Meta和Arm提高了AI公司对Arm芯片的效率、切割电源和支持主要数据中心扩展的效率。 Meta and Arm boost AI efficiency on Arm chips, cutting power and supporting major data center expansion.

Meta和Arm联手优化Meta的AI软件,包括PyTorch和FBGEMM,用于Nvidia的Grace CPUs等基于Arm的数据中心系统,提高性能和减少电力使用。 Meta and Arm have teamed up to optimize Meta’s AI software, including PyTorch and FBGEMM, for Arm-based data center systems like Nvidia’s Grace CPUs, boosting performance and reducing power use. 合作提高了AI的推论效率,利用Arm的Enverse 建筑和技术,如KleidiAI,支持Meta从x86转向Arm基础设施。 The collaboration enhances AI inference efficiency using Arm’s Neoverse architecture and technologies like KleidiAI, supporting Meta’s shift from x86 to Arm infrastructure. 这项努力是更广泛的行业向节能人工智能迈进的一部分,包括开源贡献,并支持Meta的15亿美元的德克萨斯数据中心扩张. The effort, part of a broader industry move toward energy-efficient AI, includes open-source contributions and supports Meta’s $1.5 billion Texas data center expansion. 2025年,Arm数据中心的市场份额上升到25%,反映出在AI工作量中越来越多地采用其芯片。 Arm’s datacenter market share rose to 25% in Q2 2025, reflecting growing adoption of its chips in AI workloads.