During the MWC Barcelona 2026 event, Chen Xinyu, Vice President of ZTE, introduced ZTE’s cutting-edge ‘AI Factory’ full-stack solution. Addressing the increasing demands of AI’s large-scale models, this innovative approach promises breakthroughs by overcoming traditional hardware constraints and optimizing total cost of ownership throughout the AI lifecycle.

Chen emphasized the need for architectural redesign to enhance computing efficiency, citing ZTE’s answer: a comprehensive full-stack co-design. This includes advanced AI servers, hyper-nodes, a seamless lossless network, and smart software solutions like the AI Booster OS and AI Agent Studio platform. ZTE’s compute-network integration aims to significantly improve computing density and performance.

A notable feature, the compute-network co-design, revolutionizes traditional hardware stacking. ZTE introduces vertical scale enhancements using the OEX architecture, eliminating excessive cabling and boosting compute density. By integrating a high-capacity switching chip, it supports ultra-fast connections, ensuring smooth operations without downtime. Additionally, horizontal scaling extends infrastructure across hyper-nodes for superior scalability.

Beyond hardware integration, software-hardware co-design enhances energy efficiency and system performance. Through strategic collaborations with major GPU providers, ZTE optimizes software frameworks to increase performance up to 20 times, significantly improving ‘tokens per watt’ efficiency.

Finally, ZTE highlights their open ecosystem philosophy, striving for industry collaboration. They aim to solve connectivity and delivery challenges in AI infrastructure, enhancing global industrial intelligence and technological advancement. By fostering partnerships, ZTE pursues a future-proof AI ecosystem catering to diverse industry needs.