Physical Intelligence News — 2026-11-05 (Thu) Daily News

Overview

Data and training infrastructure for physical AI drew the most headlines: South Korea opened a humanoid robot data factory, and Baidu, ViTai Robotics with Tencent Cloud, and Docomo announced world model, data and connectivity platforms (EurekAlert!, TechTarget, Gasgoo, AIsmiley). Deployment news was mostly proof-of-concept (PoC) and field tests, though Aeon's first logistics automation site using robots with physical AI is reported to be in full operation (Response.jp). Fierce Network's special report says physical AI is moving toward the factory floor, but full autonomy remains years away. 【Humanoids and general-purpose robots】 - Korea opened its first AI humanoid data factory, which is meant to train AI models on real-world tasks (EurekAlert!, AI Insider). - ROBOTWORLD 2026 opened on November 4 with physical AI and AI-driven manufacturing as themes, and added a new humanoid pavilion (Robostar). - An online seminar will cover NVIDIA's humanoid and physical AI development stack: GR00T, Cosmos, Isaac and Jetson Thor (Robostar). - Mitsubishi Electric presented "the ears of physical AI," a technology that distinguishes abnormal sounds, among a series of autonomous robot technologies (Nikkei xTECH). - Lee Jong-won received the Korea Science and Technology Award for developing a domestic AI humanoid (Chosunbiz). - A video discussion covered humanoid hands, with Harmonic Drive Systems and Tokyo Electron named as companies to watch (Unisba Media). 【Robot foundation models and world models】 - Baidu's new world model highlights both the value and the challenge of physical AI (TechTarget). - ViTai Robotics and Tencent Cloud established a strategic partnership focused on embodied AI data (Gasgoo). - STMicroelectronics expanded its STM32 roadmap for cloud AI and physical AI (EE Times). - Teradyne has a new physical AI push, and Yahoo Finance asks whether it is changing the company's robotics story (Yahoo Finance). - ABB Robotics is co-developing an education program for physical AI (A3 Association for Advancing Automation). - Advantech plans to exhibit a physical AI platform at EdgeTech+ 2026 (Robostar). - NTT Docomo developed a "Physical AI integration platform" that connects AI with diverse robots and adjusts their movements automatically to surrounding conditions (AIsmiley). - Hitachi published part 2 of a column on "SHUGY," a hand-technique AI analysis that makes skilled know-how visible, aiming at error prevention and skills transfer (Hitachi Global). - A column series says the physical AI market operates in seven business layers (Drone Journal).

Manufacturing and factories

- Fierce Network's special report says physical AI is moving toward the factory floor, but full autonomy remains years away (Fierce Network). - ROBOTWORLD 2026 will have 402 exhibitors from 17 countries and focus on physical AI and AI-driven manufacturing (PR Newswire, KuCoin). DN Solutions showed AI-powered manufacturing technologies there (The Korea News Plus). - Changshin adopted Palantir AI to link development and factory feedback (Unite.AI). - ugo will outsource production of its next-generation inspection AI robot, which patrols factories in place of people, to Canon Electronics (Mynavi News). Production of the ugo mini G2 at Canon Electronics' Misato site is scheduled from October 2026 (Robostar). - Mouser published technical information and compatible products for industrial AI and machine learning in factory automation (Mynavi News). - An article describes how AI delivers value in automotive manufacturing, from connected data to virtual factories (Automotive Manufacturing Solutions). - Bloomberg reported on women working in ABB's India factory as part of the AI boom (Bloomberg). - A market report on autonomous factory operations in Japan says physical AI, autonomous robotics and intelligent production systems are creating business opportunities (Dream News). - Panasonic is building a new plant building in Thailand for circuit board materials, to increase production for AI servers (ashu-aseanstatistics.com).

Logistics and warehouses

- Aeon's first logistics automation, using robots with physical AI, is in full operation at Kawaguchi East XD (Response.jp). - KNAPP established a robotics and AI innovation hub in Boston (Modern Materials Handling). - Fujitsu and three other companies are running a demonstration test of AI that judges exterior dirt and damage on home appliance packaging, aiming to reduce returned shipments (Chiba TV, TNGlobal). - JTEKT reported an AI parts-picking success rate of 97.1% and a 16% shorter cycle time (MONOist). - Satellite-AI site searching is expanding to logistics facilities (LOGISTICS TODAY). - Komeito's representative and others visited advanced companies, describing physical AI as a response to serious labor shortages in logistics (Komei Shimbun). 【Cooking, housework and food service】 No notable developments. 【Crafts, manufacturing and artisan skills】 - Lincoln Electric's XpertLINC physical AI platform is presented as redefining robotic automation (Defense and Munitions). - GlobalFoundries sees Singapore as key to equipping humanoid robots with AI chips (Singapore Economic Development Board). - NHK ONE examined whether Japan's artisans can win in the AI era, using the national skills competition as its setting (NHK ONE). 【Construction, field work and mobility】 - Shackleton AI raised a $5M seed round to expand construction agent workflows (Unite.AI). - ONESTRUCTION published a "Construction AI Leaderboard" that scores 20 generative AI models, including ChatGPT and Claude, on construction tasks (Digicon). - Baidu's AI for training self-driving cars recreates road environments, with the aim of reducing development costs (Reuters). - The Liberal Democratic Party's AI and digital society headquarters held an autonomous driving test ride event (Liberal Democratic Party).

Funding and business

- Shackleton AI raised a $5M seed round (Unite.AI). - ViTai Robotics and Tencent Cloud formed a strategic partnership on embodied AI data (Gasgoo). - abc (8783) disclosed on November 4, 2026 that it received an order for a transport PoC in its physical AI robot business from a major domestic glass manufacturer (Nikkei Company Information DIGITAL). - ugo is outsourcing production of its inspection robots to Canon Electronics (Mynavi News, Robostar). - KNAPP opened a robotics and AI innovation hub in Boston (Modern Materials Handling). - Changshin introduced Palantir AI (Unite.AI).

Japan's Physical Intelligence

- Waseda University professor Ogata was interviewed on whether Japan can truly win in physical AI (Yahoo! News, Business+IT). - Komeito reported that physical AI addresses serious labor shortages (Komeito). - Schoolchildren will tackle problems with a dual-arm robot arm and Claude at a physical AI hackathon on November 7 and 8 (Robostar). - Docomo developed a physical AI integration platform (AIsmiley), and abc received a glass manufacturer's PoC order (Nikkei). - Aeon's logistics site is in full operation (Response.jp), and ugo's production is outsourced to Canon Electronics (Robostar). - AI authority Salakhutdinov said South Korea has a chance to lead in physical AI because manufacturing is a strength (d Menu News).

Statements from key figures

No notable developments. 【Integrated analysis: what is happening】 Data and simulation for training appear in several unrelated headlines: Korea's humanoid data factory, the ViTai Robotics and Tencent Cloud embodied AI data partnership, and Baidu's world model that recreates road environments (EurekAlert!, Gasgoo, Reuters). This suggests that building training data is currently a main bottleneck across humanoids, factories and driving.

Demonstrations and deployments need to be kept apart. Exhibitions (ROBOTWORLD 2026, EdgeTech+ 2026), test programs (Fujitsu's packaging damage trial), a PoC order (abc) and platform development announcements (Docomo) are announcements or trials, not production use. Reported deployments are narrower: Aeon's logistics site is in full operation, and ugo's production has been outsourced to Canon Electronics, with the schedule given as October 2026 (Response.jp, Robostar). Fierce Network's assessment that full autonomy remains years away is consistent with this pattern.

Logistics and picking tasks appear to be where measurable results are reported first, such as JTEKT's 97.1% success rate and 16% shorter cycle time (MONOist). That figure comes from the headline alone, and its test conditions are not stated. Readers should treat PoC and trial headlines as early signals rather than evidence of scaled adoption.

What to watch

- Whether PoCs and trials (abc's glass transport, Fujitsu's packaging checks) are followed by announced production use, since this is where announcements may or may not become deployments. - Whether the humanoid data factory in Korea and the data partnerships lead to published results from models trained on that data. - Whether Aeon's full-operation site and ugo's Canon Electronics production produce reported operating results or shipments.