Alibaba unveils its Zhenwu V900 AI accelerator chip and ties it to an expansion plan for data centre capacity. The company says the new accelerator triples the performance of its prior generation, positioning it for demanding “frontier” model training.
Alibaba also sets a target to reach 20 gigawatts (GW) of data centre capacity by 2032. In its presentation, the chip is described as scalable for large AI training workloads, including the ability to be combined in very large clusters, with some reporting specifying cluster sizes of up to hundreds of thousands of units. Across outlets, the focus is on how the V900 improves compute performance while supporting an aggressive infrastructure build-out.
While both sources align on the chip’s performance claim and the 20 GW by 2032 goal, their coverage differs in emphasis: one article foregrounds the chip’s technical specs and scaling approach, while the other frames the announcement more broadly as an intensified push into AI hardware and data centre development.
