From hardware to services! Lenovo (00992) released four key achievements including the first high-end large model training AI server and AI factory solution.

On December 5, 2025, the Lenovo Heterogeneous Computing Summit Forum and New Product Launch event came to a close. Lenovo China Infrastructure Business Group (referred to as "Lenovo ISG China") simultaneously released four core achievements, including the "AI Factory" solution, the first high-end large model training AI server Lenovo Wentian WA8080a G5, the All-heterogeneous Computing Platform 4.0, and the industry's first training and reasoning service standard. Vice President of LENOVO GROUP and General Manager of Lenovo China Infrastructure Business Group, Chen Zhenkuan, stated at the conference, "From rapid growth, rational choice, to resurgence, we have experienced the technological breakthrough brought by generative AI together. At this moment, we are witnessing AI becoming a true expert in intelligent applications in a more autonomous and intelligent way. Facing the unique development status of the local industry, diverse heterogeneous computing power supply, highly coordinated core model application ecosystem, are all becoming the core forces to build China's unique AI ecosystem." It is worth noting that over the past two and a half years, Lenovo China's infrastructure business revenue has increased by 175%, and its server business has maintained a leading position in the Chinese market, becoming one of the fastest-growing AI infrastructure manufacturers in the Chinese market. The release of these four core achievements jointly constructs an AI service ecosystem with "hardware support + platform optimization + service implementation + standard guidance," marking Lenovo's departure from the valuation constraints of traditional hardware manufacturers and fully transitioning into a new track of technology service PS (Price-to-sales ratio) valuation. Combining hardware and software AI product portfolio, the Lenovo "AI Factory" released this time is a core strategic solution that transforms complex and isolated AI development tasks into a standardized and efficient modern "AI production line." This produces "finished products" that can be directly delivered, including industry intelligence bodies, domain-specific large models, and reasoning services. As the cornerstone of the AI Factory's computing power, the Lenovo All-heterogeneous Computing Platform 4.0 released this time has become an expert in full-scenario optimization of large model performance through the new upgrade of four major technical scenarios, directly enhancing the efficiency and industry adaptability of Lenovo's AI services: 1. In the large model pre-training scenario, model training time can be shortened by 35% through long sequence parallel optimization. 2. In the large model post-training scenario, computing engine optimization, small-sample reinforcement learning, and automatic parallel training can shorten the training time by 50%. 3. In inference scenarios, ROCE network load balancing optimization can improve bandwidth utilization by 60%, communication primitive performance by 30%, and large model inference performance by 30%. 4. In the super-intelligent fusion computing scenario, it supports international and domestic hardware ecology, completes comprehensive optimization from the underlying architecture to operators, and adds more than 16 job templates and scripts for manufacturing applications, achieving efficient super-intelligent fusion. From landing cases, the Lenovo All-heterogeneous Computing Platform has demonstrated its service value: in the "Dong Shu Xi Suan" intelligent computing hub cooperation, the MFU (Model Computing Utilization) training scenario has increased from 30% to 60%, significantly improving the efficiency of computing resource utilization and helping customers reduce costs. In the enterprise AI infrastructure scenario, it supported the DeepSeek R1 model to break the limit throughput of 12,000 Tokens/s, setting a new industry record. These achievements directly Lenovo's AI services' "differentiated advantages" - higher efficiency, broader adaptability, attracting more industry customers, driving the scale growth of service revenue. The logic of "efficiency improvement cost optimization service premium" is the key support for high PS valuation in the technology service industry. The Lenovo Wentian WA8080a G5 server was also introduced at the conference, further consolidating Lenovo's leading advantage in the training and reasoning integrated architecture. As Lenovo's first high-end large model training AI server, this product uses the latest Intel Xeon 6 processors (with more cores, faster memory, and each core has built-in AI acceleration), supporting the localization mainstream latest OAM2.0 GPUs, meeting the high-performance computing power requirements of large model training at the kilocalorie level, while being compatible with both air-cooled and liquid-cooled dual-mode heat dissipation, adapting to different scale data center deployment environments, and reducing customer hardware maintenance costs. Zhou Tao, General Manager of Lenovo China Infrastructure Business Group's Server Division, emphasized that this server was "built based on the real needs of local users." In addition to the above product portfolio, Lenovo, together with many industry partners, released the industry's first model training and reasoning service standard, "Information Technology Computing Service High-Performance Training and Reasoning Service Capability Requirement" standard verification results, with Lenovo as the main editing unit, fully participating in the standard formulation process. Chen Zhifeng, Director of the Cloud Computing Research Lab of the Software Application and Service Research Center of the China Electronics Technology Standardization Institute, believes that the core value of this standard lies in filling the industry gap and providing critical standard support for the standardized landing of the AI Factory. It not only provides enterprises with a unified architecture reference but also promotes the standardization and scaling of AI production from scattered exploration to standardization and scaling, accelerating the industry's intelligence transformation. Switching valuation logic from "stacking hardware" to "optimal service" The Lenovo All-heterogeneous Computing Platform 4.0 and the Lenovo Wentian WA8080a G5 server released this time not only enhance hardware performance but more importantly provide solid infrastructure support for AI services. Lenovo is transforming its "value anchor" through the execution of a hybrid AI strategy. The valuation logic of traditional manufacturing industry focuses on revenue, net profit, and asset turnover of hardware business, while the valuation logic of technology services values service revenue growth rate, gross profit margin, customer lifetime value (LTV), and customer acquisition cost (CAC). In the past, Lenovo's hardware business mainly focused on "one-time sales," with revenue tied to hardware shipments, and gross margins greatly affected by the manufacturing industry cycle, anchoring the valuation in the low PE range; the "AI Factory" through the "service subscription + customized delivery" mode, provides continuous AI development support, model optimization, and operational support for customers, turning revenue sources towards "continuous service cash flow." For example, enterprises do not need to build a complete AI team and computing center, but can obtain AI capabilities tailored to their own scenarios through service cooperation, while Lenovo gains stable repeat purchase revenue through long-term service - this high margin, high stickiness service attribute is the core support of PS valuation logic in the technology service industry, completely freeing from the valuation constraints of traditional hardware business. In the past three years, large model technology has experienced a global capacity leap; since the end of 2023, this technology wave has already transformed into sustainable, quantifiable, and scalable computing demands in China. IDC data shows that the AI server market in China reached $16 billion in the first half of 2025, doubling year-on-year; by 2029, the Chinese AI server market is expected to exceed $140 billion; in 2024, the overall size of the Chinese intelligent computing service market reached $5 billion, and will increase to $7.95 billion in 2025, with a five-year compound annual growth rate of 57.3% from 2023 to 2028. These data collectively indicate that China's computing infrastructure has transitioned from a technological exploration phase to a deployment phase. Whether it is adjusting industry models, deploying intelligent entities, building knowledge bases, or integrating models internally, computing is shifting from "one-time investment" to "continuous operation." In this rapid rise cycle, Lenovo ISG China demonstrates strong growth momentum, with its business revenue increasing by 175% over the past two and a half years, its servers maintaining a leading position in the Chinese market, and becoming one of the fastest-growing manufacturers in the development of AI infrastructure in China. Currently, Lenovo's AI services have shown high growth potential: the implementation of the training and reasoning service standard will accelerate industry demand release; the replicability of the AI Factory solution is expected to quickly penetrate various industries; the underlying support provided by the All-heterogeneous Computing Platform and the Wentian server ensures service delivery capabilities. As the proportion of service revenue in total revenue continues to rise, LENOVO GROUP's profit capability will transition from the low-margin level of the manufacturing industry to the high-margin level of technology services, and the market's valuation multiple will also switch from the manufacturing industry PE (usually 10-20 times) to the technology service PS (usually 30-50 times). This release marks Lenovo's strategic transformation from "hardware manufacturer" to "AI service leader," where its core value is no longer in hardware production capability but in standardized AI service capability, industry barriers, and scaling potential. For the capital market, ignoring this shift in valuation logic will mean missing the opportunity to reassess Lenovo's value - in the future, Lenovo's market value will be more determined by the growth rate of its AI service revenue and market share, signaling a new chapter for "AI Lenovo" in the capital market.