ARM is expected to occupy 50% of the data center chip market this year

 

With the explosive growth of AI training and inference needs, the Arm architecture is gradually becoming the new favorite of data centers. Technology giants such as AWS, Google and Microsoft are continuing to expand data center chips based on the Arm architecture. Take AWS as an example. Since cooperating with Arm, more than 50% of its new CPU capacity has adopted the Arm architecture Graviton processor. Almost all large cloud service providers have also launched execution entities using the Arm architecture, achieving at least 40% energy efficiency improvement.

 

In the field of edge and consumer devices, the Arm computing platform has also performed well. More than 99% of smartphones worldwide are powered by Arm architecture, and now AI PCs are being designed to be more like modern smartphones. Since the first wave of AI PCs were released last year, the market has continued to gain momentum. The Arm ecosystem has brought innovation to major platforms, from Apple’s self-developed chips and Windows on Arm devices to AI-oriented, highly efficient ChromeOS and Android. It is expected that by 2025, more than 40% of PCs and tablets will use Arm architecture.

 

To further promote the development of AI applications, Arm continues to invest in the software ecosystem. Since the launch of the KleidiAI software library at COMPUTEX last year, the package has been integrated into multiple mainstream AI frameworks and has been installed more than 8 billion times. This not only optimizes the performance of various devices in multimodal AI workloads such as audio, image, text and video, but also provides strong support for approximately 22 million developers around the world to accelerate the construction of AI applications.

 

Chris Bergey emphasized in his speech that the AI ??era is a critical period for industry cooperation and continuous innovation. Arm will continue to promote global technological transformation with its powerful computing platform and realize a future where AI is ubiquitous. He pointed out that we are at a critical moment in the history of technology. Artificial intelligence is no longer just a concept, but has become a powerful force. AI workloads are becoming increasingly diverse, computing requirements are rising rapidly, and the balance between performance and energy efficiency has become more important than ever.

 

In the data center field, with the rapid evolution of AI models, workloads have unprecedented demand for electricity. In Taiwan, data center electricity consumption is expected to grow 8 times by 2028. If energy efficiency issues cannot be resolved, the expansion of AI will be severely limited. Traditional architectures can no longer adapt to the needs of the AI ??era. Arm's Neoverse CPU is tightly coupled with GPUs and DPUs, breaking through traditional bottlenecks and achieving data transmission speeds 25 times faster than in the past. Major cloud companies are rethinking infrastructure design, and Arm is at the core of this change.

 

Looking to the future, Arm is also constantly advancing technological innovation. Armv9, which was born for AI, was launched several years ago. Last year, the computing technology codenamed Blackhawk was released, achieving double-digit IPC performance growth. Now, the next-generation Armv9 flagship CPU, codenamed Travis, has been unveiled. It will bring powerful performance improvements such as higher IPC and clock speed, and will bring multiples of AI performance through the version of the Scalable Matrix Extension Instruction Set (SME) technology. The next-generation Mali GPU, codenamed Drage, will also bring more sustained performance and richer content experience, and is expected to bring more powerful AI capabilities and console-level graphics experience to mobile devices.