Arm은 2025년 상위 하이퍼스케일러에 출하되는 컴퓨팅의 약 50%가 Arm 기반일 것으로 전망했다. 이는 Arm 네오버스 출시 6년 만에 데이터센터 생태계의 확장성과 비용 효율성에 큰 변화를 가져온 결과라고 1일 설명했다.
AI Servers to Grow by Over 300% in the Next Few Years
Arm-based chips up to 60% more efficient than before
Arm forecasts that approximately 50% of compute shipped to the top hyperscalers by 2025 will be Arm-based. This is the result of a significant change in the scalability and cost efficiency of the data center ecosystem in just six years since the launch of Arm Neoverse, the company explained on the 1st.
Mohamed Awad, Senior Vice President, said that the cloud computing environment is being reorganized in the AI era, highlighting the rapid increase in computing demand for complex learning and inference workloads and the increase in data center demand. He added that AI servers are expected to grow by more than 300% in the next few years, and power efficiency has become a basic requirement of the industry.
Through Neoverse, Arm is helping market-leading partners shape their silicon technology roadmaps based on insights gained from running software at scale. As a result, 10 of the world’s largest hyperscalers, including Amazon Web Services, Google Cloud, and Microsoft Azure, are developing and deploying Arm-based chips in their data centers, with some reporting up to 60 percent efficiency improvements over previous-generation chips.
In particular, the Arm compute platform provides partners with the flexibility to develop custom silicon solutions for AI. A representative example is NVIDIA's Grace Blackwell superchip for AI-based infrastructure, which integrates Arm Neoverse-based Grace CPUs and NVIDIA Blackwell GPUs to deliver performance optimized for AI workloads.
General-purpose chipsets from Arm partners, such as AWS’s Graviton, Google Cloud’s Axion, and Microsoft’s Cobalt, enable software developers to build applications on Arm and benefit from efficiency and performance optimizations.
Major enterprises such as Paramount+, Spotify, and Uber are increasingly migrating to Arm-based cloud infrastructure due to the TCO and energy savings, he explained. Leading data platform companies such as Oracle and Salesforce are also said to be converting their services to Arm-based infrastructure.
“With power efficiency as the top priority, clouds and data centers are being designed around Arm from silicon to software,” said Awad. “Arm is becoming ubiquitous from cloud to edge, opening up new possibilities for innovation, efficiency, and performance.”