Alibaba stole the show at today’s Apsara Conference in Hangzhou, China, detailing ambitious compute deployment plans, offering tantalizing details about its upcoming Zhenwu V900 chip, and laying down the gauntlet by disclosing between 4 trillion and 10 trillion parameters for its upcoming Qwen 4.5 and Qwen 5.0 AI models. Alibaba’s upcoming Zhenwu V900 chip will have 216GB of on-package memory vs. the paltry 141GB that NVIDIA H200 GPU offers Alibaba is now claiming that the upcoming Zhenwu V900 chip will have 216GB of on-package memory, and launch in Q1 2027. We can only theorize that the V900 will leverage CXMT’s […]