Samsung and Arm Partner to Develop Custom 2nm On-Device AI Chips
Samsung Electronics and Arm have officially launched a joint effort to design and manufacture custom next-generation on-device AI SoCs. Arm approved Non-Recurring Engineering (NRE) funding for the initiative in late August, setting the collaboration into motion. This alliance accelerates the industry-wide shift toward processing AI workloads directly on smartphones and edge devices without relying on cloud computing. It also serves as a major vote of confidence for Samsung's 2nm semiconductor node as it competes with TSMC for advanced chip manufacturing contracts. Arm will supply its AI accelerator architecture, Register-Transfer Level (RTL) technology, and project management while conveying client needs to Samsung. Samsung System LSI will integrate the SoC design, and Samsung Foundry will handle mass production on its 2nm process node.
## BACKGROUND
On-device AI executes artificial intelligence tasks locally on end-user hardware, improving latency, privacy, and power efficiency compared to traditional cloud processing. A System-on-Chip (SoC) combines key compute blocks—such as CPUs, GPUs, and Neural Processing Units (NPUs)—onto a single silicon die to run these machine learning models efficiently.