Physical AI News (June 30, 2026 Issue)

Update Date: 2026/6/30

Executive Summary
On June 29, 2026, the Physical AI sector saw simultaneous progress in real-world factory applications, commercial deployment, and foundational modeling. With the production of 15,000 G2 units and over 64 hours of live factory verification, AGIBOT demonstrated that humanoids have entered a stage where they provide value on mass production lines. DEEP Robotics presented a strategy integrating a general-purpose brain, multi-form hardware, and industry-specific solutions, while XPENG unveiled a vision to deploy autonomous driving, robotaxis, and humanoids on the same AI foundation, centered on world models. Additionally, with implementations expanding across retail, painting, machining, warehousing, and robotaxis, Physical AI is transitioning from research demonstrations to industrial infrastructure.

Note: I have input the created article content into Gamma to automatically generate slides. Please take a look here if you find slides easier to view.

1️⃣ AGIBOT G2: 15,000 units produced and over 64 hours of live factory verification

📎 Source: AGIBOT Official (15,000 units line-off) / AGIBOT Official (Factory live verification)
On June 28, AGIBOT announced the line-off of the 15,000th unit of its industrial embodied task robot, the 'AGIBOT G2,' and on the 29th, it released the results of a 6-day live factory verification at the Longcheer Technology Nanchang plant. Multiple humanoids operated for over 64 hours in the quality inspection process of a tablet mass production line, contributing to 64,828 production tasks, a 99.99% success rate, and a cumulative line-out of 17,625 units. This indicates a shift from standalone demonstrations to a stage of measuring production line value.

2️⃣ DEEP Robotics '1+X+N': Commercial deployment of general-purpose brains and multi-form robots

📎 Source: Newsfile / DEEP Robotics
On June 29, DEEP Robotics held its 2026 Global Partner Conference in Hangzhou, gathering over 500 international partners to announce its '1+X+N' strategy. '1' refers to the general-purpose embodied brain that integrates perception, decision-making, and motion control; 'X' refers to multi-form hardware such as humanoids, wheeled-legged robots, and quadruped robots; and 'N' refers to standard solutions for energy, manufacturing, logistics, and public services. They also showcased the 2026 commercial deployment of the all-weather industrial humanoid DR02, a factory with an annual capacity of 20,000 units, and a deployment network spanning over 45 countries.

3️⃣ XPENG X-Mind: Presenting a world-model-based autonomous control brain

📎 Source: XPENG Official (X-Mind) / XPENG CVPR 2026 Announcement
XPENG participated in CVPR 2026 (June 3, Denver) as an invitee for the third time, unveiling its full technical blueprint for Physical AI and world models. VLA 2.0 has entered the mass production stage, achieving over a 50% share of assisted driving mileage in its first month. As world model technologies, they announced X-World, X-Foresight, and X-Cache, with X-Cache accelerating denoising backbone inference by up to approximately 2.7 times. X-Mind is scheduled to be released soon as a technical report. They presented a framework for deploying autonomous driving, robotaxis, and humanoids on the same Physical AI foundation.

4️⃣ Striding AI: Robot foundation system using World Action Models

📎 Source: GlobeNewswire / Striding AI
On June 29, Striding AI announced that it is developing a robot foundation system for real-world deployment, centered on World Action Models and next-generation reinforcement learning. Initially targeting retail, they aim to start with shelf replenishment, inventory counting, product organization, and checkout assistance, proposing a design that handles perception, planning, execution, feedback, and recovery in a closed loop. They stated that internal initial verification of Human-in-the-loop RL improved task success rates by up to 3 times, aiming to build a flywheel that feeds real operational data back into continuous learning.

5️⃣ Momenta Global: Accelerating investment in autonomous driving AI and Robotaxis via Hong Kong IPO

📎 Source: Reuters
Autonomous driving AI company Momenta Global began its Hong Kong IPO on June 29, aiming to raise up to 5.89 billion Hong Kong dollars (approximately 751 million USD), as confirmed by Reuters. The number of shares offered is 19.9 million, with trading scheduled to begin on July 8. The company plans to allocate approximately 60% of the proceeds to R&D, including AI computing resources, data storage, and engineering, and approximately 20% to accelerating Robotaxi deployment. As of the end of 2025, over 680,000 vehicles were equipped with the company's software, and they anticipate expanding overseas Robotaxi trials.

6️⃣ Hirebotics Cobot Painter: Explosion-proof, no-code collaborative painting robot

📎 Source: Robotics 24/7 / Hirebotics Official
At Automate 2026, Hirebotics announced the 'Cobot Painter,' which combines FANUC's explosion-proof collaborative painting robot CRX-10iA/L Paint with the company's no-code Beacon. Designed for use in hazardous environments for liquid painting, powder coating, and gel coating, they emphasize that it can be introduced into existing manual painting booths, avoiding the need for dedicated painting cells or large-scale exhaust and conveyor modifications. With click-and-teach tablet operation, the goal is to expand finishing automation without the need for specialized robotics personnel.

7️⃣ RoboDK CAM: Reducing robot machining deployment time by up to 40% with CAD integration

📎 Source: Automate Show / RoboDK / Robotics 24/7
At Automate 2026, RoboDK showcased 'RoboDK CAM,' which generates toolpaths for robot machining directly from CAD files. According to A3 Automate exhibitor information, it can reduce robot machining deployment time by up to 40%, decreasing the burden of manually building and adjusting robot code. In addition to CAM support, RoboDK 6.0 emphasizes speeds up to 10–100 times faster for collision checking and geometry processing, as well as support for over 1,400 robot models.

8️⃣ Infinium Robotics: AI warehouse drones to become a European listed platform

📎 Source: TechNode Global / Business Today Malaysia
On June 29, Germany-listed Deutsche Defence Beteiligungen AG (DDB AG) acquired Singapore-based Infinium Robotics for 95.6 million ringgit (approximately $22 million). The transaction was completed through a share swap involving the issuance of up to 20.5 million new DDB AG shares. Infinium automates warehouse inventory checks using AI autonomous drones and has a track record of implementation in logistics, FMCG, manufacturing, and electronics sectors. Based on customer reports, it is said to have reduced inventory costs by over 80%, signaling that aerial scanning physical AI is becoming a widespread option for warehouse automation.

General Overview

The key feature visible from the topics on June 29, 2026, is that the competitive axis of physical AI has shifted from 'performance of individual robots' to 'operational systems that continuously collect field data.' AGIBOT's real-factory verification, Striding AI's closed-loop learning, XPENG's world models, and Momenta's R&D investments are all premised on iterative learning and scaling in the real world. Meanwhile, moves like those of Hirebotics and RoboDK to lower adoption barriers by connecting to existing equipment and CAD assets are also significant. In other words, general-purpose intelligence and field-adaptive automation are progressing in parallel, and for adopting companies, the phase has arrived where they must specifically choose 'which site to apply AI to first.'

Points to Watch in the Future

• AGIBOT's live factory verification indicates that humanoid evaluation has entered a stage where it is measured by uptime, success rates, and production contribution rather than just walking or grasping demos.

• DEEP Robotics' '1+X+N' strategy symbolizes the trend of robot companies locking in customers with general-purpose AI brains and industry-specific packages rather than standalone products.

• From the moves of XPENG and Momenta, I want to pay attention to the possibility that world models and driving data accumulated in autonomous driving will be cross-deployed to Robotaxis and humanoids.

• For Striding AI's World Action Models for retail, the focus will be on whether the mechanism of turning failure data obtained in stores and warehouses into continuous learning becomes a competitive advantage.

• The cases of Hirebotics and RoboDK show that retrofitting automation to existing equipment, rather than completely new lines, will accelerate the adoption of physical AI in the manufacturing industry.

• The acquisition of Infinium Robotics strengthens the trend that not only ground robots but also aerial scanning AI drones will become a powerful option for warehouse automation.