Alibaba Cloud Launches Lingjun Zhenwu M890 Super Node Instance and Opens Ulanqab Beta Testing

Summary

At the 2026 World AI Conference, Alibaba Cloud launched the Lingjun Zhenwu M890 Supernode instance, the first of its kind offered via public cloud. It features 64-card interconnect, 800GB/s bandwidth, and supports FP8/FP4 precision, capable of handling 10-trillion parameter MoE model inference Zhitong. The service is currently open for invitation testing in the Ulanqab region Zhitong.

Impact Analysis

Alibaba is signaling that the AI arms race has moved beyond simple chip counts to “Supernode” architectural dominance. By launching the Zhenwu M890 with 64-card interconnect and 800GB/s bandwidth Zhitong, they are positioning themselves to host the next generation of 10-trillion parameter MoE models. While the market is buzzing about the Qwen-Apple integration China Finance Online+ 2, the real story is this vertical integration. Alibaba is effectively building a closed-loop ecosystem—from proprietary silicon and storage partnerships with CXMT Wallstreetcn to public cloud deployment.

This isn’t just a tech demo; it’s a direct challenge to Huawei’s Ascend 950 科创板日报 for the soul of China’s enterprise AI. By offering this as a public cloud service Zhitong, Alibaba lowers the CAPEX hurdle for vertical leaders like FAW etnet. I’d read this as a shift from commodity cloud to high-margin, AI-native infrastructure. The key metric to watch is the Ulanqab testing feedback; if the interconnect performance scales as promised, Alibaba secures its status as the primary “AI foundry” for China’s LLM developers.