Speaking at the main forum of the 2026 Apsara Conference on Sept 22, Alibaba Group CEO Wu Yongming described AI models, AI chips and AI cloud as the three pillars of the machine intelligence era, saying the company will keep investing firmly in all three as a long-term strategic priority.
On the chip side, T-Head has assembled a full lineup of data center chips, spanning its Zhenwu GPUs, Yitian CPUs, Panmai smart NICs and ICN interconnect chips — giving it complete coverage of the core chips needed to build hyperscale AI clusters.
At this year's conference, T-Head took the wraps off the Zhenwu V900, its latest AI chip for both training and inference. The chip delivers three times the single-chip performance of the Zhenwu M890, packs 216GB of memory and 1,200GB/s of inter-chip interconnect bandwidth, and natively supports FP8 and FP4 low-precision computing — enough to handle training and inference for trillion-parameter models. The V900 is due to enter mass production in the first quarter of 2027 and will be rolled out at scale across Alibaba Cloud data centers.
Using T-Head's in-house ICN Switch interconnect chip, the V900 can link thousands of cards at full bandwidth. Its new Panjiu super node server brings together the V900, ICN Switch, Panmai smart NIC and Zhenyue SSD controller, and — paired with Alibaba Cloud's next-generation network architecture — allows a single AI cluster to scale to 500,000 cards.
Super nodes built on the M890 are already in large-scale use and have trained models exceeding 2 trillion parameters, including Qwen3.8 and Kimi K3, according to the company. As of June 2026, the Zhenwu series had served more than 650 enterprise customers.