that under the backdrop of &#34;token inflation,&#34; the value anchors of Hong Kong tech stocks may refer to the following dimensions: first, the conversion efficiency of token consumption to revenue, that is, how much revenue a unit of token can generate, which determines the true profitability of the business; second, the monetization rate of traffic and customer stickiness, including customer retention rates, paid renewal rates, and other indicators, which reflect the sustainability of revenue; third, the locking ability of prepaid systems, that is, the extent to which enterprises can lock in future revenue through prepaid and subscription models, which can better reflect the stability of the business.
It is worth noting that the performance and cost-effectiveness advantages of domestic AI models are supporting token competitiveness, while token pricing and cost advantages are also one of the foundations for building valuation premiums.
Data from the third-party API aggregation platform OpenRouter shows that in late February 2026, the usage of API tokens using Chinese AI models surpassed that of American AI models for the first time.
A recent research report from Goldman Sachs obtained by reporters pointed out that the performance gap between the new generation of Chinese AI models and American models has significantly narrowed, supporting ultra-long context windows with up to 1 million tokens and powerful agent functions. Additionally, the token pricing of Chinese AI models is only 5%-10% of that of American flagship models, highlighting their cost-effectiveness. Finally, Chinese AI models perform excellently in multi-modal fields such as text, video, and audio, further expanding token application scenarios.
In the report, Goldman Sachs analyzes that the surge in token demand will bring revenue breakthroughs for Chinese AI model companies while driving growth in cloud service revenues.
In terms of competitive landscape, independent AI model companies are expected to rise rapidly through token business, competing with internet giants; leading internet companies will also consolidate their advantages in token-related infrastructure through capital expenditure investments.
Bloomberg Industry Research reports that the stock prices of companies like MiniMax and Zhipu AI have skyrocketed, essentially reflecting the market's valuation recognition of their &#34;low token cost + high demand growth rate&#34; combination, allowing their market capitalization to quickly rival that of tech giants like JD.com and Baidu.
For investors, it may be necessary to accept a reality: in the rapidly iterating AI industry, there may not be an eternal value anchor. Today's tokens may be replaced by new measurement units tomorrow; today's leaders may be disrupted by cross-industry players tomorrow
It can be confirmed that Hong Kong stocks are becoming the core stage for the integration of China's AI industry and global capital. On this stage, the story has just begun.
(Zhang Weize also contributed to this article)
<img src="https://imageproxy.pbkrs.com/https://inews.gtimg.com/om_bt/OyeZiIcYxrVDv0ww2NnoLMJQ9eqtE4vIQDkZFFFmqA4vMAA/641?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://inews.gtimg.com/om_bt/OyeZiIcYxrVDv0ww2NnoLMJQ9eqtE4vIQDkZFFFmqA4vMAA/641"/>

STECH

03413

513130

09140

513180

83140

The Hong Kong stock market's AI sector is experiencing a valuation reconstruction in early spring 2026, with the Hang Seng TECH Index rising by 1.42%. The AI theme index has increased by approximately 15% this year, with accelerated capital inflow. Domestic AI large model companies MiniMax and Zhipu have seen their stock prices rise by 9% and 6.27%, respectively, with year-to-date gains exceeding 400%. Token economics has become the core pricing unit for the commercialization of AI businesses, reshaping the market competition landscape and posing challenges for traditional internet companies. Investors can directly participate in China's AI development, freeing themselves from the constraints of traditional businesses

腾讯新闻 - 财经

Hang Seng TECH Index

A DOO AI

Huatai-PB CSOP Hang Seng Technology ETF(QDII)

CAM HKUS AI-U

ChinaAMC Hang Seng Technology ETF(QDII)

CAM HKUS AI-R

Central Observation | Hong Kong Stocks AI New Narrative: Embracing "Token Economics"