Co-hosted by the Shenzhen Memory Industry Association and the School of Integrated Circuits, Peking University and organized by JWinsights (Shanghai) Technology Co., Ltd., the Fourth GMIF2025 Innovation Summit (Global Memory Innovation Forum) was successfully concluded on September 25, 2025. Under the theme "AI Applications, Innovation Empowered", the GMIF2025 brought together leading representatives from across the memory industry value chain. Discussions centered on critical topics such as computing-storage convergence, AI deployment, and ecosystem collaboration.
The 4th GMIF2025 Global Memory Innovation Forum concluded on Sept 25. Credit:GMIF
The GMIF2025 officially kicked off with the opening remarks from Rixin Sun, the President of Shenzhen Memory Industry Association (SMIA). Mr. Sun shared that with the rapid proliferation of AI technologies on the edge side in 2025, the memory and storage industry are stepping into the spotlight. The industry growth mode is shifting into a new structural phase characterized by high performance, high efficiency, and system collaboration. Since its inception, GMIF has upheld the vision of "connecting the memory ecosystem and fostering industry win-win collaboration," aiming to unite global industry forces, drive cross-sector cooperation, and accelerate the integration of China's memory sector into the global landscape.
GMIF2025 opens with remarks from SMIA President Rixin Sun. Credit:GMIF
Industry leaders discuss AI-driven memory transformation and ecosystem integration. Credit:GMIF
Industry Leaders Gather to Discuss AI-Driven Transformation of Memory and Ecosystem Integration
Xiaomin Han, General Manager of Consulting Services from JWinsights, noted that the global memory market is entering an AI-driven upcycle, with smartphones, data centers, and new energy vehicles emerging as key growth drivers. In spite of the supply bottlenecks of high-end advanced chips, strong investment momentum from domestic internet companies is fueling faster-than-expected growth in the server market, which is projected to achieve large-scale commercialization by 2027. Meanwhile, Chinese players such as GigaDevice and UniIC are strengthening their presence in smartphones, servers, and edge AI through customized solutions, and are expected to make breakthroughs in emerging sectors by year-end.
Guangyu Sun, Professor of School of Integrated Circuits, Peking University, highlighted that to improve computing efficiency, it is essential to optimize pathways, adopt hierarchical architectures, and implement fine-grained resource scheduling to balance bandwidth, capacity, and latency. In diverse scenarios such as smart devices and edge computing, system efficiency, security, and cost are the key factors determining technological deployment. As advanced packaging and related technologies become increasingly practical, the next phase of industry development should focus on precision control, heterogeneous integration, and ecosystem collaboration to enable large-scale commercialization.
Kevin Yoon, VP and CTO, Memory Business Division, Samsung Electronics, pointed out that AI advancements are accelerating the shift in memory and storage toward higher performance, density, and energy efficiency. With data processing skyrocketing and system power nearing its ceiling, memory is now central to intelligent infrastructure. Leveraging interface upgrades, advanced packaging, and new materials, Samsung has rolled out products like GDDR7, delivering across-the-board gains in performance, capacity, and power efficiency.
Maya Zhang underlined that data requirements in the AI era are characterized by high volumes, high performance and structural diversity. It's predicted that global data is expected to reach 200ZB by 2025, and unstructured data will account for about 80% of that total. The rise of multimodal data further increases complexity. The growing "heat" of data is accelerating the shift of storage architecture toward computing-storage convergence, with heightened demands on speed, capacity, and power efficiency.
Wallace C. Kou, CEO, Silicon Motion, stressed that AI is fueling strong demand for large-capacity storage in both cloud and edge environments, with significant market growth expected next year. Industry players must strengthen collaboration in technology innovation, talent development, and ecosystem building.
Benny Ni, GAR Sales VP, Solidigm, observed that AI is moving from model building to terminal applications. Edge computing and data security are becoming increasingly vital, and storage technologies need to meet the diverse demands spanning from large capacity to high performance.
Storage and memory at the edge side emphasize more on intelligence, reliability and green innovation beyond performance. By adopting a "subtractive" approach, Maxio Technology achieves a significant balance between quality and cost, and is making more efforts to strengthen AI storage ecosystem through industry chain collaboration. Mr. Li summarized Maxio's vision as "One Chip, One Ecosystem, One Future", enabling the intelligent world with memory and co-creating a brighter future.
Haibing Xie, Director, Application Design In Center, Intel China, introduced that Intel has taken a dual-pronged AI strategy of hardware (AIPC, data center) and software ecosystems. Intel is working with partners to build an open supply chain supporting over 900 large models and enhancing user experiences with voice control and other technologies to drive AIPC adoption.
Xiaobing Wang, Director of Customer Projects from MediaTek, identified imbalances in edge AI computing power and efficiency as a major bottleneck, alongside thermal and bandwidth pressures. MediaTek is advancing its system through three key approaches: ensuring stable high-load performance with advanced cooling, reducing power consumption through process iteration, and lowering bandwidth dependency via algorithm and architectural collaboration, so as to achieve comprehensive breakthroughs through industry-wide cooperation.
Industry leaders discuss AI-driven memory transformation and ecosystem integration. Credit:GMIF
Han He, CEO of BIWIN, shared that BIWIN is transitioning from a module manufacturer to a solutions provider. With AI infrastructure investment projected to reach USD 3 trillion by 2030, BIWIN is leveraging its core expertise in media, controller and firmware R&D, and advanced packaging and testing. By continuously promoting its "Integrated Solution and Manufacturing (ISM)" strategy, BIWIN has established an all-around industry layout, aiming to deliver scenario-driven system solutions and accelerate computing-storage convergence.
Zining Wu, Founder & Chairman, InnoGrit Corporation, highlighted that the surge of AI-generated data is forcing a radical shift in storage architecture. To match the processing power of GPUs, the industry is rolling out "AI SSDs" capable of hundreds of millions of IOPS and ultra-low latency. Technologies such as GPU Direct enable direct communication with GPU memory, extending HBM in data centers and delivering deterministic responses at the edge. Storage media are being divided into high-speed cache layers and large-capacity storage layers, forming a new generation of high-performance, hierarchical architectures that provide a solid foundation for computing breakthroughs.
Matt Bromage, Head of Global Storage Business from Arm, projected that 70% of AI inference workloads will soon be processed at the edge. In response to the three key challenges of bandwidth, power cost, and software fragmentation faced by edge AI, the industry circle is making concrete efforts to comprehensively improve edge computing efficiency by innovating technologies including near-data computing, CXL high-speed interconnects, and chip-level security, combined with customized chip design, model quantization, and software-hardware co-optimization. These advances are expected to drive a thousand-fold leap in edge AI performance, accelerating the arrival of the Ambient Intelligence era.
Xuewen Chen, Chief AI Scientist from GAC Group, showed that AI has moved beyond its role as a driver-assistance tool to become a central engine of industrial change. The intelligent technologies based on big data and Transformer architecture are reshaping industry competitiveness, which raises unprecedented new expectations for automotive storage in terms of capacity, speed and reliability, and creating a broad space for storage innovation and deployment in intelligent vehicles as well.
Cheng Zhu, Product Director at China Greatwall, expressed that computing power, as a core productivity, is closely related with national economic development. Ranked as the 2nd largest computing capacity worldwide, China demonstrated a 30% growth rate in the past 5 years with over 10 million data center racks being deployed. AI computing has increased significantly. Underpinned by its full-stack technologies, China Greatwall has developed 120+ products and contributed to major national projects, providing secure computing foundations for government, finance, and other sectors.
Shawn Xiao, Director of Ecosystem Partnerships and Field Applications from Montage Technology, considered CXL as an ideal solution for optimizing memory architecture, lowering data center costs, and improving AI computing efficiency. By enabling efficient sharing and pooling of memory resources, it breaks through the bottlenecks of the traditional memory wall and provides a more flexible and scalable solution for high-performance computing and large-scale AI applications.
Shang Shang, General Manager of the Information Security from iFLYTEK, introduced iFLYTEK Spark AI PC, positioned as an AI office assistant for government and enterprises. Through integration of generation, summarization and insight capabilities, it fosters a new paradigm of human-machine collaboration, empowering users to focus on decision-making and innovation.
Ming Zhao, General Manager from OKN Technology, pointed out that against the backdrop of surging storage demand driven by AI, requirements for capacity, speed, and power consumption are becoming increasingly stringent, and specialized testing technologies are essential to ensure high-quality industry development.
Lucas Lu, Director of R&D Department from GreaTech Substrates, concluded that AI applications are driving semiconductor market growth, with a projected CAGR of 6.7% from 2025 to 2028. AI servers, smartphones, and PCs are the key drivers. Packaging technologies are rapidly evolving toward 2.5D and 3D, and heterogeneous integration solutions such as CoWoS are being widely adopted. This trend places higher demands on packaging substrates, including larger sizes, higher layers, thinner profiles, and finer wiring.
GMIF2025 offered a full-spectrum view of the latest advances in technology. Credit: GMIF
The GMIF2025 of the year presented a full-spectrum view of the latest progresses in technological breakthroughs, product innovations and ecosystem co-building of the storage and memory industry within the AI era. With a particular focus on advanced industry trends such as AI applications, storage-compute convergence, advanced packaging, and edge intelligence, the summit served as a high-level platform for industry stakeholders to gain insights, exchange technology, and explore partnerships.
By bringing together executives from leading global enterprises, academic experts from top universities, and professional attendees, GMIF2025 not only demonstrated the strong momentum of collaborative innovation in the industry, but also provided crucial support for the deep integration of storage technologies with AI, cloud computing, and smart devices. It further contributed to the creation of an open, collaborative, and sustainable global storage ecosystem, helping accelerate progress in the data-driven era.
Article edited by Joseph Tsai
