Investors looking to uncover promising thematic opportunities rely on authoritative, professional, timely and comprehensive analyst research. Recently, a seminar titled "Insights into Beijing's Trillion-Yuan New Opportunities in the Token Economy" was held at the Beijing International Big Data Exchange, where the "2026 Beijing Token Economy Development Observation Report" (hereafter "the Report") was released on site. Drawing on frontline visits and research by reporters, the Report integrates policy direction, industrial capital and market participant perspectives to offer useful reference for the better development of Beijing's token economy and the creation of "Beijing experience."
Market observation: token-based billing makes AI a universal digital tool
As the artificial intelligence (AI) industry iterates rapidly, domestic token call volumes are showing exponential explosive growth. Data from the National Data Administration shows that in early 2024, average daily token calls in China stood at 100 billion, and by June 2026 this figure had surpassed 500 trillion. Behind this surge in call volume, a new business model based on token billing is accelerating into shape. Token was originally an English term in the AI field. In March 2026, the National Data Administration designated "ciyuan" as the Chinese name for Token, marking the increasingly standardized, localized and mature development of China's AI sector. A token is the smallest computing unit when a large model processes information such as text, code, images, audio and video, and it is increasingly becoming the unit of measurement, settlement and statistics for AI services. Billing AI services by token is like paying for internet access by data usage or paying electricity bills by kilowatt-hours, allowing data value, computing power and model quality to be quantified. Research found that tokens are packaging complex computing power, models and data into standardized services, with AI capabilities supplied on demand like water and electricity. Small and medium-sized enterprises, micro-operators and even one-person companies (OPCs) no longer need to build their own computing hardware, and can access top-tier large model capabilities simply by purchasing token services. AI has sunk from a technical patent of a few major players to a universal digital tool accessible to society as a whole. Beijing concentrates more than half of the nation's token production and resource calls, which brings major industrial opportunities while also imposing entirely new requirements on token production, circulation and application, urgently necessitating stronger top-level design and systematic cultivation of the new token economy track.
First-mover advantage: building token factories covering the full chain from production to distribution
As a pathfinder in token economy development, Beijing has issued the country's first provincial-level special policy on the token economy, the "Ten Measures for the Token Economy," systematically building up Beijing's token economy industrial development system while creating a competition arena and experimental field for the industry. Beijing has also written tokens into its "15th Five-Year Plan." The "Beijing 15th Five-Year Plan for High-Tech Industry Development" proposes enriching model service supply capacity, building a token settlement system, and creating token factories covering the full chain of token production, distribution and consumption. The "Beijing 15th Five-Year Plan for Digital Economy Development" mentions expanding the scale of the token economy and cultivating new technologies, new products and new business forms such as tokens, software agents and super economic individuals. In addition, a series of detailed policy systems concerning specific fields and related businesses are being built in coordination, from the "Software Eight Measures" to the district-level token policy in Yizhuang, and then to the "Ten Measures for Agents," opening up full-chain support for computing power subsidies, token production, evaluation and distribution, agent cultivation and OPC entrepreneurship. The introduction of policies is only the starting point; the key lies in converting policy dividends into real industrial development results. As the city with the most concentrated and vibrant token commercialization applications in China, Beijing has clear advantages in the two core token fields of AIGC and programming-office work. AIGC model entities are highly concentrated in Beijing, and downstream businesses such as AI short dramas, short videos, advertising and games are developing rapidly; enterprises such as Zhipu AI and Moonshot AI have outstanding AI programming capabilities. These two tracks continue to release large-scale token consumption demand, driving the continuous gathering of token production and scheduling resources.
Trend assessment: Beijing's token economy will evolve toward "value leap"
Combining policy direction, industrial research and the needs of business entities, the Report predicts that in the next two to three years Beijing's token economy will undergo further profound changes and present five major trends: competition will shift toward "competing on quality and standard-setting voice." In the short term, the market competes on total token throughput; in the medium and long term, the core competitive points will rest on high-quality token supply, token cost-effectiveness and standard and norm leadership. Beijing is expected to export nationally replicable standards such as graded evaluation standards for token factories, token quality evaluation standards, and token measurement and settlement rules, moving from a token production concentration center to a token rule-setting powerhouse. The share of industry-specific and specialized token supply will continue to rise, and the industry will differentiate into a batch of dedicated token factories serving vertical industries such as education and teaching, industrial manufacturing, biomedicine and scientific research. The industrial evaluation system will be reformed, no longer simply using average daily token call volume as the sole indicator, but adding indicators such as the proportion of effective tokens, cost reduction and efficiency gains in the real economy, and paid conversion rates. Billing will also shift from token consumption volume to value-based billing. Agents will become the largest token consumption scenario. Tokens will not be monetized on their own; value realization will rely more on agent applications. In the next two years, general-purpose agents and industry-vertical agents will land on a large scale, and TaaS (Token as a Service) will evolve toward AaaS (Agent as a Service) and RaaS (Result as a Service) business models, with billing models gradually exploring a shift from call volume to business result value. OPCs (one-person companies), as a unique micro-sample of the token economy, are asset-light and flexible in innovation, and a large number of young entrepreneurs participate in the industry in OPC form. The measurement, pricing and trading rules of tokens directly determine the survival cost and development space of this group, and the OPC ecosystem will become an important testing ground for gauging the effectiveness of Beijing's token industry. Corporate credit should not be based solely on "call volume." Financial tools suited to digital asset-light enterprises will continue to improve. "Token loans" and "computing power loans" are not short-term financial products; behind them is an entirely new credit logic for AI digital enterprises: using desensitized token call logs, computing power contracts and business activity levels as credible operating credentials. But credit granting must not rely solely on "call volume"; multi-dimensional cross-verification must be established to prevent volume manipulation from creating "false prosperity." Token supply will move toward cross-regional coordination. At present, Beijing produces an average of 4 million to 5 million tokens per kilowatt-hour of electricity, and it is expected that by 2030, one kilowatt-hour of electricity may produce 20 million tokens. Behind this is not only Beijing's multi-dimensional advantages in talent, models, policy and application ecosystem, but also the boost brought by regional coordination. The Beijing-Tianjin-Hebei "Galaxy Computing Corridor" project is accelerating, and regional coordination gathering computing power around Beijing will effectively enhance token supply capacity and reduce per-token energy consumption and enterprise computing costs. Standardization and innovative trial-and-error will proceed in parallel. As the token industry heats up, regulation must follow in step, distinguishing "real use" from "speculation" in tokens: supporting genuine business use, strictly prohibiting speculative hype, and especially strictly prohibiting the financialization and speculation of tokens; at the same time, a token security audit mechanism should be established, complaint and verification channels improved, and focus placed on cracking down on scams such as fake token packages, so that enterprises put computing power to real use and users use their consumption in the open. Of course, regulation and encouragement are not contradictory: only by holding the safety bottom line while tolerating reasonable trial-and-error in the process of technological iteration can exploration space be preserved for new technologies and new models.
Tiered recommendations: guiding the industry from "competing on capacity" to "competing on quality, energy efficiency and industry fit"
The Report puts forward tiered recommendations. For government departments, policy release is only the starting point; the next stage of the token industry will shift its focus to detailed implementation and pilot promotion, with the core being to address the real demands of enterprises. The Report suggests that detailed rules for graded cultivation of token factories need further improvement, guiding the industry from "competing on capacity" to "competing on quality, energy efficiency and industry fit," and giving support to token factories with industry-leading influence. At the same time, policies should encourage the construction of industry-customized token production lines so that tokens of different grades are clearly labeled on trading platforms. While regulating token circulation and trading, it is recommended to carry out regular risk warnings, expose fake token packages and investment speculation scams, and standardize data desensitization and privacy protection requirements. For enterprises in the industrial chain, finding the right positioning is key. The Report suggests shifting from low-price involution to value creation. Upstream enterprises in chips, computing power and data should pay more attention to more pragmatic indicators such as "how many high-quality tokens are stably produced per kilowatt-hour," and advance the R&D and adaptation of domestic computing power, chips and models, and the supply of stable, compliant and high-quality data. Midstream models, token factories, evaluation and trading-distribution platforms should accelerate standardized processing and circulation, output auditable token consumption bills, and distinguish effective inference tokens, cached tokens and padding idle tokens; token factory enterprises should focus on raising the proportion of effective tokens, inference quality and domestic software and hardware adaptation capabilities; third-party evaluation institutions should actively participate in the development of local standards and build neutral token evaluation tool sets; large model enterprises, cloud service enterprises, operators and aggregation platforms should leverage their network and channel advantages to do a good job in token distribution, abandon the model of simply earning price differences, and transform toward value-added services for industry solutions. Downstream agents, vertical industry applications and OPCs need to complete value release. For financial and venture capital institutions, the valuation and credit logic needs to be reconstructed. Enterprises in the token field have the characteristics of "heavy upfront investment in computing power, revenue fluctuating with call volume, and lack of traditional real estate collateral," requiring exploration of a new financial logic of "computing power as credit, calls as credentials." The "total call volume" indicator can be used for reference, but emphasis should also be placed on indicators such as the proportion of effective tokens, paid customer revenue, depth of vertical scenario implementation, domestic software and hardware adaptation level, and repurchase rate of agent products. The Report warns: be vigilant against projects that rely on subsidies and volume manipulation to inflate call data. It also suggests that young developers and OPCs choose vertical niche tracks for entrepreneurship. In segmented industries and segmented scenarios, prioritizing the digitalization needs of small and medium-sized enterprises makes it easier to form a commercial closed loop; simply acting as a token agent or distributor has limited profit space, and value-added services are the core breakthrough point.
Multi-party coordination: letting the "token industry" empower all sectors of the economy
The AI industry is moving from the era of "building large models" to a new industrialization stage of "using tokens to create commercial value." Frontline research found that while the industrial scale expands rapidly, multiple practical contradictions are also affecting the high-quality development of the token economy. First, the phenomenon of "valuing capacity over quality" is emerging on the production side. Building token factories, increasing computing power and counting token call volumes are relatively easy, while cultivating industry applications that can sustainably create value takes longer. The Report's research shows that many regions and various types of entities across the country have entered the token field, but the industry has not yet fully implemented unified grading standards, and phenomena such as "token idling," ineffective tokens and inflated capacity have appeared in the market. Under the same number of calls, the knowledge density and reasoning ability of tokens output by different factories differ enormously; some vendors simply pursue the quantity of token output and ignore the actual business utility of tokens. Industry consensus predicts that in the next one to two years, the token industry will face a shakeout, with competition shifting comprehensively from a price war to a value war of "efficiency + security + ecosystem," and entities lacking technical barriers will gradually be cleared out. Second, the billing "black box" on the evaluation and measurement side remains to be cracked, and unified standards are urgent. This is an industry pain point that a series of reports previously focused on: different vendors have inconsistent tokenization calibers, and for the same passage of text, the number of tokens parsed by different models can differ by more than 50%, directly affecting billing and settlement; there is a lack of unified bill disclosure standards, and ordinary enterprises and developers cannot verify the authenticity of token consumption. Multimodal (image, audio, video) token measurement rules are even in a blank state. Third-party institutions for token quality evaluation are weak, and "who evaluates, how to evaluate, and how evaluation results are mutually recognized across platforms" has not yet formed a mature industry consensus. Third, circulation and pricing mechanisms have not yet taken shape. At this stage, token pricing is mainly led by leading large model vendors, lacking sufficient market-based competition; mechanisms for cross-entity token transfer, bulk trading and secondary scheduling are still in the exploratory stage. The "computing power-token" supply-demand mismatch persists for a long time, with computing power scarce during peak periods and idle during trough periods, and the cost for small and medium-sized tech innovation enterprises to purchase tokens remains high. Fourth, the traditional financing system does not match the business model of token enterprises. Some places such as Beijing have already launched pilots for "token loans" and "computing power loans," but the current product coverage, risk control models and trusted data connection mechanisms still need iterative improvement, and in particular, there is a need to avoid the risk of volume manipulation bubbles caused by credit granting based solely on token call volume. Fifth, the innovative vitality of young people is abundant, but entry thresholds and risks coexist. Beijing is rich in AI professional resources in universities, and young developers and OPCs are rising rapidly, making them an important innovative force in the token economy. But the procurement cost of computing power and tokens is high, industry guidance is lacking, and compliance awareness is insufficient; at the same time, the market has bred token speculation, fake token packages and investment scams, with some illegal merchants peddling fake token products and using the token concept to carry out wealth management investment scams, posing risks and threats to young entrepreneurs and ordinary users. At present, AIGC, programming-office work and token going global have become the three main revenue areas of Beijing's token economy. But it should also be noted that the development of the token economy is still in its initial stage, and in practice it still faces many practical bottlenecks and risk challenges, with the supporting policy guarantee system not yet complete. From a technical language and policy concept to an economic form with practical significance, Beijing's token economy still has a long way to go. The core of measuring the industry is "how much actual value tokens have created." This requires multi-party coordination among the government, enterprises, research institutions, capital and young innovative groups to open up the closed loop of policy implementation and resolve industrial bottlenecks in measurement, evaluation, financing and security, so that Beijing's "token industry" can truly empower all sectors of the economy and export Beijing experience for the high-quality development of the national AI industry.
Comments