(Source: Arm Technology)
Currently, large model technologies represented by DeepSeek are leading a new wave of AI industrialization. This trend not only accelerates the deep penetration of AI technology in various vertical fields and drives the large-scale application of high-performance AI chips but also reconstructs the technological development path for the next generation of intelligent devices and AI infrastructure. Among them, as the core computing unit supporting AI training and inference tasks, CPUs are continuously optimizing their performance and energy efficiency balance to meet the exponentially growing demand for AI computing.The relentless pursuit of performance and energy efficiency has been the core gene of the Arm® architecture for over 30 years. Thanks to its comprehensive advantages in performance, power consumption, and cost, Arm technology has deeply empowered fields such as cloud computing, data centers, and 5G network devices, driving innovative evolution in traditional infrastructure chip domains. Additionally, with a solid foundation in mobile terminals, smart IoT, and automotive markets, the Arm architecture is accelerating coverage of various scenarios from cloud to edge, solidifying the computing foundation for ubiquitous intelligent devices and infrastructure, leading the future of AI computing.
Arm architecture will account for half of the computing power shipped to major cloud service providers by 2025Relying on the Arm Neoverse™ technology platform, the Arm architecture can fully meet the current infrastructure construction’s demand for high energy efficiency and chip customization, thereby achieving system-level optimization of data center infrastructure and effectively accelerating the technological innovation and product commercialization process of cloud computing partners. Currently, the Arm architecture is in a rapid growth phase in the global server market, with nearly 50% of the computing power shipped to major hyperscale cloud service providers by 2025 being based on Arm architecture.Among them, the world’s top ten hyperscale cloud vendors, including Amazon, Google, and Microsoft, are actively developing server chips based on Arm architecture. In some use cases, chips using Arm technology solutions have achieved energy efficiency improvements of up to 60% compared to previous generations. Arm technology solutions also support flexible customization and exclusive optimization, such as NVIDIA’s Grace Blackwell super chip, which integrates the NVIDIA Blackwell GPU and the Arm architecture-based Grace CPU, providing excellent computing power support for AI workloads.In the domestic market, several cloud computing giants and startups are also actively promoting infrastructure construction based on Arm architecture. For example, Alibaba Cloud’s “Yitian 710” chip has been widely used in databases, big data, and video fields since its launch, with a deployment scale reaching millions of cores. In AI use cases, thanks to the multiple enhanced features integrated into Arm technology, the “Yitian 710” demonstrates strong flexibility and scalability when running large models like Llama 3 and Qwen1.5, effectively reducing related development costs and process pressures.Driving the most widespread intelligent devices and facilities to create a “core” upgrade, Arm technology is leading the future of AI computingAs the only computing architecture supporting a trillion-level ecosystem globally, Arm technology initially emerged in the mobile computing field, such as smartphones, and is now accelerating its expansion into diversified fields like PCs, IoT, smart cars, and data centers. In this process, the Arm architecture has developed into the most widely used computing architecture globally, continuously driving the “core” upgrade of the most widespread intelligent devices and infrastructure. To date, the shipment volume of chips based on the Arm architecture by partners has exceeded 300 billion units.Entering the era of artificial intelligence, Arm technology, based on a complete product matrix from CPU to computing subsystems, a mature and complete software ecosystem, and the ecological momentum covering the entire intelligent computing industry chain, is leading the future of AI computing. Currently, there are over 20 million software developers building applications on devices based on Arm architecture. It is expected that by the end of 2025, the number of Arm devices supporting AI computing will exceed 10 billion.Looking to the future, Arm Technology will continue to strengthen its role as a bridge connecting the global Arm ecosystem with the domestic intelligent computing industry, continuously introducing the latest Arm product technologies, while enhancing its diversified self-research innovation layout. Through heterogeneous integration, it aims to create system-level solutions that meet both international standards and local market demands, collaborating widely with partners in emerging fields such as edge AI to jointly seize the opportunities of AI intelligent computing.