According to the recently released "15th Five-Year Plan for the Information and Communication Industry Development", by 2030, the national intelligent computing power scale will reach 9,800 EFLOPS, and the advanced storage capacity will reach1,700 EB. China's computing power industry is undergoing a profound transformation from "scale expansion" to "high-quality development".
China is fully entering an era of rapid growth in AI (Artificial Intelligence) computing power. How should a reliable technological foundation be built? How can industrial development be continuously driven? How can the rapidly evolving AI computing power fully empower the socio-economy? All sectors of the industry need to provide new answers for the new stage of computing power.
From September 11 to 13, the 2026 China Computing Power Conference, themed "Weaving the Computing Power Network Together, Intelligently Initiating a New Future", was held at the Beijing-Tianjin-Hebei Big Data Innovation Application Center in Langfang, Hebei Province. Government authorities, research institutes, and upstream and downstream enterprises in the computing power industry chain conducted in-depth discussions on the construction of a national integrated computing power network, AI new quality productive forces, computing power technological innovation, and industry implementation.
During the conference, Guo Zhenxing, Vice President of Huawei's Enterprise Business Group in China, delivered a keynote speech titled "Unleashing AI New Quality Productive Forces to Help Thousands of Industries Leap to New Value". He analyzed the current development status of the domestic AI industry, the evolution direction of computing power infrastructure, the practical path of industry intelligence, and value leap, and introduced in detail Huawei's All Intelligence strategy and the specific path to empower the development of the computing power industry.
Based on this speech, we can analyze how Huawei builds the "three carriages" for computing power development to escort China's computing power into an era of rapid growth.
China's Computing Power Steps into a New AI Era
Data released by the National Data Administration shows that by the end of July, the total scale of national intelligent computing power had reached 2.45 million PFLOPS (FP16). The intelligent computing scale built in the eight national computing power hubs and three computing-power-electricity coordinated development zones accounts for over 85% of the national total.
The rapid expansion of intelligent computing scale is inseparable from the continuous maturation of the AI industry, and its beneficial effects on the socio-economy are becoming increasingly evident. According to forecasts by the think tank "AI 2027", the AI technology singularity will be crossed in 2027, and computing power will gradually replace labor, which is expected to push China's GDP growth rate to break through 10% again in 2028. A series of new technological trends, such as super nodes, token factories, and world models, are continuously driving up the demand for AI computing power. AI computing power, like water and electricity, has become the underlying driving force of the digital economy.
The key challenge in the next stage of China's AI computing power development lies in how to fully integrate the massive AI computing power into the socio-economy, bringing true new quality productive forces value to thousands of industries and a vast number of Chinese enterprises. To achieve this goal, on the one hand, it is necessary to possess reliable and advanced AI root technologies; on the other hand, a comprehensive and usable computing power industry ecosystem is required to realize the positive empowerment of the economy by AI technologies and AI computing power.
Guo Zhenxing pointed out that AI has now bid farewell to conceptual demonstrations and officially entered the stage of large-scale industrial implementation. Computing power has become the primary factor of production in the AI era, reshaping the production relations of thousands of industries. Data shows that the domestic daily average token invocation volume has rapidly climbed to hundreds of trillions, and AI has completed the leap from "being able to converse" to "being able to execute". Industries such as finance and electric power have deployed tens of thousands of industry agents. The daily average token consumption has become a new reference scale for measuring the intelligence level of enterprises. China is already at the forefront of the world in the scale and depth of AI applications.
So, as the computing power industry enters a new AI era, how does Huawei connect computing power with productive forces, and how does it fully release technological dividends into industrial dividends? Guo Zhenxing provided a detailed answer in his speech.
Huawei's All Intelligence Strategy Paves the Way for Full-Stack AI
To open up the path from AI computing power to new quality productive forces, the first step is to build an advanced, reliable, and comprehensive AI root technology system, and construct an AI capability closed loop from AI computing power to AI ecosystem, model capabilities, and agent capabilities. Only when AI technology takes root deep down can the productivity transformation brought by AI reach for the sky.
To this end, Huawei has persisted in advancing the All Intelligence strategy, building a complete full-stack AI system, thereby enabling hundreds of models and thousands of forms, and empowering thousands of industries.
In terms of computing power, Huawei has built a solid computing power foundation. Based on "super nodes + clusters", it builds a second plane of computing power for the world; it opens the Lingqu protocol and super node reference architecture, enabling partners to develop related products and components, and jointly build the Lingqu open ecosystem.
In terms of AI ecosystem construction, Huawei continues to promote software openness and hardware open source. It has open-sourced fundamental software such as the hardware enablement platform CANN, the AI framework MindSpore, and application enablement suites, enabling partners and developers to innovate efficiently.
In terms of AI large models, Huawei is firmly committed to saturated investment in the Pangu foundation large model, and ensures the native adaptation of Ascend fundamental software and hardware with mainstream open-source large models.
In terms of agent capabilities, Huawei has created agents such as the terminal Xiaoyi and Qiankun Intelligent Driving, allowing AI to sink to the edge; for ToB scenarios, it has built an industry agent enablement platform, constructing a technical system covering Agentic Infra, MaaS model services, and agent development platforms, and jointly built the "Industry AI Dream Factory" with partners to accelerate the intelligence process of industry users.
Thus, we can see that at every stack layer of AI computing power foundation, AI software and hardware ecosystem, AI large model capabilities, and agent capabilities, Huawei has built reliable, usable, and fully leading systematic capabilities, which is truly unique in the industry. Based on this, Huawei has not only clarified the path for the development of China's AI technology but also defined the direction for how China's AI computing power construction can be integrated with the socio-economy.
Under the guidance of the All Intelligence strategy, Huawei has successfully built the "three carriages", thereby driving China's computing power industry into an era of rapid growth with high-quality development.
Forging the "Three Carriages" to Propel China's Computing Power Forward
Huawei's core AI capabilities and All Intelligence strategy require specific transformation paths to be implemented in the industry.
Guo Zhenxing emphasized that Huawei has proposed the direction of advancing the "three carriages" of advanced products, co-built ecosystems, and valuable scenarios in tandem, aiming to comprehensively unleash the new quality productive forces of AI and help industries achieve new value leaps.
At the product level, Huawei has made continuous breakthroughs in the semiconductor underlying architecture of AI computing power, computing power system architecture, and AI clusters, ensuring that Ascend completes industrial upgrades at a pace of one iteration per year and one architecture upgrade every two years, launching breakthrough products such as the 384 super node. Facing diversified AI computing power product demands, Huawei also provides a rich selection of products, including super nodes and clusters for large computing power scenarios, high-density computing power all-in-one machines for medium computing power scenarios, and modules and boards for small computing power scenarios, meeting differentiated AI computing power product needs.
At the ecosystem level, Huawei has solidified the core strategy of ecosystem development in the computing power field. At the model ecosystem level, it has achieved 0-day adaptation for mainstream models of the Qwen and DeepSeek series, with over 10 core models delivering optimal performance out of the box. In the software development ecosystem, Ascend has achieved comprehensive coverage of key industries, and together with ISVs (Independent Software Vendors), it has covered 80% of business scenarios. In the talent ecosystem, Huawei has completed a series of industry-university-research cooperation projects with universities such as Tsinghua University, Peking University, Harbin Institute of Technology, and Shanghai Jiao Tong University, building a talent foundation for the computing power industry.
At the scenario level, facing the demand for AI to go deep from assisting office work into the core of enterprise production, Huawei has created a series of business scenarios that fully unleash enterprise AI value, based on its experience of serving over 3,000 enterprise customers in the past two years.
Huawei empowers thousands of industries through AI. In Shandong Energy, the distillation model helped customers increase the clean coal yield rate by 0.2%; in Valin Xiangtan Steel, the scrap steel grading accuracy reached over 90%; in State Grid, the transmission line defect identification accuracy improved by 18%; in Ansteel, the coal blending cost was reduced by CNY 9 million per month; and in a leading domestic bank, the intelligent customer service increased customer purchase intention by 10%.
Relying on rich industry experience, Huawei deeply integrates underlying AI capabilities with customers' production businesses, finding entry points for AI technology in high-value AI scenarios across various industries. For example, in 2025, China Baowu Steel Group launched the "2526" project. Facing challenges such as the difficulty of black-box control of blast furnaces and numerous scenarios, relying on the Ascend computing power foundation and one-stop AI development platform provided by Huawei, it implemented over 120 AI applications, including blast furnace temperature prediction, hot rolling width expansion prediction, and steel plate surface inspection. Ultimately, it achieved a blast furnace temperature hit rate of double 90%, with a single blast furnace increasing annual benefits by over CNY 10 million; hot rolling prediction accuracy improved by 5%, saving over CNY 90 million per year for a single production line; surface defect identification accuracy reached 95%, reducing annual losses by over CNY 100 million.
Huawei's cloud-edge collaborative computing power construction solution of "central training" + "edge inference" has helped numerous enterprises improve production efficiency. In Shandong Energy Group, the group's entire business embraces AI, covering 9 categories and over 300 scenarios in mining business. Huawei and Yunding Technology deeply co-created, reducing the AI development cycle from 3 months to 2 days, and improving model identification accuracy by over 20%, effectively reducing production safety risks.
In Conch Cement, Ascend AI helped it build a large model for the building materials industry, completing optimization and strength prediction for over 100 production line calcination processes, accelerating digital and intelligent transformation. In the global optimization scenario for calcination, coal consumption decreased by 1%; in the clinker strength prediction scenario, the accuracy rate reached 85%. Huawei and Conch Group jointly innovated and incubated a global best practice case of AI + APC (Advanced Process Control).
As of now, Huawei has cooperated with over 2,700 industry partners and 80 hardware partners, jointly creating over 5,800 industry solutions, deeply empowering AI to integrate into thousands of industries.
Using products to drive AI computing power to adapt to extensive demands; using ecosystems to gather strength from all parties to enhance the extensive compatibility of AI computing power; using scenarios to drive AI computing power to unleash new quality productive forces, bringing real socio-economic value enhancement. It can be seen that Huawei's "three carriages" adapt to local conditions and guide according to the situation, highly aligning with the development needs and context of China's computing power industry, thereby transforming the scale advantage of computing power into socio-economic advantages, and giving the era of rapid growth of China's computing power real acceleration.
Looking to the future, Guo Zhenxing stated that Huawei will continue to invest in the construction of the Pangu large model and industry agent enablement platforms, adhere to open source and openness, and is willing to join hands with industry partners to fully leverage Hebei's locational and energy advantages, empowering the value leap of thousands of industries with computing power foundations and a prosperous ecosystem, and jointly promoting the high-quality development of China's computing power industry.
The 2026 China Computing Power Conference is a concentrated review of the innovation achievements of China's computing power industry. Huawei's full-stack and full-link AI capability innovation, as well as the computing power industry driving force built by the "three carriages" of products, ecosystems, and scenarios, received an exciting review here.
Returning to the initial point, China's computing power industry is facing a critical stage of moving from scale expansion to high-quality development. Computing power scale, AI technology, and industrial demands have successively matured. The most important task at the current stage is to complete the connection in these industrial spaces, realizing the seamless integration of computing power - intelligence - industry, which is the true meaning of high-quality development. The "three carriages" built by Huawei are created exactly for this purpose. With advanced products, a prosperous ecosystem, and mature scenarios, computing power can continuously transform technological value into industrial value, and the AI era can truly arrive. Huawei's "three carriages" are running on the broad avenue of the computing power network at the right time.
Huawei, which "takes root deep down and reaches for the sky", is shining in tandem with the new era of China's computing power. As computing power gallops, Huawei takes the lead, and China leads the way.