AI‑Era Cloud Infrastructure: High Compute Density, Linear Scalability and Intelligent Operations – Highlights from the 2023 Open Data Center Conference

The Open Data Center Committee (ODCC) hosted the 2023 Open Data Center Conference in Beijing, co‑organized by Alibaba, Tencent, Baidu, China Telecom, China Mobile, the China Academy of Information and Communications Technology, Meituan, JD.com and others. Alibaba Cloud’s Intelligent Infrastructure Division presented its vision for the AI era, emphasizing that 2023 is considered the “year of Artificial General Intelligence (AGI)” and that AI workloads are reshaping cloud infrastructure requirements.

New‑Generation AI‑Era Cloud Infrastructure: High Compute Density, Linear Scalability

Alibaba Cloud General Manager Zhou Ming highlighted that large‑model AI workloads demand higher compute density and linear performance scaling, impacting data‑center network bandwidth, architecture, I/O, power density, cooling efficiency, and operations. Alibaba’s Lingjun intelligent compute cluster can scale to 100,000 GPUs, schedule tens of thousands of GPUs per task, and supports trillion‑parameter models using the next‑generation AI cluster network HPN 7.0.

Predictable Networks: The Key to AI Scale and Extensibility

Wang Chao, leader of the Lingjun cluster, explained that high‑performance networking is the “password” for AI model scale and extensibility. Alibaba’s HPN 7.0 implements end‑to‑end network integration with self‑developed hardware and software, addressing path selection, traffic control, fault recovery, monitoring, topology and architecture, and has been deployed in large‑scale predictable network products.

Intelligent Data‑Center Operations: Power‑Simulation System Improves Stability

Liu Guoliang described a real‑time power‑simulation and monitoring system that integrates power‑equipment telemetry, topology, simulation, change management and emergency response, dramatically enhancing data‑center power stability despite the complexity of redundant 24/7 power architectures.

Smart Battery Diagnostics Boost Operational Efficiency

Liu Wei presented intelligent diagnostics for data‑center batteries, enabling full‑state analysis of lead‑acid batteries, prediction of abnormal discharge, high internal resistance, and other health issues, thereby reducing manual inspection workload and supporting lithium‑battery intelligent diagnosis.

Liquid‑Cooling Innovation for Sustainable AI Compute

Senior expert Zhong Yangfan highlighted Alibaba Cloud’s green liquid‑cooling solutions that address the massive energy and heat demands of generative AI models like ChatGPT, contributing to carbon‑neutral goals and enabling sustainable high‑performance compute.

Modular Server Standards Accelerate Supply‑Chain Efficiency

He Yongbao introduced the Open Server Standardization Project (OSSP) specifications for motherboards and riser cables, which adopt a modular, minimal‑component design (CPU, memory, BMC) to improve scalability, cost efficiency, and supply‑chain resilience for server manufacturers.

Alibaba Cloud received multiple ODCC ten‑year anniversary awards for its white‑papers, network projects, high‑speed copper‑cable technology, and for outstanding project managers, underscoring its leadership in cloud and AI infrastructure innovation.

The conference concluded with a reaffirmation of Alibaba Cloud’s commitment to open collaboration, continuous innovation, and joint advancement of cloud computing and AI technologies for global economic and societal benefit.