NTT DATA, Hyster-Yale Materials Handling and Archetype AI have deployed what they describe as a first-of-its-kind physical AI system embedded directly into a manufacturing assembly line, as APAC manufacturers weigh how to move AI from pilot projects to production.

The system, unveiled earlier this month, uses vision sensors, edge AI and physical AI models to verify assembly quality in real time, flagging missing components or deviations before products move to the next stage. According to NTT DATA, early results showed deployment timelines cut from months to weeks compared with legacy techniques.

Intent is high, but proof points are scarce

The deployment lands against a backdrop of rapid but largely untested enthusiasm for physical AI across manufacturing. Deloitte’s April 2026 paper, Physical AI: The moment of acceleration, found that while only 5% of firms today say physical AI is transforming their organisation, 41% expect it will within three years — a six-fold jump in extensive integration over two years. Separately, Rockwell Automation’s 2026 State of Smart Manufacturing Report found 71% of APAC manufacturers plan to increase their use of AI and machine learning over the next 12 months.

The gap between intent and execution is what makes deployments like the NTT DATA, Hyster-Yale and Archetype AI collaboration notable: at-scale proof points on live production lines remain rare, even as manufacturers across Japan, Korea, China and Southeast Asia explore where physical AI can scale first.

Edge processing and private 5G

Central to the deployment is the shift from cloud-centric AI processing to intelligence running on-site, where inference happens where the work occurs rather than in a remote data centre. NTT DATA has also been building out the infrastructure layer to support this shift, including a recent partnership with Ericsson aimed at scaling private 5G networks alongside physical AI deployments — infrastructure that becomes more important as manufacturers move from proof-of-concept trials toward production-grade systems.