Alibaba introduced its latest in-house AI processor, the Cenyu M890, at the annual Alibaba Cloud Summit. The announcement forms part of the company’s effort to reduce reliance on Nvidia processors amid US export restrictions.
Technical purpose and performance
The Cenyu M890 was designed by Alibaba’s chip unit, T-Head. According to the manufacturer, the M890 delivers roughly a threefold performance improvement over its predecessor, the Cenyu 810E. The processor is specifically optimized for "AI agents" — software systems capable of executing complex, multi-step tasks with minimal human oversight. Alibaba says the chip is well suited to the high memory and data-transfer demands of these workloads, including handling large context windows and real-time collaboration between models.
Multi-year roadmap
Alibaba laid out a multi-year chip roadmap: the M890 will be followed by a V900 processor in the third quarter of 2027 and a J900 in the third quarter of 2028. T-Head says the V900 is also expected to deliver roughly a threefold performance increase compared with the M890.
New server: Panciu AL128
The company also unveiled a new server system. The Panciu AL128 packs 128 accelerator processors into a single rack. Alibaba said the system is immediately available to Chinese enterprise customers via Alibaba Cloud’s domestic model-serving platform, Pailian.
Shipments and customer base
T-Head reported having shipped more than 560,000 Cenyu units to date. The chips are used by over 400 external customers across twenty industries, from automotive to financial services.
Model update: Qwen 3.7-Max
At the Summit Alibaba presented Qwen 3.7-Max, the latest version of its large language model. The company designed this model for advanced coding and long-running AI-agent tasks; it says Qwen 3.7-Max can operate continuously for up to 35 hours without performance degradation.
Financial and strategic context
These announcements align with Alibaba’s previously announced 380 billion yuan (about 53 billion US dollars) cloud and AI infrastructure investment program, the company’s largest such commitment to date.
Why it matters
The introduction of the M890 and associated hardware is a significant step toward Alibaba’s self-sufficiency in AI infrastructure, particularly given US export controls. The combination of large-volume chip shipments and domestic model-serving capabilities suggests Alibaba is strengthening its ability to provide high-performance AI services within China.
(Information above is based on statements made by Alibaba and T-Head at the Summit.)
