ZTE unveils full-stack and full-scenario intelligent computing infrastructure at MWC Shanghai 2024

ZTE showcased its full-stack and full-scenario intelligent computing infrastructure for large model training and inference

ZTE is committed to fostering an open and decoupled AI ecosystem.

ZTE's full-stack and full-scenario intelligent computing infrastructure

Shanghai, China, 26 June 2024 - ZTE Corporation (0763.HK / 000063.SZ), a global leading provider of integrated information and communication technology solutions, has showcased its full-stack and full-scenario intelligent computing infrastructure at MWC Shanghai 2024. This initiative has laid a solid digital foundation for industries and spearheaded the AI+ era.

Driven by the wave of digital intelligence, the demand for massive computing power continues to grow, making intelligent computing power a pivotal force in advancing the digital economy. With respect to the hardware, ZTE's solution offers a comprehensive range of intelligent computing products across all scenarios, fully compatible with industry-leading GPUs, providing users with diverse options and enhancing the performance and stability of computing resources through holistic optimization of computing, storage, and network capabilities.

In terms of software platforms, ZTE provides an efficient and user-friendly AI platform that enhances engineering capabilities for large model training and inference through comprehensive toolkits. This reduce barriers for tasks such as large-model training, fine-tuning, and inference deployment.

ZTE provides a full-stack intelligent computing product that integrates leading performance and independent innovation to meet the requirements of diversified intelligent computing centers across all scenarios.

Full Series and Full Scenarios: ZTE offers high-performance training servers, cost-effective inference servers, integrated cabinets for out-of-the-box training and inference, a comprehensive series of general computing servers, high-performance file storage, and high-performance RoCE switches. These products cater to the needs of constructing diversified intelligent computing centers from the core to the edge.

Ultimate Price/Performance Ratio: ZTE supports resource pooling, virtualization, unified management and scheduling of heterogeneous computing power, and dynamic adjustment of training and inference resource ratios to maximize the utilization of intelligent computing resources.

Green and Energy-saving: ZTE adopts intelligent busbars and implements energy storage pools to achieve adjustable and flexible power consumption of cabinets. The data center incorporates global PUE management, including global thermal design in the design phase, optimal airflow design in the solution phase, customized flow channel design, and real-time AI optimization in the operation and maintenance phase. Additionally, by combining high-efficiency working media such as oil-based fluids, fluorinated liquids, and deionized water in liquid cooling servers, along with heat dissipation and diversion optimization technologies of the indium chip, ZTE achieves a final PUE reduction to 1.1.

Open Ecosystem: ZTE continuously promotes software/hardware decoupling, training and inference decoupling, and model decoupling to build a fully open and decoupled intelligent computing ecosystem.

ZTE will continue to advance its exploration in the field of intelligent computing, collaborating with global operators and industry partners to drive the advancement of AI technologies and facilitate the digital and intelligent transformation across various industries.