Computing Power Becomes a Network, Intelligence Ushers in a New Future — A Panoramic Sidelight on the 2026 China Computing Power Conference

In September, autumn was deepening in Langfang, where a national-level computing power event crucial to the foundation of Digital China opened. From September 11 to 13, the 2026 China Computing Power Conference, themed “Weaving the Computing Power Network Together, Intelligence Ushers in a New Future,” was held in Langfang, Hebei. This was the first time a national-level computing power conference was held in the Beijing-Tianjin-Hebei region. The conference adopted an overall structure of “1+2+X+Y,” with one opening ceremony, two main forums, multiple sub-forums and special events, and concurrently hosted the “Computing Power China · AI and Computing Power Industry Exhibition,” presenting the full chain of computing power infrastructure construction, core technology breakthroughs, industry scenario implementation, and industrial ecosystem cultivation. During the three-day event, thousands of participants from government departments, research institutions, leading enterprises, and the investment community gathered to witness the key turning point of China’s computing power industry from scale expansion to quality and efficiency improvement.

Top-Level Deployment Intensifies, National Computing Power “One Network” Basically Formed

On the eve of the conference opening, a major piece of news set the tone for the event. On September 11, the State Council executive meeting held a special study on work related to the construction of the computing power network. It explicitly proposed adhering to rational layout and standardized order, further improving computing power infrastructure, actively promoting the R&D and application of key technologies and equipment, and building a multi-level networked computing power system; promoting computing-power-electricity coordination and computing-power-network integration, and strengthening the monitoring and scheduling of computing power resources and the construction of backbone optical fiber networks. The meeting emphasized that “the computing power network is the foundational support for AI development.” This formulation elevated the construction of the computing power network to the strategic height of national infrastructure and also set the direction for the core topics of this conference.

Policy signals were quickly echoed at the conference. At the opening ceremony, the China Computing Power Platform officially released nationwide integrated computing power monitoring, marking the basic formation of China’s computing power development pattern of “one network, one chessboard, and integration.” This platform is a national-level comprehensive computing power service platform integrating “supply, demand, and service.” It can achieve compatibility and interoperability among different systems, platforms, and tools, and is committed to creating a “neutral, open, fair, and trustworthy” service ecosystem for all parties in computing power supply and demand. It is reported that, on the basis of completing platform access for 10 provinces, autonomous regions, and municipalities in 2025, the China Computing Power Platform accelerated expansion and networking in 2026, promoting platform interconnection and access across 31 provinces, autonomous regions, and municipalities nationwide, building a unified, standardized, collaborative, and efficient nationwide situational awareness and monitoring system, effectively breaking down regional barriers and connecting data islands. Up to now, the platform’s computing power supermarket operations layer has gathered more than 10,000 registered enterprise users, more than 200 settled computing power service providers, more than 2,000 listed computing power products, and more than 300 large models of various types, accumulating billions of pieces of computing power monitoring big data. China’s coordinated computing power governance has entered a new stage of “all-domain monitoring, whole-network collaboration, and full-chain scheduling.”

At the same time, the National Computing Power Interconnection Regional (Hebei) Node was officially launched, becoming the first regional node to be launched nationwide. The launch of this node marks that Hebei’s computing power development has been formally incorporated into the national “1+M+N” top-level architecture for computing power interconnection, and it also provides a reference sample for the construction of other regional nodes. The first batch of inclusive computing power service lists for empowering the development of small and medium-sized enterprises was announced, covering a multi-level inclusive computing power service matrix including general computing power, intelligent computing power, large model services, and industry agent applications, and offering flexible payment models such as containerized leasing and pay-as-you-go billing. The “One Network, Three Integrations” Computing Power Financial Ecosystem Co-building Initiative was launched simultaneously, aiming to use financial instruments to leverage cross-regional circulation and efficient allocation of computing power resources.

Data shows that as of June 2026, the number of standard racks in use nationwide reached 15.56 million, the scale of intelligent computing power in use reached 2,185 EFLOPS (FP16), a year-on-year increase of 177%, and storage capacity exceeded 2,000 EB. More than 70 computing power transmission channels connecting computing power hubs and key regions were built, and 17 national regional nodes for computing power interconnection were approved. In his speech, MIIT Chief Engineer Zhong Zhihong pointed out that computing power has become an indispensable new type of information infrastructure in the digital-intelligence era. It is necessary to accelerate the construction of a multi-level, tiered computing power architecture of “hub-region-edge,” continuously improve the functions of the China Computing Power Platform, and promote coordinated monitoring and efficient utilization of computing power resources. Zhao Dachun, Vice Governor of Hebei Province, said that Hebei has always adhered to taking computing power as a core element for winning the future, seized the opportunity of building a nationwide integrated computing power network, and focused on building an intelligent computing power cluster around Beijing.

During the conference, He Baohong, Chief Engineer of the China Academy of Information and Communications Technology (CAICT), interpreted the “2026 Comprehensive Computing Power Panoramic Analysis.” He pointed out that China’s comprehensive computing power growth continues to accelerate, with computing power up 177% year-on-year, storage capacity up about 11%, a three-level network architecture of 1-5-20 milliseconds already formed, and the number of generative AI service filings up 32% year-on-year. However, he also acknowledged that the industry still faces three core challenges: obvious regional development gaps, insufficient integration of computing power with industry scenarios, and many constraints on the implementation of cross-regional computing power scheduling. The computing power industry is undergoing the pains of transformation from “competing on racks and piling up scale” to “focusing on collaboration and practical results.”

Technological Breakthroughs Announced, Space Computing Power Opens a New Track

One of the most closely watched segments of the conference was the announcement of 15 “Computing Power China · Annual Major Breakthrough Achievements.” These achievements cover key fields such as intelligent computing clusters, heterogeneous scheduling, liquid cooling infrastructure, hollow-core optical fiber, and trusted computing, with many reaching domestic leading and internationally advanced levels. Among them, Runze Technology’s single 200MW/100,000-card ultra-large-scale intelligent computing center, Taichu Yuanqi’s new-generation hyper-intelligent integrated computing system HyperIntelliX, China Tower’s heterogeneous network and heterogeneous computing power edge system, Unicom Digital Technology’s Xingluo domestic heterogeneous intelligent computing cluster global scheduling system, and China Mobile’s AI Trusted Computing (AITC) key technology were selected as “Computing Power China · Annual Outstanding Achievements.”

Taking Runze Technology as an example, its single 200MW/100,000-card ultra-large-scale intelligent computing center can carry 100,000 AI accelerator cards, is compatible with air cooling and liquid cooling, and maintains PUE stably below 1.2, supporting high-performance computing businesses such as general large model pre-training, model fine-tuning, and inference deployment. China Mobile’s AI Trusted Computing (AITC) key technology realizes integrated supply of multi-form intelligent and confidential computing power and confidential tokens, widely empowering data-sensitive application scenarios such as government processing, industrial manufacturing, medical diagnosis, and university research. ZTE’s domestic high-performance Matrix cluster supernode based on the OEX architecture compresses the chip adaptation cycle to 3 to 6 months, advances dual routes of “electrical switching + optical interconnection” and “optical switching + optical interconnection,” and achieves hundred-nanosecond-level low latency.

The ten “Annual Major Breakthrough Achievements” were also full of highlights. The “Efficient Hierarchical Metadata Management for Cloud Object Storage,” jointly recommended by Baidu, USTC, and Tsinghua, provides a more efficient storage foundation for massive data scenarios; the “Cross-domain Heterogeneous Agent Interconnection Technology System and Application Practice Based on Computing Power Network” from China Telecom Research Institute opened up the technical link for cross-domain agent collaboration; YOFC’s hollow-core optical fiber computing power interconnection technology set world records including the lowest attenuation of 0.04 dB/km and the longest drawing of 91.2 kilometers, and has completed more than 13 commercial and pilot projects at home and abroad. Tianyi Cloud’s domestic supernode architecture fully liquid-cooled 10,000-card intelligent computing cluster has a computing power utilization rate exceeding 80%, and the iteration cycle of the domestic Xingchen large model will be compressed by 50%, empowering more than 10 key industries. The concentrated appearance of these achievements not only reflects the vigorous vitality and accelerating evolution of China’s computing power industry, but also injects strong momentum into building an open and win-win new ecosystem of digital civilization.

One noteworthy signal is that space computing power has officially moved from proof of concept to industrial deployment. During the conference, Beijing Economic-Technological Development Area, Tianjin Economic-Technological Development Area, Hebei Xiong’an New Area, and Langfang City jointly issued an initiative to build the Beijing-Tianjin-Hebei Space Computing Power Industry Corridor. Each of the four locations clarified its positioning: Beijing E-Town focuses on “star-rocket sourcing and computing power going to space,” serving as the country’s core source for commercial aerospace star-rocket manufacturing, from where computing power departs for space; Tianjin E-Town focuses on “Tianhe foundation and hyper-intelligence integration,” relying on the Tianhe series of the National Supercomputing Center in Tianjin to promote intelligent integration of supercomputing and space computing power; Xiong’an New Area focuses on “SatNet leadership and network-computing integration,” relying on the headquarters of China Satellite Network to promote integrated collaboration between satellite internet and computing power networks; Langfang City focuses on “millisecond links and space-ground collaboration,” relying on the millisecond-level accessibility advantage of the computing power hub around Beijing to realize space-ground collaboration between ground and space computing power.

CAICT and the Beijing Space Intelligent Computing Research Institute jointly established the “Space Computing Power Key Technologies and Scenario Verification Laboratory,” the industry’s first laboratory focused on key technologies for space computing power and integrated verification of multiple scenarios. The 15th Five-Year Plan for the Development of the Information and Communications Industry issued by MIIT also explicitly proposed “exploring forward-looking research on space computing power technology and ecosystem cultivation.” Li Yantao, deputy director of the Beijing E-Town Management Committee, said that, guided by the Space-Earth Integrated Cross-domain Intelligent Computing Network (SEE) project, it will promote space computing power from “technology verification” to “large-scale application.” From ground to space, the boundaries of computing power infrastructure are being redefined.

Inclusive Computing Power Takes Root, Industrial Ecosystem Accelerates

Computing power must not only be built strong, but also used well. The “2026 Comprehensive Computing Power Panoramic Analysis” released at the conference pointed out that China’s comprehensive computing power is accelerating its shift from “building more” to “using well and coordinating well,” and intelligent computing power has become an important support for the comprehensive computing power system. This report, for the first time, upgraded from a traditional ranking format to a coordinate matrix map format, conducting a quantitative assessment of 31 provinces (autonomous regions and municipalities) around four dimensions: computing power, storage power, transport power, and environment. The analysis results show that Beijing, Shanghai, and Guangdong perform outstandingly in model capabilities such as large model R&D, generative AI services, and industrial ecosystems; Guangdong, Hebei, and Shanghai have relatively strong storage foundations; Jiangsu, Zhejiang, and Guangdong have relatively good transport coordination capabilities; and regions such as Xinjiang, Gansu, and Qinghai have obvious resource and environmental advantages. This differentiated pattern shows that the national “one chessboard” for computing power does not pursue homogeneous development in all regions, but rather adapts to local conditions and allows each to leverage its strengths.

The first batch of inclusive computing power service lists for empowering SME development announced at the conference aims to reduce computing costs and application thresholds, transforming the “dividends of inclusive computing power applications” into momentum for SME development. Representatives from CAICT, China Telecom, China Mobile, China Unicom, China Broadcasting Network, Alibaba Cloud, ZTE, Runze Technology, SenseTime, and other enterprises attended the release ceremony, demonstrating the determination of all parties in the industry to jointly promote inclusive computing power.

At the exhibition site, a group of technical solutions close to actual industry needs were presented, demonstrating the shift of computing power from “usable” to “easy to use.” Taichu Yuanqi exhibited for the first time offline a 20-foot standardized container computing power product. This “prefabricated computing power” uses a standardized container as a carrier, is built with 64 domestically developed AI accelerator cards, and has a prefabrication rate of over 90%. After leaving the factory, the entire unit only needs to be connected to power on site for deployment, can be put into use within 24 hours, and can be installed in as little as 2 hours. Xia Zhongmou, chief architect of Taichu Yuanqi, said: “Domestic computing power is moving from ‘usable’ to ‘easy to use.’ In the end, domestic computing power competition will certainly be about cluster collaboration and full-stack ecosystem capabilities, not paper parameters of a single card.” At the same time, the conference released original technologies and innovative products such as the Wanjie Ark large model MaaS platform, the AI computing power full-lifecycle operations service plan, and the computing-network-cloud-edge integrated solution, building a window for public display and resource matching of computing power innovation achievements.

The activity on the industry side is also striking. According to a report by Kaiyuan Securities, in the first quarter of 2026, the domestic computing power leasing market reached 68 billion yuan, a year-on-year increase of 62%, and is expected to exceed 260 billion yuan for the full year. Computing power leasing has emerged with a light-asset model of “pay-as-you-go and elastic expansion,” and is becoming a core path to fill the supply gap and reduce corporate capital expenditure. CAICT data shows that in the first quarter of 2026, domestic AI computing power demand increased by 417% year-on-year, while effective supply grew by only 128%, and the supply-demand gap continued to widen. Driven by the resonance of supply constraints and demand growth, the one-year lease contract price of H100 GPUs rose from a low of $1.70/hour in October 2025 to $2.35/hour in March 2026, a cumulative increase of nearly 40%; the oral quotation for top-spec B300 in China reached 14.5 million yuan, and high-quality computing power entered a “scramble for goods” mode.

The three major telecom operators concentrated on displaying full-stack computing power capabilities at this conference: China Mobile released achievements such as the AIDC integrated operations and maintenance management platform and the MoMA government-enterprise edition, connecting the complete service chain of “technical architecture—intelligent computing foundation—model platform—implementation applications”; China Telecom exhibited the Intelligent Transmission Network, which won the highest SAIL Award at the 2026 World Artificial Intelligence Conference, and released a domestic intelligent computing Token efficiency optimization system; China Unicom displayed the Xingluo domestic heterogeneous intelligent computing cluster global scheduling system. This end-to-end capability display from infrastructure to industry applications shows that operators are accelerating their transformation from simple “pipe providers” to “comprehensive computing power service providers.”

Regional Coordination Deepens, Hebei Sample Draws Attention

As the host of the conference, Hebei’s computing power practices have become a window for observing regional coordination in China’s computing power industry. In the past two years, Hebei Province’s comprehensive computing power index has consistently ranked first among 31 provinces, and Langfang’s urban computing power sub-index has ranked first nationwide for two consecutive years. The province’s intelligent computing power scale has reached 556 EFLOPS, with 2.5 million standard racks in use. It has developed 441 vertical large models and 98 agents in 26 fields including steel, chemicals, transportation, and healthcare, and innovatively launched support measures such as “computing power vouchers, token vouchers, model vouchers” and “computing power+” industry funds, attracting nearly 200 upstream and downstream enterprises to settle, with the province’s computing power utilization rate exceeding 80%.

Langfang’s positioning is particularly clear. Liu Yuan, deputy secretary of the Langfang Municipal Party Committee and mayor, introduced at the opening ceremony that Langfang is located in the core hinterland of the Beijing-Tianjin-Xiong’an golden triangle, and data round trips to Beijing’s core nodes take only 1 millisecond. It focuses on developing real-time computing power for AI inference and internet services, forming complementary and differentiated development with the western hub nodes of the “East Data, West Computing” project. Seven of China’s top 10 computing power service providers have deep layouts in Langfang. The city has 36 large data centers in operation, with intelligent computing supply capacity exceeding 200,000 P. Langfang also launched the “China Computing Power · Computing in Langfang” computing power scheduling service platform, further strengthening its functional positioning as the bridgehead for real-time computing power in the Beijing-Tianjin-Hebei region. Up to now, all 36 data centers put into operation in Langfang have begun transforming into intelligent computing centers, and Runze Technology has put into production 8 internet data centers and 4 artificial intelligence data centers in Langfang. From “not in Langfang, but computing in Langfang” to building a bridgehead for real-time computing power in the Beijing-Tianjin-Hebei region, Langfang is becoming a node city that cannot be ignored in the national computing power landscape.

Computing power coordination in the Beijing-Tianjin-Hebei region is accelerating and deepening. At present, the intelligent computing power scale in the Beijing-Tianjin-Hebei region has reached 656,000 P, accounting for 30% of the national total; the number of racks in use has reached 3.547 million. At the Beijing-Tianjin-Hebei Intelligent Computing and Advanced Manufacturing Supply-Demand Forum held during the conference, the Beijing-Tianjin-Hebei Computing Power Supply and Demand List was released, building an authoritative platform for efficient matching among government and enterprises, industry and research, and supply and demand, and promoting efficient allocation of computing power resource elements. At the same time, the Xiong’an International Computing Power Integrated Scheduling Center was launched, and the Xiong’an Computing Power Alliance was established simultaneously, releasing typical application scenarios of computing power empowering the future city and an integrated construction system for computing-network integration. From Langfang’s real-time computing power hub to Xiong’an’s exploration of computing-network integration, from the space computing power industry corridor to the computing power supply and demand list, the Beijing-Tianjin-Hebei region is forming a regional computing power coordination pattern with clear levels and complementary functions.

In terms of green computing power, this conference also sent a clear signal. The computing power industry is accelerating its transformation from scale expansion to efficiency, green, and safety, and solutions such as liquid cooling technology, prefabricated deployment, and local consumption of green electricity have become high-frequency keywords at the exhibition. Runze Technology’s intelligent computing center is compatible with air cooling and liquid cooling, with PUE stably below 1.2; Tianyi Cloud’s domestic supernode architecture fully liquid-cooled 10,000-card intelligent computing cluster effectively controls energy consumption while achieving high computing power utilization. Nationwide, the optimal PUE of computing power cabins can be no more than 1.15, significantly better than the national average of 1.34 for data centers. Green computing power is moving from an “optional action” to a “mandatory option.”

Computing Power Available on Demand, Inclusive Path Still Needs “Network Weavers”

The three-day conference has ended, but the process of “weaving the computing power network together” has only just accelerated. From the top-level deployment of the State Council executive meeting to the full access of platforms in all 31 provinces, from the hardcore announcement of 15 annual breakthrough achievements to the cross-regional coordination of the space computing power industry corridor, from the release of the inclusive computing power service list to the co-building action of the computing power financial ecosystem, this conference conveyed a clear signal: computing power is changing from a scarce resource into public infrastructure, and from the exclusive capability of a few leading enterprises into a production factor within reach of thousands of industries.

When computing power is available on demand and inclusively accessible like water and electricity, and when the national “one network” of computing power turns from a blueprint into reality, the foundation of the AI era is becoming solid and clear. As the theme of this conference indicates—intelligence ushers in a new future—every warp and weft of the computing power network is weaving a more intelligent and more balanced Digital China. In this process, service platforms connecting the two ends of computing power supply and demand play an indispensable role. Yunqing Tianxia Computing Power Leasing Platform (www.omniyq.com) is one of them. The platform focuses on bulk GPU computing power supply and leasing services, covering the full spectrum of high-performance computing clusters such as H20, H200, A100, B200, and B300, providing flexible and cost-effective access to computing power for small and medium-sized enterprises and research teams, and promoting inclusive computing power from a policy vision to an industry daily reality through practical actions. As more and more “network weavers” participate, every warp and weft of the national “one network” of computing power will be woven into the most solid foundation of Digital China.