OKI cuts AI server PCB inspection time 80% with new AI technology rolling out to Marugoto EMS customers July 2026

Title

OKI cuts AI server PCB inspection time 80% with new AI technology rolling out to Marugoto EMS customers July 2026

Content

OKI Electric Industry, the Japanese electronics and telecommunications manufacturer headquartered in Tokyo, developed an AI-based inspection technology that reduces the manual post-AOI (Automated Optical Inspection) visual check time for large, high-density printed circuit boards used in products such as AI servers by approximately 80%, and is rolling it out to customer production lines served by its Marugoto EMS electronics-manufacturing service from 1 July 2026. The driver for the technology is the rapid growth in size, fineness and layer count of semiconductors mounted on PCBs as AI server adoption expands. Larger parts and narrower-pitch components make solder-defect judgements much harder for AOI equipment to make reliably, and human visual inspection has had to absorb the residual false positives, slowing throughput and creating quality-control bottlenecks for AI-server production. The OKI system uses AI to inspect four categories of defects: solder wetting, misalignment, missing components, and lifted parts. The inspection model is built on OKI's accumulated manufacturing know-how for large, high-density boards, and incorporates a proprietary programme that treats some component-specific variations as acceptable when they do not affect function, including differences in serial or lot numbers and faint printing. Crucially, the system uses a "good-product-only learning" method: the AI is trained exclusively on non-defective data, and judges products based on the degree of deviation from that baseline. This is a deliberate choice for high-mix manufacturing environments handling tens of thousands of material types and thousands of equipment types, where assembling a balanced defect dataset is impractical. The technology was developed for OKI's Marugoto EMS service, which has been operating since 2025 as a one-stop electronics-manufacturing outsourcing model for customers who want to delegate operations and improve management indicators. By embedding the new AI into Marugoto EMS customer production lines from 1 July 2026, OKI expects to reduce the labour cost and time of post-AOI visual checks by roughly 80% while improving inspection accuracy and reducing false-positive soldering-defect calls. The deployment also represents a deliberate long-term bet: OKI plans to extend the same AI-based inspection approach to other inspection tasks in electronics manufacturing beyond post-AOI, and to continue developing AI-based systems for manufacturing sites in general. For AI-server OEMs and their EMS partners in particular, the 80% time reduction on the most error-prone inspection step is a material throughput gain that compounds across high-volume server production.