Taipei—Korea IT Times Reporting Team
The COMPUTEX 2026 show floor, which opened on June 2, was once again dominated by artificial intelligence. But the shift felt on the ground was not simply about more powerful AI chips or larger data centers. The center of gravity is moving toward Physical AI, where artificial intelligence senses, decides and acts in the real world.
Korean companies had a stronger presence in this transition. A total of 39 Korean firms participated in this year’s COMPUTEX. While their number was smaller than that of Taiwanese exhibitors, their technologies in AI semiconductors, robotics, displays, memory and edge computing left a clear impression. From Samsung Electronics, SK hynix and LG Electronics to ultra-low-power AI chip startups, Korean companies showed that they are becoming part of the core architecture of the next AI industry.
Korean giants underpin AI infrastructure
Across the exhibition halls, Korea’s influence was visible in the technologies that support AI infrastructure, including memory, displays and system solutions. Samsung Electronics and Samsung Display focused on AI memory, including HBM, next-generation displays and system-level solutions, as they sought to expand cooperation with global customers. Their presence pointed to a broader strategy: moving beyond supply-chain strength toward application markets such as AI data centers, robotics and AI PCs.
Samsung’s core competitiveness lies in memory, system semiconductors and the on-device AI ecosystem that supports AI computation. HBM has become a critical component for large-scale AI training and inference because it reduces data bottlenecks and enables high-speed processing alongside GPUs and AI accelerators. Samsung Display highlighted high-resolution, low-power and high-brightness panel technologies aimed at AI PCs, robotics and automotive displays. In the Physical AI era, displays are becoming essential interfaces through which robots, vehicles and industrial equipment interact with humans.
SK hynix focused on HBM, now one of the most important components in the AI data-center market. As the performance of AI training and inference increasingly depends not only on compute chips but also on memory bandwidth and power efficiency, SK hynix appears to further strengthen its position as a key partner in NVIDIA’s supply chain. HBM vertically stacks multiple DRAM chips and places them close to GPUs, enabling large volumes of data to move quickly for large AI model operations.
LG Electronics and LG Display presented technologies centered on robotics, AI PCs and industrial display solutions. LG Electronics showed a strategy that connects robots, smart spaces and AI-powered home appliances, while LG Display emphasized the role of high-performance displays in shaping the user experience of robots, mobility systems and AI PCs. As Physical AI expands into factories, hospitals, logistics centers and homes, the role of screens that allow people to monitor outcomes and communicate commands will become increasingly important.
Industry sources at the venue said major Korean companies also took part in private networking events such as Korea Partner Night. These meetings reportedly covered supply-chain cooperation with leading AI companies as well as longer-term strategic partnerships. The role of Korean conglomerates is shifting from component suppliers to strategic partners that help design AI infrastructure and application ecosystems.
Korean AI chip startups stand out at InnoVEX
Among Korean exhibitors, AI semiconductor startups drew particular attention. At InnoVEX, the startup-focused exhibition area, Korean firms showcased next-generation AI computing technologies to investors and global companies. Their shared message was clear: low power, compact design and real-world deployment.
DEEPX was one of the most closely watched companies. Located at booth M1311a in TaiNEX 1, DEEPX focused on ultra-low-power NPU technology for the Edge AI market. An NPU, or Neural Processing Unit, is a semiconductor optimized for neural-network operations such as image recognition, object detection, voice processing and anomaly detection with lower power consumption.
DEEPX’s DX-M1 delivers 25 TOPS of performance while keeping power consumption in the 1–5 W range. This allows some AI inference tasks that previously required high-power GPUs to run directly inside cameras, industrial equipment, robots and embedded systems. Its DX-H1 Quattro delivers up to 100 TOPS and targets more complex Physical AI environments, including multi-camera input, real-time video analytics, robot control and smart-factory equipment.
The company emphasized more than 20 times higher power efficiency compared with GPUs and promoted its slogan, “AI for Everyone & Everywhere.” At the booth, DEEPX presented mass-production plans based on cooperation with industrial computer company AAEON, along with several industrial demonstrations. Visitors paid close attention to edge AI systems that process AI workloads directly inside field devices rather than relying on the cloud.
Mobilint also drew attention with AI inference chips designed for image recognition and edge AI processing. Its solutions target CCTV, drones, autonomous driving and industrial inspection systems where real-time decision-making is essential. By processing inference locally rather than sending data to the cloud, edge AI can reduce latency and lower privacy and security risks.
Alsemy presented low-power, high-efficiency NPU design technology. As AI semiconductor competition shifts from raw performance to power efficiency and product applicability, design-focused companies such as Alsemy can provide AI acceleration technologies optimized for industrial equipment and embedded devices. A Barristor Company introduced next-generation semiconductor materials and device technologies, taking a materials-level approach to the power, heat and integration challenges facing AI chips.
3D Architech showcased 3D stacking packaging technology, an increasingly important field as AI chip performance competition intensifies. AI semiconductors need to place more compute units and memory in smaller spaces, making inter-chip connection speed and thermal management critical. 3D stacking enables high-density chip structures and can improve both performance and power efficiency by reducing data movement distances.
K-AI expands into robotics and AIoT
Beyond AI semiconductors, Korean companies were active in robotics and AIoT, or AI combined with the Internet of Things. COSMOS LAB, BWAVE and others introduced Physical AI scenarios in which AI interacts directly with the physical world. AIoT combines sensors, networks and AI inference to collect data, make decisions and control equipment on site.
The direction shown by COSMOS LAB and BWAVE points to a shift from IoT systems that merely collect data to AIoT systems that make decisions. For Physical AI to work in industrial settings, factory equipment, logistics devices, security cameras and robot platforms must read data and respond in real time. Edge inference is especially important in environments where cloud connectivity is unstable or immediate reaction is required.
PC hardware company Zalman also sought new opportunities by presenting cooling and thermal-management solutions, which are becoming more important as AI PCs expand. AI PCs and high-performance workstations generate heavier computing loads and more heat than conventional PCs. Zalman’s cooling and case technologies serve as essential infrastructure that allows high-performance GPUs and AI accelerators to operate reliably.
Regional economic bodies, including the Busan-Jinhae Free Economic Zone and Daegu-Gyeongbuk Free Economic Zone, also participated to promote Korea’s AI and semiconductor clusters and attract overseas investment. Their role went beyond showcasing individual companies. They promoted Korea’s industrial locations, manufacturing base and R&D ecosystem. As the global AI supply chain is reorganized, regional clusters are becoming important factors in attracting companies, joint research and production bases.
New opportunities in Edge AI and Physical AI
The common strategy among Korean companies at COMPUTEX was clear: low-power Edge AI and Physical AI. Industry executives expect the AI sector to evolve through a combination of cloud-based AI centered on data centers and Edge AI that can make immediate decisions and act on site.
A hybrid architecture combining high-performance platforms such as NVIDIA’s Jetson Thor with ultra-low-power NPUs such as those from DEEPX could emerge as a new industrial standard. High-performance platforms are suited to complex robot control and multimodal perception, while ultra-low-power NPUs are well suited for continuous AI inference inside cameras, sensors and compact devices. Physical AI is therefore likely to become a systems competition that combines cloud and edge, memory and displays, sensors and robots.
Large Korean companies are building the foundation of the AI ecosystem through memory, displays and system infrastructure. Startups, meanwhile, are differentiating themselves through power efficiency, cost competitiveness and deployable solutions for real industrial sites. Industry officials at the show said Korean firms are expanding their influence beyond component supply into AI system design and applications.
From AI supply chain to AI solutions
COMPUTEX 2026 showed both the current position and future direction of Korea’s AI industry. Korea has long been a key pillar of the global AI supply chain through its strengths in memory, displays and semiconductor manufacturing. At this year’s exhibition, however, there was a clear move beyond component supply toward Physical AI solutions that can be used in real industrial environments.
If the innovation of ultra-low-power AI chip startups is combined with the infrastructure strength of Samsung Electronics, SK hynix and LG Electronics, Korea could play a more important role in the global robotics and Edge AI markets. As AI expands from data centers into factories, logistics centers, hospitals, vehicles and homes, power efficiency, reliability and field deployment will become decisive sources of competitiveness.
The Korean booths at COMPUTEX 2026 were not simply exhibition spaces. They were small laboratories showing how AI is moving beyond data centers and into the real world. Korea’s next task is no longer just to supply excellent components. The new standard of competition will be how quickly and reliably it can deliver AI solutions that work in actual industrial settings.
Editorial Contact: Korea IT Times Technology Team (monica@koreaittimes.com | info@koreaittimes.com)
