At a time of rapid technological development, the convergence of the Internet of Things (IoT) and artificial intelligence (AI) is reshaping various fields. Arm, a giant in chip architecture, has set its sights on the South Korean market and is looking to further expand with AI chip designs built specifically for IoT devices.
On February 27, 2025, President Hwang Sun-wook of Arm Korea unveiled the company's latest platform for edge AI for IoT devices, the Armv9 Edge AI platform, at a press conference in Seoul. The platform integrates the design of the Cortex-A320 CPU, the Ethos-U85 neural processing unit, and is equipped with the associated architecture and software tools to support developers from silicon production to other solution builds.
Artificial intelligence at the edge, in simple terms, enables devices such as smartphones, self-driving cars, home appliances, and sensors of all kinds to directly handle AI tasks without relying on remote servers. As IoT devices become more intelligent, more and more devices are equipped with AI functions, and the load of AI in cloud computing is becoming more and more heavy, which makes the importance of AI at the edge more prominent. A large amount of data is processed on local devices, which not only reduces data transmission latency, but also reduces dependence on the network and improves the user experience. In areas such as smart homes, smart cities, and industrial automation, AI at the edge has become a key force driving innovation.
The Armv9 platform is a high-performance platform with an 8x increase in machine learning performance compared to the Cortex-M85-based platform introduced last year. Among them, the new Cortex-A320 CPU is optimized for IoT applications and excels in energy efficiency. It leverages the features of the Armv9 architecture to deliver a 10x improvement in machine learning performance and a 30% improvement in scalar performance over the previous generation Cortex-A35. The Cortex-A320 completes Arm's CPU product line, from ultra-efficient models for IoT devices to high-performance models for servers, providing partners with a wealth of choice.
Figure: Arm is expanding in the Korean market with the design of IoT AI chips
Huang said that from 2026, it is expected that customer companies will start using the Cortex-A320 to produce physical chips, although these chips may not require a state-of-the-art manufacturing process such as 2nm. The Cortex-A320 and ETHOS-U85 are already being evaluated by a number of customers for a wide range of home appliances and chips.
Previously, Reuters reported in January that Arm was developing a long-term strategy to increase licensing fees by up to 300%, a move that could have an impact on chip and smartphone makers, including Samsung Electronics. At the press conference, Hwang Sun-wook said he was inconvenient to respond to questions about the plan to increase the licensing fee, but stressed that Arm's goal is to help Korean partners succeed. He mentioned that South Korea has a large number of customers using Arm's solutions, and that South Korea has probably the highest percentage of users in the world. Arm wants to ensure the continued success of many startups that build and validate their solutions and products on Arm's technology to develop competitive products that are a springboard to global markets.
Arm's presence in the Korean market with the Armv9 Edge AI platform is expected to drive the development of IoT AI devices in South Korea, but the potential adjustment of licensing fees also adds uncertainty to its relationship with partners, and the future development is worth keeping an eye on.
