Table of Contents
Alibaba’s semiconductor subsidiary T-Head introduced the Zhenwu M890 AI processor, which the company said delivers three times the performance of its previous-generation chip
In sum – what to know:
New AI model – Alibaba launched the Qwen3.7-Max model for agentic AI workloads, including coding, reasoning and long-duration autonomous task execution involving extensive tool usage.
Infrastructure expansion – Alibaba Cloud introduced the Panjiu AL128 Supernode Server, integrating 128 AI accelerators in a single rack to support large-scale AI training and inference.
Chip upgrade – T-Head unveiled the Zhenwu M890 processor and ICN Switch 1.0 networking chip to improve performance and bandwidth for AI compute clusters.
Chinese company Alibaba Cloud announced a broad update to its AI portfolio during its Alibaba Cloud Summit, introducing a new large language model, upgraded cloud infrastructure, and next-generation AI processors aimed at supporting enterprise AI agent deployments.
The company unveiled Qwen3.7-Max, a new foundation model designed for coding, reasoning, and long-duration autonomous AI tasks. According to Alibaba, the model is optimized for software engineering workflows, office automation, and multi-step agent operations involving extensive tool usage. The company said the model can support extended autonomous execution and will be made available through its Model Studio platform.
Alibaba also introduced the Panjiu AL128 Supernode Server, a rack-scale AI system designed for large-scale AI training and inference workloads. The platform integrates 128 AI accelerators in a single rack and uses the company’s ICN Switch 1.0 networking technology to increase bandwidth between processors. Alibaba said the system is intended to support higher concurrency for AI agent workloads and reduce bottlenecks in large model deployments.
The infrastructure announcement reflects increasing industry focus on AI systems capable of coordinating multiple models, tools, and workflows over longer operational periods. Large-scale inference capacity and interconnect bandwidth are becoming key requirements as enterprises deploy more autonomous AI systems.
During Alibaba Group’s latest earnings call, Eddie Wu said the company is seeing rapid growth tied to AI training, inference, and agent orchestration workloads. “We are at a pivotal inflection point in the evolution from conversational chatbots to autonomous AI agents,” Wu said, adding that the shift is driving “explosive growth” across AI infrastructure demand.
Wu said AI-related product revenue now accounts for 30% of Alibaba Cloud’s external revenue and is expected to surpass 50% within about a year. He also said the company expects continued acceleration in cloud growth as demand shifts from “traditional compute and storage to models, AI compute and agent services.”
Alibaba’s semiconductor subsidiary T-Head also introduced the Zhenwu M890 AI processor, which the company said delivers three times the performance of its previous-generation chip. The processor includes 144 GB of memory, supports precision formats ranging from FP32 to FP4, and is designed for both AI training and inference workloads. The launch of the new chip shows how the Chinese technology giant intensifies efforts to build domestic alternatives to Nvidia chips amid tightening U.S. export restrictions.
Alongside the processor, T-Head launched the ICN Switch 1.0 networking chip and its proprietary T-Head SAIL software stack for managing AI compute infrastructure. Alibaba said more than 560,000 Zhenwu chips have been shipped to date, with deployments spanning sectors including automotive and financial services.
Alibaba last year pledged to spend more than CNY380 billion ($53 billion) on cloud and AI infrastructure over three years, according to previous reports.
In October 2025, Alibaba Cloud opened its second data center in Dubai, United Arab Emirates (UAE) as part of the firm’s new international expansion plans.
