Key Highlights
- Alibaba partnered with China Telecom to activate a 10,000-chip artificial intelligence computing facility in Shaoguan, located in Guangdong province, utilizing Alibaba’s proprietary Zhenwu semiconductor technology.
- This facility represents the inaugural large-scale Zhenwu chip deployment within China’s Greater Bay Area, engineered to handle AI model training involving hundreds of billions of parameters.
- The computing facility delivers a 30% improvement in both training and inference performance, with individual card throughput achieving nearly tenfold gains over earlier technology.
- Expansion plans target scaling from 10,000 to 100,000 chips, while smaller enterprises can purchase computing capacity through China Telecom’s distribution platform.
- This development comes shortly after a comparable 10,000-chip facility powered by Huawei Ascend 910C processors became operational in Shenzhen the previous month.
Alibaba (BABA) has partnered with China Telecom to bring online a substantial 10,000-chip artificial intelligence computing cluster situated in Shaoguan, Guangdong province. The entire infrastructure runs on Alibaba’s internally developed Zhenwu AI processors, created by its T-Head semiconductor design division.
Alibaba Group Holding Limited, BABA
This deployment represents a significant milestone as the first large-scale implementation of Zhenwu technology throughout the Greater Bay Area region. Alibaba Cloud characterized this initiative as advancing China’s AI computing capabilities “from high-performance technological achievements to widespread commercial deployment.”
The infrastructure employs an advanced high-performance networking framework enabling all 10,000 processors to function collectively as one unified supercomputing system. Alibaba reports this configuration achieves 30% enhanced training and inference performance, while individual card processing capacity has increased roughly tenfold compared to previous-generation technology.
This computing cluster possesses the capability to train artificial intelligence models containing hundreds of billions of parameters — positioning it alongside the most sophisticated AI systems under development worldwide.
Network latency measurements reach 4 microseconds, a performance metric Alibaba credits to the sophisticated networking infrastructure connecting the processor array. This specification proves critical for commercial AI applications where minimal response delays are essential.
Beijing’s Domestic AI Computing Strategy
This launch aligns with a comprehensive national initiative. Chinese authorities incorporated intelligent computing infrastructure into the country’s 15th five-year development blueprint last month, while an August State Council AI strategy document emphasized optimizing computational resource distribution throughout China.
As of June last year, China’s aggregate computing capacity reached 962,000 petaflops — representing 21% of global capacity and marking a 73% annual increase, based on data from the China Academy of Information and Communications Technology.
The Shaoguan facility has already been implemented across healthcare and sophisticated manufacturing applications. Small and medium-sized enterprises can purchase computing resources via China Telecom’s platform using card-based or hourly billing arrangements.
Alibaba has also revealed intentions to expand the cluster from its current 10,000 chips to 100,000 processors. This scaling initiative aims to reduce operational costs while enhancing overall resource utilization.
Industry Background: Huawei’s Role and China’s Semiconductor Competition
This activation follows a comparable achievement last month when China’s inaugural 10,000-card intelligent computing cluster — constructed with Huawei’s Ascend 910C processors — commenced operations in Shenzhen.
That infrastructure provides 11,000 petaflops of computational capacity and has been integrated with a separate 3,000-petaflop cluster launched in 2024. Shanghai is simultaneously developing a 10,000-card facility through an INESA state-owned subsidiary, designed for compatibility with various domestic processor architectures.
Although Chinese semiconductors still lag behind Nvidia in individual chip performance metrics, Beijing’s approach emphasizes large-scale cluster design and optimized networking to narrow the performance differential.
United States export limitations on Nvidia processors have expedited China’s domestic semiconductor development efforts. Alibaba’s T-Head division has emerged as a cornerstone of this initiative, operating alongside Huawei.
BABA experienced a 7.79% price increase on the announcement date, with after-hours trading adding another 0.82% gain on its Hong Kong-listed shares (728-HK).
