Arm and Alibaba released chips for AI agents operation - ForkLog: cryptocurrencies, AI, singularity, future

# Arm and Alibaba Release Chips for AI Agents

Arm Holdings has introduced its own data center chip optimized for AI inference.

The first customer to receive the new CPUs is Meta. According to Arm CEO Rene Haas, by 2031, this development is expected to generate around $15 billion in annual revenue. The company’s total revenue could reach $25 billion, with earnings per share of $9.

CPUs are once again in high demand due to the development of agent-based AI, which is changing computing requirements. According to Arm’s head, demand for such chips will grow fourfold.

“In modern data centers, CPUs manage thousands of tasks: coordinating accelerators, managing memory and storage, distributing loads, and moving data. With the advent of agent AI, they also coordinate the work of multiple assistants,” the company said.

The release of Arm AGI CPU signifies a major change in the company’s business model. Previously, it had no manufacturing facilities and earned revenue solely from licensing and royalties for 35 years.

“Arm AGI CPU provides high performance per task when operating thousands of cores in parallel, while remaining within power consumption and cooling limits,” the company’s blog emphasizes.

Arm is known for its architecture, which underpins most modern smartphones. In 2018, it began competing with server chips from Intel and AMD by launching the Neoverse platform.

“For the first time in over 35 years, Arm is releasing its own chips — expanding Neoverse beyond IP and Arm Compute Subsystems to give customers more options for deploying computing: from creating custom chips to integrating platform solutions or using processors developed by the company itself,” the announcement states.

Alibaba’s Equivalent

Chinese tech giant Alibaba has introduced its own solution — the XuanTie C950 CPU for agent capabilities.

The chip can handle multi-step tasks performed by AI assistants. It will be used in data centers for inference.

Until recently, the industry mainly focused on graphics accelerators (GPUs), dominated by Nvidia. These chips are needed for training large-scale models because they can perform many calculations in parallel.

In contrast, CPU architecture involves sequential operations. This makes them an optimal choice for the developing ecosystem of AI agents focused on implementing specific chains of actions.

XuanTie C950 “can be customized for specific inference scenarios, allowing clients to adapt processors to their tasks.” The processor is based on the open RISC-V standard, an alternative to Arm’s architecture.

Recall that in December 2025, Amazon Web Services introduced a new version of its own AI chip — Trainium3.