Park Sung-hyun

Rebellions designs AI inference accelerator chips for data center deployment, targeting the workload characteristics of large language model inference and transformer model serving with a custom silicon architecture that aims to deliver better performance per watt than general-purpose GPU solutions for inference-specific workloads. The Seoul company has announced a merger with Sapeon, the SK Telecom AI chip subsidiary, combining two of South Korea most prominent domestic AI chip efforts.\n\nThe company raised approximately $124 million including a Series C round from investors including KT, POSCO, and Korean government-affiliated investment funds. Rebellions ATOM chip has been benchmarked for LLM inference performance and deployed in early customer trials at Korean cloud and telecommunications operators. The proposed merger with Sapeon, if completed, would create a combined entity with broader resources and shared technology for competing against NVIDIA in the Korean AI infrastructure market.\n\nRebellions competes in the AI accelerator market against NVIDIA, AMD, and domestic chip efforts from Samsung and SK Hynix, as well as global AI chip startups including Groq, Cerebras, and Tenstorrent. South Korea semiconductor industry expertise, centred around Samsung and SK Hynix memory chip production, provides a technical ecosystem that supports AI chip design through access to advanced packaging, high-bandwidth memory, and process node expertise. The Korean government has identified domestic AI chip capability as a strategic priority, providing funding support that supplements commercial venture investment in companies like Rebellions.