Arm, a renowned global frontrunner in semiconductor intellectual property (IP), recently took the wraps off its inaugural self-designed CPU for data centers—the Arm AGI CPU—at an event held in San Francisco, USA. This milestone signifies Arm's maiden venture into the direct sale of self-developed chips after 35 years in the business, marking its official entry into the fiercely competitive core computing power market for data centers.

Constructed utilizing TSMC's cutting-edge 3nm process technology and boasting a dual-Chiplet architecture, this innovative chip is meticulously crafted for AI agent infrastructure, promising unparalleled performance. Reports indicate that its single-rack performance can surpass that of x86 platforms by over twofold, potentially slashing computing power capital expenditures by up to $10 billion for every 1GW AI data center. Moreover, the Arm AGI CPU offers a diverse range of deployment options, adeptly catering to the varying needs of data centers across different scales.

During the launch event, NVIDIA's CEO, Jensen Huang, extended his heartfelt congratulations. Arm's CEO further disclosed that the cumulative shipments of Arm-architecture chips have soared past the staggering figure of 350 billion, a number that not only dwarfs the global population but also stands at seven times the cumulative shipments of non-Arm architecture CPUs, underscoring Arm's dominant position in the semiconductor landscape.