[November 16, 2025] KSIA and K-ASIC Spotlight Korea’s Top AI Chip Innovators at SC25

St. Louis, MO | November 16–20, 2025 – The Korea Semiconductor Industry Association (KSIA) and the Korea AI & System IC Innovation Center (K-ASIC) are proud to present six of Korea’s leading AI semiconductor companies at SC25, the world’s premier supercomputing and AI infrastructure event.

This joint initiative highlights Korea’s growing presence in global AI hardware, from edge intelligence and generative AI acceleration to advanced memory and cooling solutions.

Meet Korea’s AI Chip Innovators at SC25

• ADTechnology

  • Korea’s largest ASIC design house, offering turnkey SoC solutions across advanced nodes (5/4/3nm) for AI and HPC applications.

• FADU

  • Leading SSD controller company, recently introduced PCIe Gen6 ‘Sierra’ controller with 28.5 GB/s speed at sub-9W power—designed for next-gen AI storage.

• HyperAccel

  • Developer of the LPU (Language Processing Unit), purpose-built to accelerate large language models (LLMs) efficiently using Samsung’s 4nm process.

• KoolMicro, Inc.

  • Thermal engineering startup pioneering ultra-dense microchannel liquid cooling (up to 2,000W/cm²) for AI chips and chiplet-based architectures.

• Mobilint, Inc.

  • On-device NPU innovator; its MLA100 MXM delivers 80 TOPS at 25W with 8-core parallelism—ideal for robotics, smart factories, and edge servers.

• Supergate

  • Custom SoC and chiplet interconnect design house providing scalable logic IP solutions for HPC and AI accelerators.

These companies represented the forefront of Korea’s national AI semiconductor strategy delivering efficient, scalable, and competitive technologies for a world increasingly shaped by artificial intelligence.

Independent Korean Exhibitors Expand Korea’s Footprint

In addition to the Korea Pavilion, several Korean companies made independent appearances at SC25, reinforcing the country’s emergence as a full-stack AI hardware player.

Dnotitia made waves with the debut of its Vector Data Processing Unit (VDPU) prototype—a dedicated AI storage accelerator purpose-built for semantic search and RAG (Retrieval-Augmented Generation) workflows. With drop-in compatibility for open-source vector DBs like Milvus and FAISS, the VDPU promises up to 9x throughput gains over CPU-based systems. The company also unveiled updates to its SaaS platform Seahorse Cloud, pushing the boundaries of RAGOps and AgentOps in enterprise LLM applications.

CEO MK Chung emphasized that VDPU is the cornerstone of “AI-native infrastructure,” enabling hyperscale environments to move beyond traditional storage bottlenecks and toward real-time, context-aware data pipelines.

MangoBoost, a rising DPU (Data Processing Unit) innovator, introduced a novel AI server acceleration platform that redefines DPU not as a GPU competitor but as the network and storage brain of modern AI systems. CEO Jangwoo Kim, a former Sun Microsystems CPU architect, emphasized the role of DPUs in eliminating data movement bottlenecks an increasingly critical factor in AI model scaling. MangoBoost’s SC25 showcase featured a hybrid system architecture combining high-capacity HDD arrays, DPU acceleration, and AMD GPUs to offer a cost-effective alternative to traditional SSD-based systems. With growing strategic alignment with AMD and visible support from SK hynix, MangoBoost is positioning itself as Korea’s answer to NVIDIA’s acquisition of Mellanox.

As the AI-driven computing landscape accelerates, SC25 offered a vivid glimpse into the technologies shaping the next frontier—from Korea’s edge NPU breakthroughs to global advances in AI-native storage, networking, and high-performance silicon.

K-ASIC will continue to spotlight Korea’s top semiconductor innovators on the global stage, while building bridges between Korean startups and leading global partners.

Stay tuned as we expand our ecosystem of deep-tech collaboration next stop, CES 2026 and beyond.