CFi.cn: Alibaba ships Zhenwu V900 chip and Pan Jiu nodes; Qwen3.8 users up 8.5x in two months
A CFi.cn roundup on September 27 of broker research noted that Alibaba unveiled the Zhenwu V900 AI chip (single cluster supporting 500K cards) and Qwen3.8-Flash at the September 24 Apsara Conference; Qwen3.8 users grew 8.5x and token usage 12x in two months.
A CFi.cn roundup on September 27 of broker research notes that Alibaba unveiled the Zhenwu V900 AI chip at its 2026 Apsara Conference on September 24. Zhenwu V900 supports 500,000 cards in a single cluster, triples compute over the previous generation, and is expected to be ready in Q1 2027. The conference also saw the launch of Pan Jiu AL64 and AL144 hyper-node products, completing Alibaba’s chip + server + cloud full stack. On the model side, Alibaba said Qwen3.8 users grew 8.5x and token usage 12x in two months; the new Qwen3.8-Flash maintains low pricing while approaching frontier performance.
The broker research argues Zhenwu V900 combined with Alibaba Cloud’s AgentCore closes a "compute + toolchain" loop, strengthening Alibaba Cloud in the domestic AI infrastructure market and benefiting domestic AI chip and server vendors. The same note emphasizes that the rapid Qwen3.8 user and token growth reflects a sharp pickup in inference demand, with Alibaba’s combination of capability and pricing closer to real developer usage than competitors.
On the application layer, Alibaba upgraded Alibaba Cloud to "Agentic Cloud" and released runtime platforms such as AgentCore, pushing agents from "can generate" to "can deliver." In the same window DeepSeek disclosed its DSec sandbox platform can support 3 million agent training sandboxes per day and Meta Muse passed 2.5 million downloads within 13 days of launch, together marking a shift to scaled agent training and commercial deployment.