※ URL(Korean): https://www.kisdi.re.kr/bbs/view.do?bbsSn=114701&key=m2101113055776&pageIndex=3&sc=&sw=


KISDI Publishes Report on “Policy Directions for Strengthening the Competitiveness of On-Device AI Semiconductors in the Age of Physical AI” 
– The advent of the Physical AI era is reshaping the AI landscape through “on-device AI semiconductors”
– Policy support needed to systematize the AI semiconductor ecosystem, following the platform strategies of major global AI chipmakers 

The Korea Information Society Development Institute (KISDI, President Sangkyu Rhee) recently released the 21st issue of AI Outlook, which provides an in-depth analysis of the importance of on-device AI semiconductors in responding to the era of Physical AI. The report examines global and domestic trends in technology, markets, and policy, and presents strategic directions for enhancing future competitiveness.

The report analyzes the concepts of Physical AI and on-device AI, forecasts for the on-device AI semiconductor market, and semiconductor companies’ domestic and international strategies for customized and optimized solutions tailored to different device types. Building on this analysis, it reviews current policy support for on-device AI semiconductors and proposes directions and implications for strengthening national competitiveness.

AI technologies are now evolving beyond cloud-based generative AI toward Physical AI, which autonomously perceives and acts within real-world environments. Physical AI refers to technologies that can move beyond simple perception to understand and interact with their surroundings in human-like ways.

Physical AI is realized through Edge AI and on-device AI approaches that utilize lightweight, compact AI models tailored to specific end-use services. These approaches address many of the limitations of large, cloud-based generative AI models—such as high infrastructure costs, energy consumption, latency, and privacy risks.

Most notably, to implement and support Physical AI in real-world autonomous machines and various devices, energy-efficient on-device AI semiconductors are indispensable. Such semiconductors must be custom-developed for each type of device, accompanied by “full-stack” technical support systems that enable the optimized use of these customized chips.

As Physical AI becomes more widely implemented, on-device AI semiconductors are expected to drive the growth of the overall AI semiconductor market. According to Omdia, the on-device AI semiconductor market is projected to surpass the growth rate of cloud and data center AI semiconductor markets starting in 2026, achieving an average annual growth rate of 12.6% through 2029. In contrast, growth in cloud and data center AI chips is expected to slow as those markets mature. While adoption of on-device AI semiconductors is already high in smartphones, PCs, and tablets, penetration in other sectors—such as automobiles, robotics, and security cameras—remains low, suggesting significant potential for continued expansion.

Currently, global AI semiconductor leaders such as NVIDIA and Qualcomm are pursuing platform-based strategies that integrate hardware, software, and service layers across the on-device AI semiconductor ecosystem. In Korea, small and venture firms such as Nextchip (automotive AI SoC), DeepX (NPU-based edge AI), and Mobilint (ASICs for drones and robots) are developing proprietary on-device AI semiconductors and supporting software solutions, creating new market opportunities.

Associate Research Fellow Minsik Kim noted, “In Korea, policy support for AI semiconductors in response to the spread of Physical AI is structured but fragmented across ministries under the ‘AI–Semiconductor Initiative Strategy Plan.’ The Ministry of Science and ICT supports data center AI semiconductor development through the ‘K-Cloud and AI Semiconductor Demonstration Project,’ the Ministry of Trade, Industry and Energy promotes on-device AI semiconductor R&D under the ‘K-On-Device AI Semiconductor Technology Development Project,’ and the Ministry of SMEs and Startups assists fabless AI semiconductor firms through the ‘1,000+ Deep-Tech Startup Project.’”

The report also found that global AI semiconductor companies such as NVIDIA, Qualcomm, AMD, and Intel are responding to market changes with strategies that leverage their respective strengths. In particular, NVIDIA and Qualcomm—leaders in the AI semiconductor field—are providing comprehensive platforms that integrate enabling technologies, AI semiconductors, AI models, and application development to secure and maintain competitiveness in the on-device AI semiconductor market.

Associate Research Fellow Kim emphasized, “Similar to the platform strategies of global AI semiconductor leaders, Korea must establish national support measures to strengthen the competitiveness of its on-device AI semiconductors in the era of Physical AI. Key priorities include: (1) providing targeted support and collaborative networks for weak links in the horizontal value chain of on-device AI semiconductors; (2) enhancing support tailored to diverse end-user needs resulting from small-batch, multi-product production; (3) expanding support for the various technical layers essential to the use of on-device AI semiconductors; and (4) establishing a governance framework for continuous policy updates and monitoring under the national AI strategy jointly led by relevant ministries.”

KISDI AI Outlook can be downloaded from the KISDI website (www.kisdi.re.kr).