The year 2026, April, marks a pivotal transition period for China's digital economy. It has been designated by the National Data Administration as the "Year of Data Factor Value Release" and also serves as the inaugural year of the 15th Five-Year Plan period.
On April 29th, at the opening of the 9th Digital China Summit in Fuzhou, Liu Liehong, member of the National Development and Reform Commission's Party Leadership Group, Party Secretary and Director of the National Data Administration, delivered the opening address. While the speech was concise, it was dense with critical data. One of the most significant and thought-provoking signals is that the underlying logic of China's data processing is clearly shifting from large-scale transmission of traditional files towards more granular, precise, and metered intelligence—namely, the "Token" economy. Director Liu formally disclosed that the national daily Token call volume grew from over 1 trillion at the beginning of 2025 to 100 trillion by year's end. By March 2026, this figure had further increased to 140 trillion.
This represents more than just an iteration of a technical term. For observers of Digital China and the AI economic landscape, it signifies a "systematic deep dive" into a new phase for China's entire digital infrastructure and business models.
**First Signal: A Major Shift in Data Production Methods** Liu Liehong began by presenting a set of highly significant foundational data: in 2025, the total national data production reached 52.26 zettabytes (ZB), an increase of nearly 11.2 ZB from the previous year, representing a year-on-year growth of 27.28%.
For decades, the overwhelming majority of data output has consistently come from IoT sensor data and manually input data. However, a turning point arrived in 2025: for the first time, data produced by AI surpassed that collected through traditional methods. Data generated by system software, artificial intelligence, and other emerging production factors amounted to 26.92 ZB, accounting for over half of the total.
Furthermore, at the summit's thematic exchange on "High-Quality Datasets," Liu announced another key achievement: as of the end of March 2026, the number of high-quality datasets built nationwide had exceeded 116,000, with a total volume surpassing 960 petabytes (PB). To put 960 PB into perspective, it is roughly equivalent to 336 times the digital resource volume of the National Library of China. The current competition in large AI models ultimately hinges not only on floating-point operations of core computing power but, more critically, on the scarcity of high-quality training corpora and industry-specific data.
**Second Signal: Inference is No Longer a "One-Sided Affair"** In his address, Director Liu highlighted a second, highly significant trend shift. In recent years, public perception of the large model and AI industry has largely remained fixated on the notion of "expensive training." But in 2025, an inflection point occurred: The total data volume used for AI training and inference throughout the year was 199.48 exabytes (EB), a year-on-year increase of 42.86%. Notably, the data volume for "inference" surpassed that for "training" for the first time, reaching 101.34 EB.
Liu's assessment on this is very clear: this marks AI's entry into a new stage characterized by "equal emphasis on training and inference, with application-driven iteration." Enterprises are no longer solely spending unlimited funds to "develop a large model." Instead, they are using models daily in practical applications, continuously "reverse-validating" them through high-frequency user interactions within real consumption and production scenarios. For instance, driven by the large-scale implementation of trusted data spaces in sectors like energy, transportation, and automotive, the cost and efficiency of AI agents are rapidly approaching the optimal zone for commercial viability. In other words, doing more and using more has become more important than mere training.
**Third Signal: The "Token" Holds Vast Value Potential** Just days earlier, on April 25th, the National Data Administration released the "Implementation Plan for Promoting the Construction of High-Quality Industry Datasets (Draft for Comments)," which, for the first time at the national policy level, formally proposed "exploring Token trading" as a new model for data transactions. At the summit, Liu Liehong directly articulated a statement carrying significant weight: "Tokens are painting a new picture for the release of data factor value." His complete judgment is: Tokens are the smallest unit of information processed by large models, making intelligent services measurable, pricable, and tradable. As a comprehensive embodiment of data, models, and computing power, Tokens reflect the vitality of the intelligent economy. How to understand this change? Let the data speak: "The national daily Token call volume grew from 1 trillion at the beginning of 2025 to 100 trillion by year-end, and further increased to 140 trillion by March this year." This data indicates that the value of data factors is being released at an accelerating pace, and the intelligent economic form is rapidly taking shape.
Simultaneously, at the China Mobile AI Ecosystem Conference held on the same day, Liu Liehong elaborated on four directions regarding the "Token economy": 1. Deepen the construction of high-quality datasets (wherever AI goes, high-quality data follows). 2. Construct computing power infrastructure and a nationally integrated computing power network with high quality. 3. Actively promote business model innovation. 4. Build a collaborative innovation ecosystem integrating "models, chips, cloud, and applications."
The summit also revealed concrete actions from China Telecom and China Mobile. China Telecom officially launched a one-stop Token service platform at its Intelligent Cloud Ecosystem Conference. This platform aims to provide efficient and inclusive services across the entire chain of Token production, scheduling, distribution, and value realization. Chairman Ke Ruiwen explicitly stated that "the essence of Token operation is to provide AI services for users." Leveraging its "Tiangang" platform, China Mobile has, by April 2026, implemented over 50 industry-specific large models and more than 3,000 AI+ projects, and for the first time released its intelligent service matrix.
At the summit's Trusted Data Space sub-forum, Director Liu emphasized that a key task during the 15th Five-Year Plan period is to establish a secure and trusted foundation for data circulation and to build a benefit-sharing mechanism based on mutual trust in rules and co-creation of value. It is reported that the National Data Administration has previously selected the first batch of 63 trusted data space pilot projects. At the launch ceremony, trusted data spaces for cities including Shanghai, Ningbo, Wenzhou, Xiamen, Guangzhou, Jinan, and Xiong'an, as well as those undertaken by industry and enterprises such as China Southern Power Grid, China Automotive Technology & Research Center, and Xincaidao, announced the commencement of services.
Looking at the private enterprise level, on April 28th, the day before the summit opened, Liu Liehong presided over a symposium with private enterprises in the digital economy. He clearly stated that in the next step, the National Data Administration will deeply advance the "5+3+1" work system for the market-oriented allocation reform of data factors, making coordinated efforts in areas such as data factor supply, scenario opening, and industrial ecosystem cultivation to accelerate the process of marketization and valorization of data factors. According to public information, the "5+3+1" work system specifically refers to: - "5" represents five aspects: improving the basic data system; constructing and operating data infrastructure; scenario cultivation and data integration application; building a nationally integrated data market; and strengthening the data industry. - "3" represents three foundations: breakthroughs in core data technologies; data standardization; and cultivation of digital talent. - "1" refers to the key direction of data empowering AI innovation and development.
The gathering of the three major telecom operators and leading tech private enterprises represents the urgent desire of millions in the industry to step onto the "new continent" created by data factors at the outset of the 15th Five-Year Plan in 2026. The underlying driving force of China's internet economy is shifting from APP traffic to "new quality productive forces" driven by computing power and Tokens. As Liu Liehong concluded his address: "I wish the 9th Digital China Summit a complete success." However, for Digital China, this is merely the first step in the inaugural year of the 15th Five-Year Plan construction. For the millions of practitioners in the data factor market, it is the true starting signal to "ignite 2026." Digitization is no longer novel, but stepping into the intelligent "Token era" has only just begun.