Overview
- At its Apsara Conference on Sept. 22, Alibaba introduced the Zhenwu V900, saying it delivers roughly three times the performance of the M890 and offers 216 GB on‑package memory, 1,200 GB/s inter‑chip bandwidth, and planned mass production in Q1 2027.
- Alibaba confirmed Qwen 4 is in training and outlined plans for future Qwen 4.5 and Qwen 5 models that the company projects will span roughly 4 to 10 trillion parameters, with a separate target for a 5–10 trillion‑parameter model.
- The group set a long‑range infrastructure goal of more than 20 gigawatts of global data‑center capacity by 2032 and said it will open new cloud regions in countries including Türkiye, Finland and the Netherlands while expanding sites in Europe, the Middle East and Asia.
- Investors showed a positive near‑term reaction with Alibaba shares rising about 3–4% after the announcements, reflecting confidence in the company’s $53 billion AI spending pledge and recent capital raises to fund the buildout.
- Key claims remain vendor‑reported and unverified, and deployment faces hurdles such as production yields, memory and chip supply, software and tooling compatibility with existing ecosystems, and the need for independent benchmarks.