Arm said on the 17th (local time) that it plans to apply Nvidia's semiconductor consolidation technology "NVLink" to its "Neoverse" semiconductor design platform for artificial intelligence (AI) data centers. Nvidia unveiled NVLink under the name "Fusion," building an open ecosystem that allows other hardware manufacturers to use the technology.
As AI data centers scale up and as many as millions of semiconductors such as central processing units (CPUs) and graphics processing units (GPUs) are installed, the importance of the technology that consolidates them is growing as much as the performance of a single chip. Vast workloads must be quickly distributed across multiple chips for simultaneous processing and then consolidated again, and only by resolving the bottlenecks that arise there can the efficiency of data center operations be maximized.
With Arm, a powerhouse in the semiconductor intellectual property (IP) market, adopting Nvidia's NVLink technology, the technology is increasingly likely to become a de facto industry standard. From Arm's perspective, using the technology of Nvidia, the No. 1 AI Semiconductor, is expected to lead more data centers to choose its platform, which appears to have informed this decision.
Rene Haas, Arm chief executive officer (CEO), said, "Arm and Nvidia are working to set a new bar for AI infrastructure," and added, "By expanding the Arm Neoverse platform with Nvidia NVLink Fusion, we will deliver 'Grace Blackwell'-class performance to all partners based on Arm."
Jensen Huang, Nvidia CEO, said, "NVLink Fusion is the consolidation fabric of the AI era, consolidating all CPUs and GPUs into a single unified architecture," and added, "We will work with Arm to extend this vision across Neoverse."