All in AI sprints in an all-round way to solidify the foundation of AI infrastructure in the era of intelligent agents. The three major business lines of ARM Technology collectively deliver results

Intelligent Things Author | Yunpeng Editor | Mo Ying The popularity of "Lobster" and "Hermès" has kicked off the era of Agentic AI. From the beginning of the year to the present, AI agents have almost become the only main thread running through the fields of intelligent terminals, applications, models, computing power, and embodiment. How to build the "AI computing power infrastructure" in the era of intelligent agents has become the focus of attention of various technology giants and startups. As the computing power paradigm bids farewell to the stage of pursuing the ultimate performance of a single chip, factors such as overall system efficiency, energy efficiency, and cost have become more critical considerations for enterprises. Facing the era of intelligent agents, finding a computing power platform that is more suitable for one's own business needs and more "more efficient, more economical" has become a "rigorous need" for almost every player in the AI ​​​​track. 1. The transformation of AI computing power is a foregone conclusion, and new solutions are urgently needed to break the chip problem. Today, AI is accelerating from the cloud to the terminal. In the field of the Internet of Things represented by AIoT, devices are moving from "perceptual connection" to "intelligent computing". A large number of embedded devices face multiple resource constraints such as performance, power consumption, latency, and storage, and urgently need breakthroughs at the chip layer. At the same time, small and medium-sized models are rapidly improving to accelerate the implementation of intelligent agents. As the core engine of AI inference, NPU faces new challenges in dimensions such as accuracy support, memory bandwidth, computing efficiency, and ecological compatibility. In the embodied field, from the evolution of the robot's "brain", "eyes" and "hands" to the robot's truly delicate perception of physical reality and interaction with people, solid support from the chip layer is required. The skyrocketing demand for cloud inference has made giants increasingly demanding solutions with more advantages in energy efficiency, power consumption, and cost. The era of "big chips" has become a thing of the past, and the flourishing computing power market has brought about massive opportunities. From the edge to the device to the cloud, AI computing power is about to change, bringing a series of new opportunities and challenges to the chip industry. The creation of these excellent chips is inseparable from the support of underlying key IP technologies and solutions. Under such industry trends, Arm Technology, as an enabler of underlying core IP and solutions, did not hesitate to fully embrace AI. Its goals in the AI ​​era are extremely clear: since last year, it has set the "AI Arm"