BEIJING – China is transitioning its artificial intelligence sector toward a formalized “token economy,” repositioning the basic unit of AI processing as the primary settlement mechanism between technological providers and commercial users.
The National Data Administration, the country’s central data regulator, has designated tokens as the settlement unit linking technological supply with commercial demand, extending Beijing’s broader effort to treat data and computing power as nationally managed strategic resources under the framework set by the Data Security Law. This policy shift coincides with a massive scaling of domestic infrastructure; China now processes 140 trillion tokens daily, a sharp increase from 100 billion at the start of 2024.
The strategic pivot seeks to monetize AI capabilities while navigating severe hardware constraints and leveraging a significant advantage in power grid capacity. While U.S. developers focus on proprietary frontier models, Chinese firms are increasingly utilizing open-source strategies to capture market share across Southeast Asia and the Middle East. For policymakers in those regions, the transition from flat subscription pricing to token-based billing is likely to reshape procurement decisions, cloud budgets, and data-sovereignty negotiations with Chinese providers.
Corporate Restructuring and Model Competition
Large-cap technology firms are reorganizing their internal operations to align with this token-centric model and to qualify for state-backed computing subsidies. Alibaba has consolidated five separate entities-including its foundational research arm, Tongyi Laboratory, and its enterprise division, Wukong-into the Alibaba Token Hub (ATH), effectively turning compute, data and model access into a unified internal marketplace.
“ATH is built around a single organising mission: create tokens, deliver tokens and apply tokens,” said Alibaba CEO Eddie Wu.
The competitive environment is split between open-source and proprietary approaches. Alibaba’s Qwen models have gained global traction due to their low barrier to entry, with Meta’s Muse Spark model trained partly on Qwen. In contrast, ByteDance has maintained a proprietary ecosystem. Its Doubao chatbot recorded 100 million daily active users during the February 2026 Chinese New Year holiday, reinforcing the company’s direct-to-consumer model and its ability to test monetization schemes at national scale.
Tencent has integrated AI directly into its WeChat interface via ClawBot, launched in March 2026, allowing over one billion monthly active users to access OpenClaw services without leaving the super-app. That design effectively embeds the token economy inside everyday payments, government service portals and enterprise workflows that already run through WeChat, giving provincial authorities a ready-made channel to deploy local AI services.
Capital Markets and AI Economics
The surge in AI development has pushed IPO activity on the Hong Kong Stock Exchange to a five-year high, turning the city into a preferred venue for Chinese AI listings aimed at both domestic and international investors. Public listings for labs such as MiniMax and Zhipu AI have provided a window into the high-burn nature of frontier AI development and the extent to which these firms rely on discounted access to state-backed compute.
| Company | 2025 Revenue | Net Loss/Profit | Note |
|---|---|---|---|
| MiniMax | $79 million | ($250 million) | 70% of revenue from overseas |
| Zhipu AI | $104.8 million | ($680 million) | R&D spending increased 45% |
Despite these losses, market sentiment remains aggressive. Zhipu’s shares have risen over 570% from its IPO price, and MiniMax has increased over 470%. Moonshot AI, currently valued at $10 billion following a January funding round, is considering a Hong Kong listing, a move closely watched by regulators in both Hong Kong and Beijing as a test of investor appetite for companies whose cash flows are tied to volatile compute prices.
Infrastructure and Physical AI
China is leveraging its domestic manufacturing supply chain to lead in “physical AI,” particularly in humanoid robotics and autonomous transit, areas explicitly encouraged under national industrial policies. Unitree Robotics has filed for a 4.2 billion yuan ($610 million) IPO on Shanghai’s STAR Market, reporting an adjusted net profit of approximately 600 million yuan ($87 million). The Shanghai exchange has positioned STAR as a listing venue for strategically important high-tech firms, giving Unitree access to deep pools of onshore capital.
In the autonomous vehicle sector, Pony AI launched Europe’s first commercial robotaxi service in Zagreb, Croatia, in early April 2026. Simultaneously, WeRide has entered a partnership with Uber to deploy commercial robotaxis in Dubai. These cross-border deployments serve not only as technology demonstrations but also as templates for how Chinese firms negotiate safety standards, liability frameworks and data-transfer rules with foreign regulators.
This expansion is supported by a massive investment in energy infrastructure. Goldman Sachs estimates China will possess 400 gigawatts of spare power capacity by 2030, which is roughly three times the projected global demand for data centers. That excess, concentrated in coal-heavy and renewable-rich provinces, is being earmarked for data and AI “industrial parks,” giving officials a tool to steer both domestic and foreign AI workloads toward regions they want to develop.
Operational Constraints and Trade Barriers
The growth of the token economy faces systemic headwinds from U.S. export controls on high-end semiconductors. Chinese firms are increasingly reliant on Huawei hardware or home-designed chips, such as Alibaba’s Zhenwu chips, which powered a new data center unveiled on April 8, 2026. For corporate buyers, that shift raises questions about long-term interoperability with U.S.- and European-made systems and about exposure to future sanctions.
The funding gap between Silicon Valley and Beijing remains stark. While Moonshot AI relies primarily on domestic capital, U.S.-based Anthropic raised $30 billion in February 2026, reaching a $380 billion post-money valuation. The disparity underscores how much of China’s AI buildout still depends on policy-driven financing rather than purely private market returns.
These pressures have led some founders to relocate. Manus AI reincorporated in Singapore before being acquired by Meta for $2 billion in late 2025. This transaction resulted in regulatory friction, with CEO Xiao Hong and chief scientist Ji Yichao currently subject to exit bans. The case has become a cautionary example for Chinese AI entrepreneurs and global investors attempting to structure cross-border deals without triggering capital-control or national-security concerns.
Capital Expenditure and Monetization
The cost of maintaining AI competitiveness is placing significant pressure on corporate balance sheets and, increasingly, on public policy as governments weigh the systemic risks of concentrated compute spending.
- Alphabet: $94 billion (2025 Capex)
- Meta: $75 billion (2025 Capex)
- ByteDance: $23 billion (projected AI infrastructure spend)
- Alibaba: 123 billion yuan ($17 billion) (2025 Capex), contributing to a 66% plunge in net income
- Tencent: 79 billion yuan ($11.6 billion) (2025 Capex)
To offset these costs, firms are shifting away from free access. Alibaba and Z.ai have transitioned some recent models to closed formats, while Baidu and Zhipu are increasing prices for cloud services and model access. For enterprise and government customers, the move from experimentation-era freebies to metered access is forcing a reprioritization of which AI use cases can demonstrate measurable returns.
The Chinese government continues to subsidize “one-person companies” specializing in AI agents, positioning them as a way to diffuse innovation beyond the country’s largest platforms, while simultaneously proposing regulations for AI companion apps and issuing warnings regarding security vulnerabilities in OpenClaw-based agents. Combined with tighter content and safety rules under the country’s interim measures for generative AI services, the emerging token economy is being built inside a dense layer of algorithmic accountability-one that foreign regulators will increasingly have to understand as Chinese AI systems and pricing models move across borders.
