Alibaba, which is known for being one of the largest retailers and e-commerce companies, said it is preparing for the significant shift towards agentic AI.

Its new server chip, XuanTie C950-5, has been built using open-source RISC-V chip architecture, Alibaba said, which has previously been referred to as world-leading RISC-VCPU technology.

It is also expected that the chip will perform three times faster than its predecessor, the XuanTie C920, according to reports.

“RISC-V’s open-standard nature allows chip designers ⁠to customise instruction sets and accelerate specific AI workloads with no or low licensing fees. This is particularly important for the development of AI agents,” Alibaba said via its blog post.

Having committed to accelerating in-house chip development, Alibaba is mainly focusing on its Zhenwu810E chip series for AI training and inference. The XuanTie series, on the other hand, is more focused on high-performance cloud systems and agentic AI.

The news comes shortly after Alibaba launched Wukong, its enterprise agentic AI platform that can coordinate multiple AI agents to perform complex business tasks. Wukong is part of Alibaba’s broader strategy to advance further in the global agentic AI race – particularly as interest in China peaks on account of OpenClaw.

Previously known as Clawdbot and Moltbot, OpenClaw is an open-source self-hosted AI agent platform that is able to act autonomously and connect to large language models (LLMs) to handle workflows. Interest has soared across China and Alibaba has even launched its own agentic AI platform, Accio Work, to run complex business operations for small and medium-sized enterprises (SMEs).

The company is eager to ensure it earns profit as rising competition across China has led to token prices dropping.

“We distinguish ourselves ⁠by being a specialised B2B tool rather than a generalist platform,” Alibaba International vice president Kuo Zhang said, reported by Reuters. “We draw a very clear ⁠line at high-stakes operations … any action involving financial transactions, payment execution, or access to private files requires explicit, granular permission from the user.”

The company has also said it plans to separate its AI business from its cloud computing arm, as it appears to be shifting focus to digital assistants.

Zhang said the high-stakes global push to define agentic ‌AI carries inherent risks that can only ​be mitigated with controlled, ​specialised models that balance automation with security.

“We believe the greatest risk lies in using horizontal, generalist models for vertical business tasks,” Zhang added. “By focusing on specialised B2B ​agents and implementing AI ‌alongside human approval layers, we can deliver the benefits of an autonomous workforce ​without the traditional risks associated with unconstrained AI.”

Related stories

AI agents are starting to talk: Are the networks ready?

Nvidia CEO predicts $1tn compute demand as AI reaches ‘platform shift’

Oracle targets ‘autonomous’ enterprise operations with new agentic applications

Google Cloud’s journey towards the ‘agentic telco’ to reshape network AI