ZTE unveils NewStart AIOS and AI infrastructure at MWC Shanghai 2026

ZTE used MWC Shanghai 2026 to present a broad AI stack built around lower token production costs, AI service orchestration and future network infrastructure, while also using the event to frame its wider corporate AI strategy. The company launched NewStart AIOS, which it described as the "technological foundation in the AI era," and paired it with its Co-Claw enterprise-level agent platform and a range of AI devices aimed at enterprise, home and individual use cases. The company said its AI factory approach focuses on improving token efficiency, a metric it argued is becoming more important as AI agents spread and industry competition shifts from raw computing scale to the economics of inference. To support that, ZTE showcased its SuperPod system, low-cost inference architecture and an AIDC (AI data center) design centered on energy efficiency. SuperPod uses multi-chip open collaboration design and Orthogonal Electrical eXchange, or OEX, architecture to support plug-and-play deployment. ZTE said one rack can support 128 GPUs and scale to 16,000 GPUs. It also said its inference setup combines a non-blocking network with GSLB (Global Server Load Balancing) to coordinate computing clusters from thousands to tens of thousands of GPUs. On the network side, ZTE tied its AI push to 6G development and SAGIN (space-air-ground integrated network), highlighting LEO satellite communication payload solutions, its GigaMIMO technology and what it called the world's first 256 TR U6G prototype. It also promoted AIR MAX, AI HI-NET and a 10G AI-Optical Network solution, alongside applications including AI-assisted calling, ISAC (integrated sensing and communication) for the low-altitude economy and highly autonomous networks. Separately, Cui Li, Chief Development Officer, said ZTE is pursuing an "All in AI, AI for All" strategy to embed AI across products and solutions while shifting toward a data-driven organization centered on human-machine collaboration. ZTE said it is building a resilient AI system to handle uncertainty across four areas: openness and decoupling, flexible scalability, extreme synergy and a scenario-first approach, and said it sees AI ushering in a new phase of human-AI symbiosis.