On the 27th, LENOVO GROUP held its 2025 Innovation Open Day at its Beijing headquarters, showcasing the company's latest developments in AI PC and AI agent matrix. Lenovo's core focus remains on its AI PC strategy, while simultaneously opening innovation exhibition areas covering multi-dimensional products for consumer, enterprise, and industry applications.

Haitong International's Yao Shuqiao team released a research report stating that Lenovo's AI PC strategy and smart agent ecosystem are gradually forming synergistic effects. Combined with the massive traffic portal brought by approximately 400 million active devices, the company has diversified monetization pathways in smart agent collaboration, industry solutions, and AI+cloud services, opening long-term growth space for its AI+hardware strategy.

The research report provides specific analysis as follows:

**Comprehensive AI Application Matrix**

Lenovo has constructed a complete AI application matrix with "Tianxi Personal Super AI Agent" and "Lexiang Enterprise Super AI Agent," covering two core scenarios: personal and enterprise. Tianxi targets personal users, leveraging multimodal perception, cognitive decision-making, and autonomous evolution capabilities, combined with five major functions of AI control, search, translation, note-taking, and services to enhance learning, work, and life efficiency. Lexiang positions itself as an enterprise-level silicon-based employee, based on hybrid AI architecture and cross-departmental data scheduling, achieving process orchestration, task automation, and multi-form interaction, demonstrating value in customer management, consulting services, and business operations. Both agents operate in a unified multi-terminal approach with edge-cloud hybrid deployment, balancing intelligence and data security while promoting personal experience upgrades and assisting enterprise digital transformation.

**Complete AI PC Product Matrix**

Lenovo's AI PC matrix covers high-performance to entry-level full scenarios, forming the most complete commercial AI PC system. Its core is achieving computing power stratification through local heterogeneous computing power of CPU+NPU+GPU. High-performance dGPU models (ThinkPad T14p, Zhaoyang X7, ThinkStation P368, etc.) target complex AI tasks, high-end NPU models (ThinkPad X1, X9, X/T) meet mainstream government and enterprise applications, while entry-level NPU models (ThinkPad L, Zhaoyang X5, ThinkCentre/Qitian M) ensure popularization and cost-effectiveness. Through the synergy of three major brands - ThinkPad (high-end), Zhaoyang (domestication), and ThinkCentre/Qitian (government and enterprise scale) - Lenovo has built a full-line AI PC product matrix covering laptops, desktops, and workstations, forming leading advantages in government, enterprise, and industry markets.

**AI Integration in Moto Phone Series**

Lenovo has comprehensively integrated AI capabilities into its Moto phone series. The new generation of AI phones is equipped with Tianxi Personal Super AI Agent, supporting multi-scenario applications such as AI translation, AI calling, AI screen recognition, and AI travel, achieving full-time intelligent interactive experiences. In terms of hardware, the Razr 60 series adopts foldable design and diverse materials, combined with AI imaging functions to optimize group photos and shooting effects, enhancing user differentiated experiences. Meanwhile, lightweight models like edge 60 cover younger demographics, building a multi-layered product matrix. We believe that Moto AI phones expand Lenovo's AI application scenarios on the consumer side, strengthen user stickiness, and form a complementary ecosystem with AI PC and enterprise smart agents. Overall, this product achieves deep integration of intelligence and hardware, providing a new growth engine for Lenovo's AI+hardware strategy.

**Technology Platform Evolution: From Computing Power Scheduling to Smart Computing 3.0**

Lenovo launched the Wanquan Heterogeneous Smart Computing Platform 3.0, with its core being the achievement of more efficient large model training, inference, and applications through differentiated technology:

• Large Model Pre-training: Provides computing power matching, super-intelligent fusion scheduling, checkpoint resumption, and GPU kernel virtualization, significantly improving efficiency (reducing training costs and increasing utilization from 80% to 95%).

• Large Model Training and Inference: Strengthens AI inference acceleration algorithms (5-10x performance improvement), AI compilation optimizer (15%+ cost reduction), and expert parallel algorithms (3x reduction in inference latency).

• Key Breakthrough: Through integrated optimization of "pre-training—training—inference," comprehensively improves model development cycles, computing power utilization, and energy efficiency indicators.

**Hardware Foundation: Thirteen Years of Liquid Cooling Evolution**

From the 2012 iDataplex dx340M4 to the 2025 ThinkSystem M1380, spanning 6 generations. Heat recovery rate at 45-50°C improved from ~85% to ~97%, achieving single rack >180kW and 100% liquid cooling in 2025. Supports AMD, Intel, and NVIDIA CPU/GPU accelerated computing, compatible with CXL expansion. Liquid cooling technology is not just about energy saving and carbon reduction, but is key to addressing AI high-power training (hundreds of kW per rack level).