Seoul, South Korea, September 29 (Reuters) – SEMIFIVE is a global provider of customized AI semiconductor (ASIC) solutions. The company announced on Monday that it has signed a contract worth approximately 70.3 billion Korean won (52 million US dollars) with an American AI fabless company to develop the next generation of AI inference accelerators.
This project marks SEMIFIVE's first “Spec Hand - off” collaboration in the North American market. This single contract alone exceeds 40% of the company's total orders for 2025, which amounted to 168.4 billion Korean won, and also accounts for about 60% of its new orders in the first half of 2026, which totaled 118.9 billion Korean won. It is by far the largest single contract for SEMIFIVE. The company stated that this major project will help to solidify its presence in North America and strengthen its technical competitiveness in the global custom AI semiconductor ( ASIC ) market.
In the Spec Hand - off mode of SEMIFIVE, customers only need to provide key performance requirements and specifications. SEMIFIVE will be fully responsible for the entire development process, including detailed design, software development, packaging, testing, and mass production. Unlike traditional turnkey collaborations, where customers typically complete chip design first before outsourcing subsequent implementation and manufacturing processes, SEMIFIVE takes the lead from an earlier stage of chip development.
Companies specialized in ASIC are playing a more extensive role in the global custom semiconductor market, starting from the demand phase. For example, large technology companies such as Google and Amazon have collaborated with semiconductor firms like Broadcom and Micron to develop their own AI chips.
For this project, SEMIFIVE will integrate advanced technologies to provide the high memory bandwidth and data processing performance required for large-scale AI model inference, while reducing power consumption and total cost of ownership ( TCO ). The accelerator will utilize LPDDR6 – a next-generation low-power DRAM technology – to reduce power consumption related to memory; it will also employ PCIe Gen5 to minimize bottlenecks in large-scale data transfer. Based on its experience in multiple projects with areas of up to 800 square millimeters, SEMIFIVE also plans to adopt an architecture that balances high computational throughput with operational stability.
According to the plan, the project will enter the chip manufacturing phase in the first half of 2027, followed by mass production on a global scale starting in 2028 by the two companies. The target customers are large-scale cloud service providers and cloud service suppliers worldwide ( CSPs ).
SEMIFIVE, Chief Executive Officer and Co-Founder, Brandon Cho stated: "As the AI inference market continues to expand rapidly, the demand for custom chips optimized for customer services and workloads is growing at an accelerated pace. By successfully commercializing this next-generation inference accelerator and improving the economics of large-scale AI models, we aim to position SEMIFIVE as an indispensable partner in the development of custom chips for the global hyperscale infrastructure market."
