Chinese technology giant Lenovo Group has launched its Wanquan Hybrid Token Factory initiative in China, marking a major expansion of its artificial intelligence (AI) infrastructure business. Announced on November 10, 2026, in Zhangjiakou, Hebei province, the initiative aims to commercialise AI token production and make large-scale computing resources more accessible to businesses.
Lenovo also inaugurated its first Wanquan Hybrid Token Factory facility at the Beijing-Tianjin-Hebei computing power hub node in Zhangjiakou. The project comes as demand for AI computing capacity increases across China, driven by the growing adoption of generative AI, large language models and AI-powered enterprise applications.
The initial phase of the facility will feature a computing cluster with a capacity of 15,000 petaflops and an intelligent computing centre with approximately 100 megawatts of power capacity. Once fully operational, the facility is expected to generate more than two quadrillion tokens annually, with peak processing capacity reaching 200 million tokens per second. It will also support up to 100,000 simultaneous users.
Lenovo expects the facility's key performance indicators to improve more than tenfold by 2030, reflecting its ambition to establish an industrial-scale AI production ecosystem.
The Token Factory concept treats AI computing as a production system in which tokens generated by AI models represent measurable output. Lenovo aims to integrate computing infrastructure, AI models and cloud services to help organisations move beyond experimental AI projects towards commercial deployment.
The initiative builds on the company's industry framework outlined in its June 2026 monograph, Token Factory: The Wealth Production Paradigm of Token Economy. It focuses on four areas: Token Factory, Intelligent Production Capacity, Effective Computing Power, and Secure and Controllable AI Infrastructure.
The facility will operate through Lenovo's xCloud intelligent cloud platform, which manages private and public cloud resources and allocates computing capacity according to workload requirements. Its hybrid architecture combines different computing resources, chips and AI models, including open-weight and closed-source systems.
Through intelligent model routing, the platform directs complex tasks to larger AI models while assigning simpler requests to smaller models. This approach is expected to improve computing efficiency, optimise resource utilisation and reduce operational expenses.
Lenovo is targeting a reduction of more than 70% in the cost of processing one million tokens. It also aims to maintain a token output rate above 99.9% and a power usage effectiveness ratio below 1.1 across its data centre operations.
The initiative arrives as China's AI industry experiences increasing demand for inference capacity, which enables trained models to generate responses, produce content and complete tasks. According to China's Centre for Information Industry Development, daily AI token usage in the country has exceeded 175 trillion.
Lenovo's Token Factory forms part of its broader three-tier AI architecture. Its Wentian portfolio provides underlying infrastructure, while Qingtian agent services target government and large enterprises, and Baiying services cater to small and medium-sized businesses.
With the launch, Lenovo is positioning itself to capitalise on China's expanding AI infrastructure market by helping enterprises access computing power, optimise token generation, and scale AI applications without maintaining extensive in-house infrastructure.