Alibaba unveils Zhenwu V900 and maps a 20-gigawatt AI buildout
Alibaba says its Zhenwu V900 triples its predecessor’s performance, while plans for larger models and expanded data centers outline a broader push for AI computing capacity.

Key takeaways · 3
- 01
Cloud teams should evaluate claimed chip gains against real training and inference workloads before comparing platforms.
- 02
Alibaba’s model roadmap raises the upper target from Qwen3.8-Max’s 2.4 trillion parameters to as many as 10 trillion.
- 03
Infrastructure planners should track whether Alibaba’s proposed 20-gigawatt capacity translates into wider cloud availability.
A faster Zhenwu chip
At Alibaba’s annual flagship conference in Hangzhou, CEO Eddie Wu said the new Zhenwu V900 is “the most powerful AI chip in China today” and can deliver three times the performance of the previous-generation Zhenwu M890. [2] Alibaba’s Zhenwu chips run in its data centers, providing computing to the company and cloud clients for frontier-model training and inference, the calculations a trained model uses to respond to users. [2]
Models and capacity scale
Alibaba said it plans to train a new AI model with five to 10 trillion parameters, compared with 2.4 trillion parameters for Qwen3.8-Max, its most powerful Qwen-series model. [2] Moonshot says Kimi K3, released in July, is the world’s largest open model with 2.8 trillion parameters. [2]
Alibaba plans to expand the data centers supporting its cloud business and expects to have more than 20 gigawatts of computing capacity by 2032. [2] SpaceX had approximately 1.4 gigawatts of AI computing capacity around mid-2026 and says it aims to exceed 10 gigawatts in 2027. [2]
What it means
Alibaba is aligning its chip, model and data-center roadmaps rather than presenting the Zhenwu V900 as an isolated component. Its proposed model would substantially exceed both Qwen3.8-Max and Moonshot’s Kimi K3 in parameter count, although parameter scale alone does not establish model quality. The capacity target would also put Alibaba’s stated buildout above SpaceX’s cited goal, but on a later timetable. What the sources don't address: how the Zhenwu V900 performs on standardized benchmarks or when Alibaba’s larger model will become available.
Alibaba is coordinating proprietary accelerators, larger models and expanded data-center capacity around the same AI strategy. Practitioners evaluating its cloud platform will need performance benchmarks and availability details to determine whether those plans produce practical alternatives for training and inference.
Why it matters
Put this to work — one session a day, built for your industry.
Create a free account for a daily session — eight questions and one real-work challenge, on the news that affects your role.
Start freeHow this developed
24 September 2026
Alibaba unveils Zhenwu V900 and maps a 20-gigawatt AI buildout
24 September 2026
Event created from source cluster.