Designed for Agent workloads, we explore a new paradigm of AI-Native data infrastructure.
As foundation model adoption deepens, the AI industry is entering a phase of surging inference demand. Token costs, compute resource utilisation and infrastructure efficiency during inference have become critical factors constraining large-scale AI deployment. Faced with growing inference workloads, traditional compute infrastructure calls for innovative technical approaches. How to boost overall efficiency through breakthroughs in memory and compute architectures has become a key industry focus.
From super memory processors to the reimagining of intelligent infrastructure, the AI inference era is driving compute systems toward diversified architectural forms. How can super compute nodes and distributed computing resources coexist and work in synergy? How will emerging memory technologies reshape the organisation and utilisation of compute resources? Hongqiao International Science and Technology Week 2026 will centre on innovations in AI inference infrastructure and compute architectures. It examines the technical value of the Super Memory Processor Unit and the development roadmap of infrastructure to meet future intelligent computing demands.
Hongqiao International Science and Technology Week 2026 will take place at the National Exhibition and Convention Center (Shanghai) from November 20 to 22. As one of its six thematic stages, the Startup Ecosystem Stage brings together industry practitioners. Through keynote speeches and panel discussions, it explores the global tech startup ecosystem, corporate growth and financing, cross-border innovation and collaboration with international industrial capital.
The agenda covers startup case studies, investment and exit strategies, and scaling from early-stage ventures. It creates a dialogue platform connecting founders, investment institutions and strategic acquirers.
At this forum, Shuaijun Duan, Chairman and CEO of CoreArk (Shanghai) Integrated Circuit Co., Ltd., will deliver the keynote speech The Inference Era: Super Memory Processor Unit (SMPU) and the Restructuring of Intelligent Infrastructure. Addressing token cost challenges in the AI inference era, he will discuss how SMPU reshapes intelligent infrastructure, analyse the trend of coexistence between super compute nodes and distributed computing, and share technical insights and infrastructure evolution pathways for next-generation AI computing demands.
Keynote The Inference Era: Super Memory Processor Unit (SMPU) and the Restructuring of Intelligent Infrastructure The Token Cost Challenge in the Inference Era How Super Memory Processor Units Reshape Intelligent Infrastructure An Era Where Super Compute Nodes and Distributed Computing Coexist
Speech Date: November 22, 2026 Stage: Startup Ecosystem
Speaker Bio
Shuaijun Duan CoreArk (Shanghai) Integrated Circuit Co., Ltd. CEO
Shuaijun Duan holds a Master’s degree in Microelectronics and Solid-State Electronics from Jilin University. He currently serves as Chairman and CEO of CoreArk (Shanghai) Integrated Circuit Co., Ltd.
He previously worked at Renesas Semiconductor, AMD, the Research Center for Fundamental Software and General-Purpose Chips of the Chinese Academy of Sciences, CoreAlliance Technology and FAW Group. He has participated in and led the development and mass production of dozens of GPUs and APUs. He spearheaded key R&D for indigenous X86 CPU cores, and headed projects for L4 autonomous driving chips, AI processors and 3D high-bandwidth memory chips.
He led the development of the world’s first mass-produced 3D-stacked zero-cache AI accelerator chip and holds a portfolio of relevant patents.