Zheng Weimin, Chairman of the Open AI Infra Community Advisory Board, Member of the GCC Strategic Advisory Committee, Academician of the Chinese Academy of Engineering, and Professor at Tsinghua University; Jin Hai, President of the Global Computing Alliance, Professor and Doctoral Supervisor at Huazhong University of Science and Technology; Zhang Chun from China Mobile, Chairman of the Open AI Infra Community Management Committee; Gao Xiaojun from ByteDance, Co-Chairman of the Community Management Committee; and Long Pan from Huawei, Co-Chairman of the Community Management Committee, among other prominent guests, attended.

Leaders and technical experts from numerous community member units, including those from the internet, operators, core technology providers, etc., as well as over 1,500 ecosystem partners and industry peers in the field of AI Infra gathered on-site, jointly creating an industry event in the field of AI computing power facilities and working together to explore new paths for industrial development featuring open collaboration and ecosystem win-win results.

Jointly Building the Ecosystem Blueprint and Charting the Strategic Course - The Strategic Leadership Chapter of the Open AI Infra Community

Jin Hai, the President of the Global Computing Consortium (GCC), delivered a welcome speech at the conference. He pointed out that since its establishment, the Open AI Infra Community has brought together over 150 core units in the industrial chain, covering partners across the entire value chain including internet, telecommunications carriers, chips, complete machines, liquid cooling, power supply and distribution, and data centers, and has built  a high-quality industrial ecosystem characterized by demand-driven, technology-supported, open and collaborative, and effective implementation .

Looking ahead to the future development of the community, Chairman Jinhai put forward three initiatives: First, adhere to the original aspiration and mission, consolidate the foundation of co-governance, and always uphold the core principles of fairness, justice, and openness; Second, focus on technological breakthroughs, overcome key bottlenecks, and make in-depth breakthroughs around core directions such as megawatt-level computing power systems, supernode clusters, high-speed interconnection, and green and efficient infrastructure, solidify technical specifications, and strengthen the implementation of applications; Third, promote the implementation of standard achievements, guided by industrial needs and aiming at large-scale implementation, transform industry technical consensus into executable, replicable, and commercially viable de facto standards.

Zheng Weimin, an academician of the Chinese Academy of Engineering and a professor at Tsinghua University, said in his opening speech that the Open AI Infra Community has been deeply involved in and strategically deployed in cutting-edge fields such as open supernodes, AIDC, and component ecosystems, basing itself on national strategies, solving industry challenges, and leading industry upgrades, which is of great significance and far-reaching impact. He proposed thatin the future, computing power will be a new type of intelligent complex featuring deep integration of computing, storage, transmission, control, and cooling, providing inclusive, efficient, and green computing power support for the intelligent transformation of various industries.

The community continuously links with the industrial chain, promotes the standardization of technologies such as liquid-cooled full racks, improves the test and verification system, provides performance verification and quality assurance for open super nodes, AIDC, and core components, standardizes industrial development, and lowers the application threshold; meanwhile  continues to make efforts in areas such as optical interconnection, megawatt-level computing power systems, complete machine R&D, and evaluation systems , jointly builds an open, collaborative, secure, and reliable global computing power ecosystem, and injects strong impetus into the high-quality development of the computing power industry in China and even globally.

Jointly Focus on Ecosystem Co-construction and Co-build a Benchmark of Achievements —— Open AI Infra Community's Comprehensive Development Chapter

At the Open AI Infra Community Panorama Session of the Main Forum on the morning of April 10, Zhang Chun, Chairman of the Open AI Infra Community Management Committee, shared with the theme of "Open Innovation, Collaborative Win-Win - A Review of the Growth and Outlook for the Development of the Open AI Infra Community".

During the sharing, Zhang Chun stated that since the establishment of the Open AI Infra community, it has always adhered to the core idea of "open collaboration, technology leadership, and industry closed loop"  , closely linking the two main bodies of core users and core technology parties, gathering industry users, equipment vendors, evaluation institutions, academic institutions and other forces, focusing on demand definition, open collaboration, full stack integration, continuous evolution, technological innovation, business closed loop, and industrial chain to promote ecosystem construction. Committed to building a smart computing infrastructure that adapts to the AI era and supports the rapid development of the AI industry.

Zhang Chun emphasized that every step of growth in the Open AI Infra Community is inseparable from the wisdom and dedication of community members. In the future, the Open AI Infra Community will continue to uphold the concept of "Open Innovation, Collaborative Win-Win", work hand in hand with all member units, jointly build a more prosperous, efficient, and sustainable new ecosystem for AI computing power, and make every effort to boost the high-quality development of the intelligent economy.

At this conference, the Open AI Infra Community also held a series of appointment and recognition ceremonies to showcase the achievements of community building, pay tribute to the enterprises, institutions, and experts who have dedicated themselves to community development, and further strengthen the organizational structure and consolidate the synergy of the ecosystem.

At the ceremony for presenting letters of appointment to members of the Community Advisory Board, Chairman Jin Hai and Professor Zheng Weimin jointly presented certificates to the appointed experts, pooling top-notch intellectual resources in the industry to safeguard the high-quality development of the community.

The community simultaneously held an appointment ceremony for project team leaders, where a group of senior industry experts from the internet field were officially hired to take the lead in key areas such as AI full racks, AIDC infrastructure, core components, and Benchmark tools, providing strong support for the development of community technical specifications, standard setting, and project implementation.

Meanwhile, to pay tribute to community partners who have been deeply involved in and dedicated to community work, the conference will simultaneously hold an annual recognition ceremony to honor experts, enterprises, and individuals who have made outstanding contributions to the advancement and implementation of community projects, thus rallying the driving force for ecosystem progress.

Experts from nine enterprises, including CenturyLink Cao Weibing, Huawei Mo Pengbo, GDS Liu Tao, H3C Wang Xinxin, Sinopec Lubricant Zhang Kaijiao, China Telecom Services Guo Cong, Hyper Fusion Ding Junfeng, ZTE Xia Jinsong, and Inventec Ji Yidong, have received the  Open AI Infra Community 2025 Annual Expert Contribution Award .

Beijing Douyin Information Service Co., Ltd., Huawei Technologies Co., Ltd., Hebei Jingmai Technology Co., Ltd., and Shenzhen Envicool Technology Joint Stock Company, these four enterprises, have won the 2025 Annual Member Contribution Award of the Open AI Infra Community.

Beijing Douyin Information Service Co., Ltd.'s Wang Shifeng, Huawei Technologies Co., Ltd.'s Cai Zhiping, and JD Cloud Service Co., Ltd.'s Jin Yuehong were awarded the 2025 Outstanding Service Award by the Open AI Infra Community.

At the release session of the Standard / Specification / Benchmark series of achievements, relying on the strong support of all member units of the community and the dedicated efforts and collaborative breakthroughs of all experts, Yu Xiongjie, Co-Chair of the Community Technical Steering Committee and Technical Expert at ByteDance, said that  the community adheres to the principle of seeking truth and being pragmatic, with specifications, certifications, and implementability, ensuring the safe delivery and stable operation of IT equipment. The community has jointly released multiple significant achievements with multiple enterprises and experts, comprehensively demonstrating the effectiveness of its phased work.

The <GCC Liquid-Cooled Rack System Architecture Design Specification - Type A> and <GCC Liquid-Cooled Rack Interface Specification>, which focus on the standardization of liquid-cooled rack system architecture and interfaces, enable AIDC to be compatible with multiple types of AI liquid-cooled racks, and provide unified design basis and implementation guarantee for high-density intelligent computing deployment, have been officially released.

The <AIDC Infrastructure Specification>, which meets the demand for a single cabinet power of 300 kW in the next three to five years, is compatible with the evolution of 2 to 3 generations of whole cabinets, and realizes the decoupled design of the data center and cabinets, was officially released.

Three series of technical specifications focusing on the safe, reliable, and efficient operation and maintenance of liquid cooling systems, including <Technical Specification for Intelligent Operation and Maintenance of Cold Plate Liquid Cooling Systems>, <Key Specifications and Evolution Roadmap of CDU>, and <Process Specification for Secondary Side Pipeline>, have been officially released.

AISBench V3.0, the intelligent computing supernode, and ClusterBench V1.0, the general computing supernode, have been officially released, representing the core tool achievements of the Benchmark for Supernode Performance Benchmark.

The launch of the Open AI Infra Community AIDC Model Site Lighting Plan and the AIDC Model Site Certification Project demonstrates the benchmark leadership and standard certification for the large-scale implementation of AI infrastructure.

Insight into Technological Frontiers, Sharing Practical Insights —— Open AI Infra Community Technology Insights

During the main forum session of this conference, multiple Subject-Matter Experts from the Open AI Infra community took the stage to give speeches. Relying on the community's technological accumulation and practical experience, they shared the latest exploration results around the cutting-edge technological trends, core challenges, and implementation paths in the field of AI computing power facilities, providing professional guidance and reference ideas for the industry's development.

Gao Xiaojun, Co-Chair of the Open AI Infra Community Management Committee and from ByteDance, delivered a speech titled "Technological Trends and Challenges of the New Generation of Intelligent Computing Supernodes". He pointed out that currently, innovation in large models has slowed down while application scenarios continue to explode, and infrastructure has become the core barrier to the development of the AI industry. The key requirements of the new generation of intelligent computing supernodes cover multiple dimensions such as computing, storage, network interconnection, systems, and energy, and there is an urgent need to achieve elastic expansion and rapid deployment through open architectures and diversified computing power layouts. Gao Xiaojun emphasized that  only by relying on the Open AI Infra community platform, rallying the strength of the entire industry, and collaborating in key areas such as computing power and core component applications, system interconnection, heat dissipation, and power supply, and continuously promoting technology R&D and implementation, can we achieve the evolution of next-generation supernode technology and the development of the ecosystem.

Jing Tangbo, a member of the Technical Steering Committee of the Open AI Infra Community and the head of R&D and design management of ByteDance's Data Center, shared insights on "Technical Trends and Challenges of Gigawatt-scale Data Centers". He summarized the three core technical trends of gigawatt-scale data centers:the upgrade of hardware and AI computing platforms gives rise to a new ecosystem, liquid cooling technology moves towards mainstreaming and diversification, and the electrical system evolves towards 800V DC and electro-hydraulic integration.Meanwhile, he also pointed out that the large-scale construction of gigawatt-level data centers faces multiple challenges, including increased reliability risks brought about by high-density and ultra-large-scale deployment, insufficient resource utilization and prominent carbon emission pressure, misalignment between AIDC metrics and business computing power requirements, as well as compatibility and delivery cycle pressures caused by uncertainties in the hardware supply chain.

Members of the Community Technical Steering Committee mostly come from the frontline of the internet field, providing substantial support and professional advice for the community's technological development. As a representative of the members of the Technical Steering Committee, Jing Tangbo stated that the OAII Community Infrastructure Project Group has already made early arrangements, completed the reference design and technical specification formulation for infrastructure, and will continue to promote version iteration in the future, focusing on the evolution of core projects such as liquid cooling and power supply and distribution.

Long Pan, Co-Chair of the Open AI Infra Community Management Committee and Vice President of R&D of the Computing Product Line at Huawei Technologies Co., Ltd., shared insights on the topic of "Technical Exploration and Planning of the Megawatt-Level Computing Power System in the Open AI Infra Community".

He believes that the intelligent computing system needs to break through the limitations of single computing power and integrate diverse resources such as NPU, CPU, and NAND media through the supernode bus, thereby becoming an intelligent agent (Agent AI) with productivity.

In the future, the basic unit of AIDC will be a megawatt-level computing power system, on which a gigawatt-level Data center will be built. To achieve this goal, all areas of infrastructure urgently need to be reconfigured, and efficient power supply, closed-loop liquid cooling, and high-density optical interconnection will become the core foundation of the computing power system. Long Pan said,  The Open AI Infra community uses the three areas of power supply, heat dissipation, and interconnection as entry points to build AI infrastructure specifications, which precisely meet the needs of industry development and are of great significance for promoting the upgrade of computing power infrastructure.

Zhang Guangbin, the host of this conference and the Secretary-General of the Open AI Infra Community, mentioned thatthe technology stack related to megawatt-level computing power systems and gigawatt-level data centers can be summarized into 5 scales of K/M/G/T/P, including KW-level chips, KV-level power supply and distribution, MW-level systems (such as racks, CDU, and SST), GW-level AIDC, TB/s-level high-speed interconnection, PB-level storage (such as memory pools and SSDs), etc. The Open AI Infra Community is steadily carrying out related work, and welcomes members to actively participate!

At this conference, the "Megawatt-level Computing Power System" project of the Open AI Infra Community was launched. This project was jointly initiated by users such as China Mobile, ByteDance, Baidu, JD.com, and Kuaishou, computing power system partners such as Huawei, H3C, Inspur Hyper, and Ruijie, and community industrial partners in areas such as liquid cooling, power supply, and interconnection. It will unite the forces of the entire industrial chain in the community, focus on tackling the core technical bottlenecks of megawatt-level computing power systems, and help AI computing power infrastructure achieve higher density, higher efficiency, and larger-scale innovation breakthroughs.

Wang Guilin, a member of the Technical Steering Committee of the Open AI Infra Community and the Director of Hardware R&D at JD Cloud, delivered a keynote speech titled "Exploration and Practice of JD Cloud's Liquid Cooling Technology". In his view, liquid cooling technology has become an inevitable choice in the era of computing power, and multiple Tian Shu liquid-cooled full rack servers independently developed by JD have achieved large-scale deployment in several of its self-built data centers.As a founding member of the community, JD Cloud has deeply identified with the development philosophy of openness, integration, and co-construction of the Open AI Infra Community since its inception, fully participated in community building and standard setting, jointly led key tasks such as the liquid cooling working fluid standard, cold plate liquid cooling test standard, and AIDC reference design with industry partners, and, as a management team unit of the "Double Zero Initiative", collaborated and innovated with community partners to promote the rapid implementation of liquid cooling technology.

Wang Guilin said that in the future, JD Cloud will contribute JD Tianshu full rack technology design and implementation experience to the Open AI Infra community, helping the industry to unify liquid cooling technology standards and accelerate the large-scale implementation of the industry. He emphasized that JD Cloud will take technological innovation as the core driving force, and under the leadership of the Open AI Infra community, jointly build an open ecosystem with partners across the industry to jointly promote the upgrading and implementation of green computing power infrastructure.

Yan Kun, a member of the Open AI Infra Community Management Committee and vice president of 21Vianet, shared his insights on "Specification Leadership, Implementation First - 21Vianet's Overall AIDC Solution Planning and Practice". He mentioned that the core challenge faced during the AIDC construction process is how to adapt to the future needs of rapid chip iteration and upgrade at a lower cost. The various technical standards formulated by the OAII Community have formed a comprehensive system from dimensions such as modular design, power supply and distribution solutions to performance indicators, providing clear guidance for enterprise customized product R&D and architecture optimization, and enabling better technical adaptation. CenturyLink will fully embrace community infrastructure specifications, empower the implementation of AIDC construction based on standards, and the AIDC delivered this year has been fully designed in accordance with the standards, looking forward to creating the first model site based on the "AIDC Infrastructure Specification".

Liu Peiguo, the co-team leader of the Power Supply Project Team of the Data Center Infrastructure Project Group of the Open AI Infra Community Data Center Infrastructure Project and the head of the ICT Power Supply Solutions Department of Huawei Digital Power, shared on the topic of "Industrial Exploration and Reflection on AI Data Centers".

He proposed that computing power, as the foundation of the AI industry, is showing a trend of diversified development, with power consumption demand continuously rising. There is an urgent need to build friendly AI infrastructure that meets the industry's needs,  achieving chip friendliness, grid friendliness, environmental friendliness, and customer friendliness, ultimately achieving the goals of reliable power supply, efficient transmission, orderly heat dissipation, and protection of investment value . In the future, continuous innovation should be carried out around the 4T link to build a secure, reliable, resilient, agile, green, and low-carbon AIDC system.

Meanwhile, Liu Peiguo stated that Huawei Digital Power will rely on the OAII Community Platform, collaborate with upstream and downstream players in the industrial chain to improve the AIDC standard system, and jointly promote the maturity of the industrial ecosystem.

Li Yunjiao, a representative of the member unit of the Open AI Infra Community Management Committee and the Marketing Director of the Super Fusion Intelligent Computing Center Field, delivered a keynote speech titled "Technical Sharing on Open Computing AI Supernode Architecture".

He believes that diversified computing power has become the norm in the industry, with large model applications rapidly being implemented and the computing power ecosystems of the East and the West coexisting in the long term. Each generation of AI Infra needs to achieve precise matching between computing power throughput and latency. Super Fusion has always been committed to promoting the construction of computing power infrastructure with an open architecture, advocating for unified standard interfaces and specification systems,  and has contributed its diversified computing power architecture design to the Open AI Infra community. Currently, it is leading the formulation of the design specification - Type B for the GCC liquid-cooled full cabinet system architecture of the OAII community .

In the future, Hyper Fusion will collaborate with supply chain and industry manufacturers to deepen the design of the supernode architecture, fully meeting the core requirements of AIDC for computing power infrastructure.

Li Xinqi, the representative of the leading unit of the CDU project in the Open AI Infra Community and the Product Director of the IDC Business Unit at Envicool, delivered a speech titled "AIDC Liquid Cooling Technology Trends". Based on sharing Envicool's participation in community standard setting and related work achievements, he interpreted the new development trends of AIDC liquid cooling technology, namely that larger-sized cabinets, stronger cooling capabilities, higher power distribution support, and faster network transmission have become the core requirements, which also pose higher demands on large-granularity CDU, all-round liquid cooling solutions, and low-fluctuation and low-resistance liquid cooling systems.

Li Xinqi said that two-phase liquid cooling technology has the potential for disruptive innovation and is an important breakthrough direction for future high-end heat dissipation technology; at the same time, the ultimate collaborative design of the entire cabinet at the levels of computing, interconnection, power supply and distribution, and cooling is more in line with the development direction of next-generation AI infrastructure, which takes the entire cabinet as the smallest high-performance collaborative unit.

Wang Feng, Technical Expert of the AI Rack Project Group in the Open AI Infra Community and General Manager of the Cloud and Computing Storage Solutions Department of H3C Technologies Co., Ltd., gave the concluding speech titled "Super Nodes Empowering the Construction of a Diverse, Open, and Efficient AI Infra System".

He pointed out that  the Agentic era has arrived, with four characteristics - massive Token consumption, ultra-high concurrent access, inference tidal fluctuations, and intensive storage-computation interaction - reshaping the demand for computing power. The industry urgently needs to build a diversified, open, and efficient standardized AI Infra innovation architecture to achieve sustainable, high-performance, and low-cost inference services.

Wang Feng believes that the next generation of AI Infra is evolving from hardware stacking to deep integration, with integration becoming the core trend. H3C integrates multiple key technologies to build the AI Infra architecture for the era of intelligent agents, creating an efficient and stable computing power base for the industry. He emphasized that the development of AI infrastructure cannot be separated from collaborative innovation across the entire industry, and H3C is willing to join hands with GCC and various manufacturers to jointly promote the standardization of computing power architecture and jointly build a diversified, open, and efficient AI Infra ecosystem.

Multidimensional Exploration and In-depth Research: Joint Exploration of Technological Breakthroughs - Open AI Infra Community Theme Forum and Ecosystem Exhibition Area

In addition to the high-profile agenda of the main forum, the 2026 Open AI Infra Summit will simultaneously host multiple professional closed-door meetings and a series of themed forums, focusing on key areas such as high-speed interconnection, 800V HVDC power supply and distribution, advanced liquid cooling, supernode ecosystem, GW-level AIDC, and supernode performance testing. The conference will conduct in-depth exchanges on hot topics across the entire industrial chain from megawatt-level computing power systems to GW-level data centers, precisely tackle the core technical bottlenecks in the clustered deployment of computing power, and provide solid support for the industry to break through development dilemmas and achieve efficient upgrades.

The detailed and exciting content of the six thematic forums on high-speed interconnection, 800V power supply and distribution, full stack liquid cooling, super node ecosystem, GW-level AIDC, and super node evaluation will be presented to you one by one in special topics. Please stay tuned.

During the conference, a featured ecosystem exhibition area will be created, precisely building a platform for industry display and docking, bringing together over 30 leading enterprises and core institutions in the AI Infra industrial chain, and centrally showcasing cutting-edge innovative technologies, new architectures, and mature solutions in core fields such as computing power, interconnection, liquid cooling, power supply, and storage. All kinds of new technologies, new architectures, and new solutions will make a concentrated appearance and panoramic presentation, comprehensively demonstrating the innovative vitality and fruitful practical achievements of the intelligent computing infrastructure ecosystem, and building an efficient bridge for technical exchange and cooperation docking for all parties in the industrial chain.

From the results release and cutting-edge exploration in the main forum to the in-depth discussions and intellectual exchanges in the closed-door meetings and six thematic forums, the 2026 Open AI Infra Summit has built an efficient industrial communication platform with its diverse agenda. Every sharing session is a collision and integration of technologies, and every exchange is a gathering of strength within the ecosystem, jointly drawing a clear and promising new blueprint for the development of AI Infra in the industry.

Walking hand in hand with the Open AI Infra community, we open up and jointly create a new future for AI Infra. This summit, with openness as its foundation and collaboration as its wings, brings together the wisdom and strength of the entire industrial chain to jointly explore technological paths, jointly build standard systems, and jointly promote ecosystem prosperity. In the future, let us continue to work side by side, innovate and collaborate, continuously delve deep into the field of AI computing infrastructure, jointly build a new intelligent computing ecosystem that is efficient, inclusive, green, and sustainable, and jointly embark on a new journey towards high-quality development of the intelligent industry!

( end)