Hong Kong AI New Narrative: Embracing Token Economics

In early spring 2026, Hong Kong’s AI sector has drawn intense global capital attention as valuation frameworks undergo a structural shift. At midday on March 24, the Hang Seng Tech Index stood at 4,779.52 points, up 1.42%, while the Hang Seng Stock Connect Artificial Intelligence Theme Index has climbed roughly 15% year‑to‑date, indicating accelerating capital allocation into the AI theme.

Domestic large‑model developers MiniMax and Zhipu have been standout performers in this wave. By midday on March 24, MiniMax traded at HKD 999, up 9%, with a market capitalization of HKD 313.322 billion, and Zhipu traded at HKD 627, up 6.27%, with a market capitalization of HKD 279.544 billion. Since listing, both stocks have risen more than 400% year‑to‑date.

A coherent valuation thesis is emerging: as intelligent agents such as OpenClaw catalyze application deployment, token consumption per task is rising exponentially. Tokens—the smallest units processed by large models—are increasingly treated as the principal metric for assessing how deeply AI solutions are being adopted and their commercialization potential. Investors and analysts in Hong Kong are recalibrating valuation measures to reflect this shift.

The competitive landscape is being reconfigured. Although the Hang Seng Tech Index remains weighted toward traditional internet and software companies, large models are becoming new gateways for user engagement and value creation, encroaching on the user time and monetization space of incumbent platforms. In January 2026, MiniMax and Zhipu became among the first publicly listed large‑model developers globally, and their market capitalizations now rival those of Baidu and JD.com. . Market observers note that these listings allow investors to gain direct exposure to China’s AI advancement without the legacy business exposures typical of the major tech incumbents.

Analysts project that MiniMax and Zhipu will each deliver year‑on‑year revenue growth exceeding 150%, far outpacing the single‑digit growth expected for China’s largest tech firms and the low double‑digit growth seen among major U.S. peers. Incumbent giants such as Tencent, Alibaba, and Baidu have responded by increasing capital expenditure to secure AI entry points, investing heavily in compute infrastructure and model development. Tencent launched WorkBuddy, an AI agent compatible with OpenClaw skills; Alibaba Cloud introduced dedicated image services; and JD Cloud initiated pilots across retail and logistics. These moves quickly influenced market sentiment: Tencent’s shares jumped 7.27% on March 10 following WorkBuddy’s launch, Alibaba’s stock recovered steadily amid token‑driven growth in its MaaS business, and JD.com  posted consecutive gains from March 12 to 16 as pilot news emerged.

The transition from model competition to scenario positioning underpins the token‑centric valuation shift. At Nvidia’s GTC in March 2026, CEO Jensen Huang described data centers as “token factories,” reframing them as continuous production lines that convert power and data into tokens. Tokens have thus been cast as a form of hard currency in the digital economy, with token generation efficiency directly influencing a technology company’s revenue trajectory.

For investors, tokens provide a unified, quantifiable scale linking an AI company’s revenue and cost structures. Valuation analysis increasingly emphasizes token consumption volumes, the conversion efficiency from tokens to revenue, and unit token profitability rather than relying solely on traditional profit metrics. MiniMax’s M2 series drove daily token consumption more than sixfold from December 2025, lifting 2025 revenue 158.9% to USD 79.04 million and expanding gross margin from 12.2% to 25.4%. Zhipu’s GLM‑5 flagship model supported daily token consumption of 4.2 trillion by November 2025, fueling multiple‑fold revenue growth; its MaaS platform has aggregated over 3 million enterprises and developers, and API pricing rose 83% cumulatively in the first quarter of 2026. In both cases, token call volumes have become direct revenue indicators, enabling a shift from concept‑driven to revenue‑driven valuations.

Third‑party data show Chinese models gaining traction: by late February 2026, API token usage for Chinese models on OpenRouter exceeded that of U.S. models. A recent Goldman Sachs report cited narrowing performance gaps, support for ultra‑long context windows up to one million tokens, and token pricing at roughly 5%–10% of U.S. flagship models, highlighting China’s cost‑performance advantage across multimodal tasks. Bloomberg research attributes the rapid market re‑rating of MiniMax and Zhipu to investor recognition of their “low token cost plus high demand growth” profile, which has propelled their market capitalizations toward parity with established tech names.

Market participants should recognize that valuation anchors in a fast‑evolving industry are inherently transient: today’s token metric could be supplanted by new measures tomorrow, and current leaders may be displaced by cross‑sector entrants. What is clear is that Hong Kong is emerging as a central venue for connecting China’s AI ecosystem with global capital, and the sector’s narrative is only beginning to unfold.