Alibaba Cloud Accelerates RMB 40 Billion Shanghai Computing Center with Xuanwu Chips
On March 9, 2026, the Jinshan District Government of Shanghai and Alibaba Cloud formally signed a strategic cooperation agreement to accelerate construction of the Alibaba Feitian Cloud Intelligence East China Computing Center (hereinafter referred to as "Alibaba Cloud Jinshan Computing Center"), which will create an ultra-large-scale computing facility based on "Xuanwu" chips.
The collaboration between Jinshan District and Alibaba Cloud dates back to 2021, when the RMB 40 billion (US$5.5 billion) Alibaba Cloud Jinshan Computing Center first broke ground in the district. The project covers approximately 510 acres with a planned construction area of roughly 600,000 square meters. Upon completion, it will deploy hundreds of thousands of servers, positioning it to become one of the largest AI computing hubs in the East China region and supporting Shanghai's development as an artificial intelligence highland.
Under the new cooperation agreement, the center will prioritize deployment of Alibaba's chip subsidiary T-Head's "Xuanwu" computing chips. By building a fully autonomous computing infrastructure spanning from underlying chips and intelligent computing platforms to upper-layer applications, the project aims to establish a benchmark for green computing centers.
Leveraging the Alibaba Feitian Cloud Intelligence East China Computing Center, Alibaba Cloud and Jinshan District will jointly advance computing ecosystem development and support the district's construction of an all-time, all-domain AI city. Alibaba Cloud will comprehensively assist Jinshan District enterprises with intelligent upgrades, deeply participate in the digital transformation of government services and urban governance, support construction of secure and stable government service infrastructure, and apply AI technology to improve real-time response efficiency of public services.
In late January 2026, T-Head's official website released its latest Xuanwu 810E high-usability AI training and inference integrated chip. The chip employs a self-developed parallel computing architecture and ICN (Inter-Chip-Network) interconnection technology, coupled with a fully self-developed software stack to achieve hardware-software integration. The chip integrates T-Head's proprietary "Xuanwu" processing unit (PPU), with 96GB HBM2e memory and inter-chip interconnection bandwidth of 700 GB/s. It can be used for AI training, AI inference, autonomous driving, and multimodal model scenarios. Multiple 10,000-card clusters have been deployed on Alibaba Cloud, capable of meeting complex business requirements.