According to the company, this chip triples the performance of its predecessor and is designed to handle the demanding workloads of frontier-model training.
The announcement marks a significant step in the e-commerce giant’s transition into a major provider of AI infrastructure.
The release of the Zhenwu V900 is intended to provide a domestic alternative to hardware from Nvidia Corp.
as Alibaba expands its global data center capacity.
By developing its own high-performance chips, the company aims to support a massive buildout of computing power, targeting 20 gigawatts of data center capacity by 2032.
This internal hardware strategy allows Alibaba to scale its cloud computing resources while navigating the competitive landscape of the global AI race.
Alibaba CEO Eddie Wu stated that the new chips can be integrated into massive clusters of up to 500,000 units to support large-scale AI development.
This high-density scaling mechanism is essential for training the most advanced generative models, which require vast amounts of interconnected processing power.
The deployment of these units will primarily affect the company’s infrastructure capabilities as it seeks to underpin future technological growth.