Alibaba upgrades AI stack with new model, chip

Alibaba updated its AI stack with the Qwen3.7-Max model, new cloud infrastructure, and a new chip for AI agent workloads.

The company said Qwen3.7-Max is built for coding, reasoning, and long task execution.

The model will be offered to developers through its Model Studio platform.

Alibaba Cloud also launched the Panjiu AL128 Supernode Server for training and inference, while Alibaba chip unit T-Head introduced the Zhenwu M890 processor and ICN Switch 1.0.

Alibaba said the T-Head has shipped more than 560,000 Zhenwu units to over 400 external customers.

🔗 Source: Alibaba Cloud

🧠 Food for thought

Implications, context, and why it matters.

US export controls shape Alibaba’s AI plans

The Zhenwu M890 chip has 800 GBps of inter-chip bandwidth, above the 600 GBps interconnect limit in US export controls for some advanced chips sold to China 1.
The rules also cover manufacturing equipment and chip-design software. The Center for Strategic and International Studies (CSIS), a Washington-based policy research group, said the policy aims to “actively strangl[e]” parts of China’s tech industry 1.
Those limits can steer Chinese tech companies toward “technological sovereignty” through more homegrown AI infrastructure 2.
Alibaba rolled out a new chip, an updated Qwen3.7-Max model, and new cloud infrastructure as access in China to some foreign advanced chips gets tighter 3.

Alibaba is tying chips, models, and cloud services together

Alibaba is building a more self-contained AI setup by pairing its own chip, model, and cloud services, which it calls an “AI golden triangle” 2.
That structure can trim costs and tune performance. It also follows the playbook of global hyperscalers, or very large cloud computing providers, that want less reliance on one hardware supplier.
The result could give enterprises in China and other markets another way to spread AI supply-chain risk beyond US-centered technology 2.