ZTE pivots to token-centric infrastructure at MWC Shanghai 2026

ZTE is rebranding its approach to infrastructure by prioritizing token efficiency as the primary metric for value creation. At the core of this strategy is the TCO-optimal AI factory, a system designed to minimize the cost-per-token through the integration of its SuperPod architecture. This hardware solution, which supports up to 16,000 GPUs using an Orthogonal Electrical eXchange design, enables high-density computing for large-scale model training and inference.

Beyond raw hardware, the company introduced NewStart AIOS to serve as the technological foundation for token scheduling and orchestration. This operating system supports the Co-Claw agent platform, which aims to bridge the gap between AI-native terminals and end-user applications. By integrating this software stack with network innovations like the AIR MAX and AI HI-NET solutions, ZTE intends to transform existing communication architectures into intelligent ecosystems capable of supporting 6G connectivity. This end-to-end framework covers the entire lifecycle of AI services, from initial compute provisioning to final service delivery.