CCF HPC2026 Opens in Shenzhen With Focus on Supercomputing-AI Integration

The conference was hosted by the China Computer Federation and jointly organized by CCF's Technical Committee on High Performance Computing and the National Supercomputing Center Shenzhen (Shenzhen Cloud Computing Center), with Tsinghua Shenzhen International Graduate School and Shenzhen University's College of Computer Science and Software Engineering as co-organizers. The High Technology Research and Development Center of the National Natural Science Foundation of China supported the event.

Organizers said more than 20 Chinese and foreign academicians and more than 500 HPC experts and industry executives took part in sharing sessions. The program included 11 invited reports, eight corporate reports, 50 thematic forums and 40 side events. It received 72 Chinese-language paper submissions via EasyChair and 49 English-language submissions to the CCF THPC journal. The exhibition covered more than 12,000 square meters and brought together 65 companies and frontier institutions, including National Supercomputing Center Shenzhen, Shenzhen University's College of Computer Science and Software Engineering, Huawei, Sugon, the High-Throughput Ethernet Alliance, Xin'aisuan, Lenovo, Jinan Supercomputing, Parallel Technology, Shishi Technology and Silang Technology. More than 5,400 industry guests registered to attend or exhibit.

At the opening ceremony on Oct. 18, Lu Yutong, conference chair, CCF Fellow, professor at Sun Yat-sen University and director of the National Supercomputing Center Shenzhen/Guangzhou, said computing power was flowing into research front lines, industrial sites and sectors. She said the National Supercomputing Center Shenzhen's Lingcheng fully domestic exascale supercomputer system ranked first in this year's TOP500 list. She added that the Guangdong-Hong Kong-Macao Greater Bay Area has national supercomputing centers in Guangzhou and Shenzhen and can combine them with Hong Kong and Macao's international research strengths and advanced manufacturing systems to explore new systems, applications and ecosystems for supercomputing-AI integration.

The opening also featured the release of the HPC+AI supercomputing-intelligence collaborative open community and the High-Throughput Ethernet (ETH+) Alliance's full-stack domestic intelligent computing network. The organizers said the releases covered frontier programs, open communities, domestic intelligent computing networks and computing applications, with the aim of building a supercomputing-AI integration ecosystem and landing domestic computing innovation.

In the keynote sessions, Liao Xiangke, academician of the Chinese Academy of Engineering and director of Qiyuan Laboratory, discussed preliminary practices and lessons in AI for Science, describing AI4S as a fast-developing fifth research paradigm and saying the laboratory is building a one body, three wings architecture and a Wuzhen base. Lu Yutong presented on a new generation supercomputing-AI integration system and scientific engineering applications. She said HPC plus AI for Science, combining numerical simulation with artificial intelligence, is becoming an important new paradigm for scientific discovery and that China's supercomputing advances from Tianhe to Lingcheng relied on independent innovation in system architecture.

Ding Zhaohui, chief technical expert for Kunpeng high-performance computing at Huawei, said HPC is evolving toward research agents in the AI era and that CPU value is becoming more prominent in the agent era. Cao Zhennan, vice president of Sugon and deputy director of the National High Performance Computer Engineering Technology Research Center, presented Sugon 8000's applications in earth science, materials, life sciences, astronomy, high-end equipment and AI. Fu Binzhang, co-chair of the High-Throughput Ethernet Alliance technical committee and head of Alibaba Cloud's high-performance network, described network evolution from training to inference, from HPN to TPN and from scale-out to scale-up, with the goal of integrating scale-out and scale-up and building a domestic intelligent computing network ecosystem.

Satoshi Matsuoka, director of R-CCS at RIKEN and a professor at Tokyo Institute of Technology, discussed beyond-exascale AI-quantum discovery engines for AI for Science for industry and listed solving intractable science problems, leading advanced science, starting from basic science and achieving societal impact as goals and challenges. Torsten Hoefler, professor at ETH Zurich and head of the Scalable Parallel Computing Laboratory, asked whether an AI climate scientist can be built and said AI can help supercharge the scientific method. Cao Jiannong, vice president and chair professor at Hong Kong Polytechnic University, discussed edge-AI-native service provisioning. Yang Zhi, co-founder of Xin'aisuan, said large-model inference is approaching an inflection point and outlined demand, efficiency, chips, form factors and capital. Xiao Limin, chief scientist at Lenovo, presented work on memory-computing fusion to accelerate large-model inference. Wang Xinhui, vice president of ZTE, called for an open, decoupled and secure AI ecosystem. Yang Canqun, researcher at National University of Defense Technology and director of the National Supercomputing Center Tianjin, discussed large-scale thermal-mechanical-chemical multi-effect coupled simulation.

The conference theme was Deep Integration of Supercomputing and Intelligence, Strong Computing Power in Shenzhen. Discussions focused on domestic computing foundations, two-way empowerment between supercomputing and large models, computing-network coordination and green intelligent computing. The organizers said China's computing industry is moving from scale expansion to high-quality development, with AI large models and AI for Science driving demand for intelligent computing, while the supercomputing internet, computing-power-electricity coordination, green low-carbon systems and full-stack self-controllable hardware and software remain core issues.