This summary covers the 14-day period from 2 to 15 September 2026. It reviews key physical AI news, centered on the impact of deployments at Japanese companies, the expansion of service models, and overseas efforts to secure foundation models and computing resources.
Japanese companies have announced cases that directly reduce staffing requirements in operational settings, in addition to developments involving general-purpose robots. Mujin announced the automation of mixed-load pallet unloading at 松田電機工業所, stating that staffing in the target process had been reduced from 3 to 0 and that 5,000 boxes were unloaded daily. This has reduced the operational burden at the site and increased the importance of using operational data to assess investment payback.
Meanwhile, Hitachi launched an operational data platform centered on HMAX, together with operational support, demonstrating a framework that continuously supports not only the collection of operational data but also the updating of decision rules and AI operations. FANUC America and Palladyne AI are testing the integration of learning capabilities into existing industrial robots. Panasonic Connect presented a service and pricing model for centrally managing AMRs from multiple manufacturers. These initiatives target recurring revenue from ongoing site operations and updates, rather than hardware sales alone.
In overseas markets, Unitree announced UnifoLM-WLA, a foundation model for humanoid robots, while Nscale and Figure announced a strategic partnership based on a long-term plan. Unitree said the former can handle 64 tasks with a single model, while Nscale and Figure presented a financing plan that includes deploying up to 100,000 GPUs. In addition, NEURA Robotics and SECO are advancing the joint design of computing architectures, while Vecna Robotics has raised funding to strengthen field deployment. Expansion is progressing across the industry on both fronts: operational deployment and computing resources.