On August 23rd afternoon, the 2025 China Computing Power Conference was held in Datong, Shanxi. Chen Zhenkuan, Vice President of LENOVO GROUP and General Manager of China Infrastructure Solutions Group, delivered a keynote speech titled "Lenovo's Full-Stack AI: Making AI a Productive Innovation Force" at the main forum. He also revealed that Lenovo is collaborating with authoritative institutions such as the China Electronics Standardization Institute to jointly promote the development of the industry's first high-performance training and inference service standard, and officially released the "2025 AI Solution Empowerment White Paper" at the conference.
Chen Zhenkuan stated that "computing power, models, and applications are the three main lines of AI development. Their integrated development jointly drives the AI trend through cycles." Lenovo has been comprehensively deploying artificial intelligence since 2017 and has now formed a full-stack AI portfolio including AI terminals, AI infrastructure, and AI solutions and services.
At the event, Chen Zhenkuan highlighted Lenovo's construction of diversified AI computing power infrastructure. In the computing power domain, Lenovo possesses a comprehensive AI server product portfolio covering large model training, inference, integrated training-inference, and edge inference scenarios, along with a "1+3+N" long-term development architecture - namely one system architecture, three CPU platforms, and compatibility with multiple GPU vendors. In the storage domain, the company provides both large-capacity, high-density, green and energy-efficient core storage solutions and has released the latest all-flash software-defined storage. In the network domain, Lenovo has launched the AI RoCE intelligent computing network solution.
Chen Zhenkuan introduced that "Lenovo's ThinkSystem heterogeneous intelligent computing platform has been upgraded to version 3.0. Building on the original five differentiating technologies, it has newly released four innovative technologies, providing precise support for new trends in model development." Based on this foundation, Lenovo has also launched three major management platforms targeting cluster computing power optimization, AI integrated machine management optimization, and high-performance computing optimization.
Currently, the innovative technologies of Lenovo's ThinkSystem heterogeneous intelligent computing platform have been validated in customer scenarios. In national-level high-quality AI cluster scenarios, Lenovo has collaborated closely with the largest intelligent computing hub in the "East Data, West Computing" initiative, improving model floating-point computation utilization from 30% to 60% in thousand-card training scenarios. For enterprise AI infrastructure scenarios with local model deployment, the full-speed operation of the full-performance DeepSeek R1 model achieves extreme throughput exceeding 12,000 tokens per second.
Simultaneously, in partnership with the Heterogeneous Intelligent Computing Industry Ecosystem Alliance and numerous ecosystem partners, the latest "2025 AI Solution Empowerment White Paper" was jointly released. The content encompasses Lenovo's understanding and insights into the latest industry trends and technological innovations, as well as technological innovations and practices completed in collaboration with ThinkSystem ecosystem partners.
Chen Zhenkuan stated, "The AI trend is unstoppable. We hope that computing power providers, model developers, and application enablers will deeply integrate to jointly promote intelligent transformation and achieve inclusive computing power."