Alibaba has set a target for the global data center capacity operated by Alibaba Cloud to exceed 20GW by 2032 as the Chinese technology group expands its full-stack artificial intelligence strategy across models, chips, infrastructure and agents.
The target was announced at the 2026 Apsara Conference in Hangzhou, where Alibaba detailed a roadmap spanning its Qwen model family, proprietary T-Head processors, AI infrastructure and a new agentic cloud architecture.
Alibaba Group Chief Executive Officer Eddie Wu said the company expects demand for machine intelligence to continue expanding and is planning infrastructure around that longer-term growth. The 20GW figure is a future capacity target rather than currently deployed capacity.
Qwen 4 enters training
Alibaba said its next-generation Qwen 4 model is now in training. It also outlined future Qwen 4.5 and Qwen 5 series models that the company says could scale to between 5 trillion and 10 trillion parameters.
The company also highlighted experiments in recursive self-improvement. Alibaba said Qwen3.8-Max completed 33 automated improvement cycles over one month and increased its score on Artificial Analysis from 40 to 45. In a separate chip-design experiment, Alibaba said the model made more than 10,000 electronic design automation tool calls over more than 60 hours and reduced chip area by 42 percent without a performance reduction. These figures were reported by Alibaba and are not independent benchmark results.
Other model updates include Qwen3.8-LiveTranslate for simultaneous interpretation, additions to the Qwen Audio family and Qwen-Image 3.1, which Alibaba plans to launch later this year. Alibaba also introduced Qwen Intelligence, a business-facing agent platform intended to give smartphone manufacturers access to Qwen-powered cross-app capabilities.
Zhenwu V900 scheduled for 2027
T-Head, Alibaba’s chip design unit, unveiled the Zhenwu V900 accelerator for AI training and inference. Alibaba said the processor has 216GB of memory and 1,200GB per second of inter-chip bandwidth and delivers three times the performance of the earlier Zhenwu M890. The company plans mass production and commercial availability in the first quarter of 2027.
Alibaba said Zhenwu processors are already used by more than 650 customers across sectors including automotive, finance, large language models, embodied intelligence, energy and manufacturing.
The group also outlined new Yitian processors for agentic workloads. The Yitian 720 and Yitian 730 are scheduled for 2027, with the latter based on T-Head’s own microarchitecture.
Agentic cloud architecture
Alibaba Cloud described its agentic cloud strategy in three layers: AI Native Cloud for model training and inference, Agent Native Cloud for enterprise deployment and security, and a Context Engine for real-time data and long-term memory.
The new roadmap extends Alibaba’s previous investment in global AI infrastructure and partner ecosystems. TNGlobal has previously covered Alibaba Cloud’s plans to expand its international AI partner network and related investments.
Alibaba Cloud accelerates global AI partner ecosystem with new incentives, investments