Physical Intelligence News — 2026-10-16 (Fri)

Overview

Physical AI news in the past 24 hours centered on data infrastructure, factory and warehouse deployments, and inspection work. Voxel51 and Hesai announced data tools for physical AI teams, and JDSC and Daifuku reported a 97% sorting success rate in a logistics verification (PR Newswire, Engineering.com, Dellows News). Nvidia CEO Jensen Huang called for controls on OpenClaw-related robotics technology (TechRadar). 【Humanoids and general-purpose robots】 - A "Hama Robo Expo" in Hamamatsu, Shizuoka brought together 42 companies, including major manufacturers, to show advanced humanoids and AI-equipped robots (d Menu News). 【Robot foundation models and world models】 - Voxel51 announced Voxel51 3.0 as data infrastructure for physical AI, and Unite.AI described it as a unified data platform aimed at physical AI teams (PR Newswire, Unite.AI). - Hesai launched a 3D data engine for physical AI (Engineering.com). - Rivian CEO RJ Scaringe said physical AI will have its "science fiction" moment by 2030 (Inc.com). - ABB Robotics brought industrial-grade physical AI to fashion logistics (ABB). - H2L announced "BodySharing Physical AI," which learns human motion and force, and reported a 95% task success rate from 15 minutes of experience data (au Web Portal). - NEC will present physical AI and next-generation road control technology at the ITS World Congress 2026 (Response.jp). - A physical AI robot is in a demonstration experiment on coexisting with people in Marunouchi (TV Tokyo BIZ). - Business Network ran a piece on building security and safety into the design of physical AI systems (Business Network). - Mitsubishi UFJ eSmart Securities published a piece on analog semiconductors, for which demand is rising on physical AI and other uses (Mitsubishi UFJ eSmart Securities).

Manufacturing and factories

- Standard Bots announced a new AI platform that lets any manufacturing worker become a robot operator (The National Law Review). - Manufacturing Dive reported on what manufacturing plant managers think of the AI and automation boom (Manufacturing Dive). - WION and The Economic Times covered how factory automation is changing, including the move of computer vision, robotics and edge automation into the real world (WION, The Economic Times). - Shandong Province is upgrading manufacturing with industry-specific AI, with field-led industrial innovation accelerating (China Economic News). - Yamaha Motor will give a special lecture at iPros AI-Industry World 2026 (Dellows News). - Nvidia and semiconductor industry leaders are promoting AI use across chip design and manufacturing for an "AI-defined fab" (Niconico News). - AI Data Inc. announced "AI Komei on," a production optimization AI positioned as the step after generative AI (Niconico News).

Logistics and warehouses

- Ultra raised $62M in funding to expand deployment of its AI-powered warehouse robots (AI Insider). - JDSC and Daifuku achieved a 97% sorting success rate in a physical AI verification for logistics (Dellows News). - Business Insider reported that AI companies are paying workers worldwide to record their jobs so robots can learn to do them (Business Insider). - UBTech signed FAW-Volkswagen, and LDRobot moved to buy Germany's Sabo, in a China physical AI roundup that also listed C&D deals and funding rounds (AI Insider). - Morrisons rolled out an AI-powered workload planning tool across stores and its logistics network (Retail Systems). - Haoxin is evaluating the AI-powered LogiAgent platform (Pluang). - The AI inspection platform "AUDER" won a 2026 Good Design Award (PR TIMES). - IFS Softeon announced "Fulfillment Flow," context-aware AI for warehouse operations (Infoseek). 【Cooking, household and food service】 No notable developments. 【Crafts, manufacturing and artisan skills】 - Murata Manufacturing's glove with a 3-axis LF antenna captures fingertip movement, aiming to turn artisans' tacit knowledge into physical AI (EE Times Japan). - SCIEN began offering "EnCite OS," an AI platform that passes on manufacturing artisans' "memory" inside companies and turns it into a corporate asset (PR TIMES). 【Construction, field work and mobility】 - Osaka Gas is using autonomously piloted drones to inspect gas production facilities and analyzing damage with AI (Nikkei). - Seibu Railway is using an AI robot to patrol and inspect inside a substation, aiming to make maintenance more efficient (Asahi Shimbun via Yahoo! News). - TomTom partnered with DeepRoute.ai to advance AI-based autonomous driving technology (PR TIMES). - The Atlantic reported on how Ukraine plans to paralyze Moscow airports with AI-guided drones (UNITED24 Media). This is a military topic and is covered in the international politics section.

Funding and business

- Ultra raised $62M in funding to expand deployment of its AI-powered warehouse robots (AI Insider). - UBTech signed FAW-Volkswagen, and LDRobot moved to buy Germany's Sabo (AI Insider). - TomTom partnered with DeepRoute.ai on autonomous driving technology (PR TIMES). - SCIEN began offering EnCite OS, and Standard Bots announced its new AI platform (PR TIMES, The National Law Review).

Physical Intelligence in Japan

- Toyo Keizai Online argued there is no need to look for a "Japanese Nvidia," because physical AI plays to the strengths of Japanese companies that can supply the "five senses, muscles, joints and nerves" (Toyo Keizai). - Japanese activity in this period included the Hamamatsu robot expo, the JDSC-Daifuku verification, H2L's announcement, Murata's glove, and the Osaka Gas and Seibu Railway inspections (d Menu News, Dellows News, au Web Portal, EE Times Japan, Nikkei, Asahi Shimbun).

Remarks by key figures

- Jensen Huang, Nvidia CEO, praised OpenClaw as helping the future of robotics but called for controls, "so we can realize the incredible future potential of the technology that we're building" (TechRadar). 【Integrated analysis: what is happening】 Several items share one structure: the problem is moving from robot hardware to the data that teaches robots. Voxel51 and Hesai announced data tools, H2L reported learning from short experience data, Murata and SCIEN target artisans' tacit knowledge, and Business Insider reported workers being paid to record their jobs (PR Newswire, Engineering.com, au Web Portal, EE Times Japan, PR TIMES, Business Insider). This suggests competition is likely shifting toward who can collect and structure human work data.

Demonstrations and deployments need to be told apart. The Hamamatsu expo, the Marunouchi experiment and the JDSC-Daifuku verification are exhibitions or trials (d Menu News, TV Tokyo BIZ, Dellows News). Osaka Gas's drone inspections and Seibu Railway's substation patrols are described as in use or aimed at maintenance work, and Ultra's funding is aimed at expanding deployment (Nikkei, Asahi Shimbun, AI Insider). The headlines do not confirm mass production or scale of operation for most items. Reported figures such as 97% and 95% come from the announcing parties' own tests, and the headlines do not show how they hold up in daily operation.

Safety is the second theme. Huang's call for controls and the Business Network piece on building security and safety into design both point to governance concerns rising alongside deployment (TechRadar, Business Network).

What to watch

- Whether the verification results (97% sorting success, 95% task success) hold up in regular operations at customer sites, which would show in follow-up announcements. - Whether data-collection approaches, such as paid job recording and wearable gloves, scale beyond pilots; this appears to be the key constraint on robot learning. - What form the controls Huang called for take, and whether safety design requirements become part of physical AI procurement.