Samsung Teams Up With Arm to Develop 2nm On-Device AI Custom Chips, OpenAI Reportedly a Key Customer

TradingKey - On September 4, according to South Korean media outlet Greened.kr citing semiconductor industry sources, Arm Holdings (ARM) has approved non-recurring engineering (NRE) funding for a next-generation on-device AI system-on-chip (SoC) and officially launched joint R&D with Samsung Electronics. The project adopts a collaborative model in which Arm provides core technology while Samsung handles design and manufacturing, with plans for mass production using an advanced 2nm process.

The report cited industry insiders as revealing that OpenAI is a candidate end customer for this project. As AI capabilities accelerate their expansion from the cloud to edge devices, if OpenAI further expands into on-device AI, its demand for custom AI chips is expected to rise accordingly.

Arm Approves NRE, On-Device AI SoC Project Officially Launches

According to reports, Arm approved non-recurring engineering (NRE) fees for the development of a next-generation on-device AI SoC at the end of August, launching a joint project with Samsung. The custom chip is designed specifically for on-device AI, allowing computations to be performed locally on terminal devices such as smartphones without routing through the cloud, thereby reducing pressure on cloud computing power while improving response speeds and data privacy.

The collaboration operates under a Limited Use License (LUL) framework, under which the technology provided by Arm is restricted to product development for specific clients. From a business model perspective, Arm is not a chip vendor, but a provider of core technology and a commissioned development partner; Samsung is fully responsible for design and manufacturing, delivering the completed chips directly to end customers and handling settlement and collection.

Clear Division of Labor: Samsung Mass-Produces 2nm, Arm Supplies Core IP

The division of labor between the two parties is clear: Samsung Electronics' System LSI Business will be responsible for SoC design based on Arm's AI architecture, while its Foundry Business will leverage advanced processes to achieve mass production using a 2nm node; Arm will provide key design technologies such as its proprietary AI accelerator architecture and RTL, while conveying end-customer requirements to Samsung and coordinating the overall project development progress.

What It Means for Samsung: Design and Manufacturing Synergy as 2nm Mass Production Serves as a Model

For Samsung, the project is expected to form a complete "design + manufacturing" AI chip solution—synergizing its System LSI and foundry businesses within the same project and allowing its advanced node capabilities to be directly embedded into the AI chip development process.

Particularly on the foundry side, practical development and mass production experience with 2nm on-device AI chips carries strong demonstrative significance. If the project progresses smoothly, Samsung can leverage it to build mass-production track records for serving AI chip clients with advanced nodes, providing a benchmark for securing future orders for AI accelerators and custom SoCs. As demand for AI chips continues to spread to edge devices, whether Samsung can open up the custom AI chip market through Arm and potential OpenAI projects will be a new focal point for the synergistic growth of its System LSI and foundry businesses.