SK Telecom (017670.KS) is joining forces with UK semiconductor design firm Arm and Korean artificial intelligence chip startup Rebellions to develop next-generation AI server solutions. The partnership aims to create new data center computing technology that combines central processing units (CPU) and neural processing units (NPU) to reduce dependence on Nvidia graphics processing units (GPU).
SK Telecom announced Thursday that it signed a strategic partnership memorandum of understanding with Arm and Rebellions for next-generation AI infrastructure innovation. Under the agreement, the three companies will jointly develop an AI inference optimization solution that combines Arm's newly introduced "Arm AGI CPU" with Rebellions' "RebelCard," scheduled for release in the third quarter of this year, in a single server configuration. SK Telecom plans to conduct verification tests at its AI data center.
The Arm AGI CPU is a data center processor that marks Arm's first direct manufacturing effort in its 35-year history. The chip is optimized for AI inference services. Rebellions' RebelCard is also an NPU specialized for large-scale AI inference. When combined in a single server, the CPU handles general-purpose computing such as data processing and system operations, while the NPU is dedicated to AI inference calculations. This configuration can improve power efficiency and reduce operating costs compared to GPU-based systems.
The collaboration reflects an industry shift where the focus of AI is moving from model "training" to "inference," the stage of actual service delivery. In AI infrastructure, the key competitive advantage is no longer simply securing massive computing power, but how quickly and efficiently services can be delivered with minimal power consumption.
SK Telecom plans to deploy servers equipped with the solution at its AI data center to verify performance and stability. The company is also reviewing options to run its internally developed AI foundation model "A.X K1" on this infrastructure. "By providing a full package that combines inference-optimized infrastructure with our proprietary foundation model, we will further strengthen our AI data center competitiveness," said Lee Jae-shin, head of AI business development at SK Telecom.
