Faraday Future aims to build a US robotics business by establishing a US robot factory by the end of 2026, with the first US-assembled robot scheduled for February 2027. The company seeks to localize production to enhance regulatory compliance, supply chain resilience, and quality control, rather than just moving assembly lines. This transition is supported by leveraging existing EV manufacturing infrastructure to create synergies with robotics development. Faraday Future views vehicles and robotics as complementary parts of a larger physical AI ecosystem, sharing core technologies like AI, perception, and computing. Success is defined by establishing a sustainable US-based physical AI and robotics ecosystem through product launches, manufacturing capacity, and supply chain growth.
Physical Intelligence is reportedly seeking at least $1 billion in funding for a valuation of approximately $20 billion. The company is opening its models to external developers via an API beta to facilitate robot model training and inference. Existing investors Avenir and Lux Capital are in discussions to lead this potential financing round. The company's ambition is to create a shared intelligence layer that can learn across different robot bodies and applications. This development aims to make their research accessible for building real-world applications.
Unitree's third-quarter revenue increased by approximately 70% to 665 million yuan. Concurrently, the company's net profit attributable to shareholders decreased by 17.4% to 112.7 million yuan. Research and development spending significantly escalated by 249.6% to 127 million yuan. This increased investment was directed toward embodied-intelligence models and computing capacity. Despite strong sales growth, the profit did not keep pace with the expanding revenue.
Living digital twins combine high-fidelity operational data with control systems to mirror industrial plant dynamism. AI is central to evolving these twins into adaptive, living models that can intervene in the physical world. Agentic AI enables systems to act autonomously in select areas, balancing human oversight to define acceptable operational boundaries. This requires robust data synchronization through technologies like IoT sensors and predictive models to create a connected operational record. Ultimately, this integration across the value chain promises greater resilience and sustainable industrial operation.
North American robot orders grew 24.8% in units and 26.1% in value in the third quarter of 2026. This growth was driven by positive returns in automotive OEM orders and significant increases across other sectors like semiconductors, life sciences, and plastics. Collaborative robots saw a substantial 56% increase in units during the same period. This positive momentum occurred despite earlier declines in the first half of 2026 for some segments. Overall, the market shows increasing diversification across various industries.
Warp Sciences raised $215 million to create chemistry infrastructure using robotics and AI for faster drug discovery. The company is building a U.S.-based contract research organization to combine chemists, robotics, and AI for molecule development and testing. Warp aims to generate real-world data to train AI, enabling labs to test more ideas more efficiently. The company focuses on lead optimization analog synthesis, allowing scientists to quickly compare molecular variations. Warp emphasizes building a sovereign biotechnology partner to ensure speed, quality, and intellectual property protection within the U.S.
AGIBOT introduced Worken, a framework to measure robot productivity using the formula: Worken = Active Robots × Autonomous Working Hours × Output Efficiency Ratio. The company also unveiled new hardware like AHA3, which integrates locomotion, interaction, and task execution capabilities. Newer AI models include GO-3 Preview for reasoning and action, and WITA-Omni 1.0 for human-robot interaction. The G2 Max robot is designed for industrial handling with a 40 kg dual-arm capacity and a working height range of 0.1–2.5 meters. Finally, Link-U OS 1.0 is an open-source operating system with a roadmap aiming for full open-sourcing by the end of 2026.
Robbyant will send 15 teams to a Shanghai final for its Embodied AI Challenge, a 48-hour hackathon focused on adapting models to physical robots. Competitors will use AgileX Robotics’ Cobot Magic platform for this task. The challenge is based on the LingBot-VLA 2.0 model, with over 1,800 teams registering initially. Final details regarding scoring criteria and robot access are still being finalized. The event aims to assess how effectively vision-language-action models can be adapted to physical hardware.
Audi is testing Agile Robots’ mobile humanoids for sequencing automotive parts at its Ingolstadt Production Lab. The project involves picking components from containers and arranging them for assembly, addressing a material-supply operation scale of one million picks daily. The system utilizes dual-arm manipulation and a mobile base to handle varied, unsorted components in logistics environments. Virtual testing using a digital twin supported by NVIDIA Omniverse and Dassault Systèmes is used to assess feasibility before wider deployment. Success in the final deployment will depend on consistent physical performance metrics like pick rates and operating costs.
MIT engineers developed a robotic enteroscope inspired by snakes to navigate the small intestine for easier diagnosis and treatment. This soft robot uses multifunctional actuators and textured coatings to move through narrow passages by mimicking snake locomotion. It is equipped with tools for imaging, drug delivery, and taking biopsies of polyps. The system allows external controllers to guide the robot through the intestine. Animal studies demonstrated the robot's capability to travel several feet while performing these interventions.
Neura Robotics and the Technical University of Munich opened a €17 million robotics center near Munich Airport. This center features the TUM RoboGym, a facility combining robotics research, training, and industrial applications in over 2,300 square meters. The initiative allows companies across ten industries to train robots in real-world tasks like handling and loading materials. Neura also unveiled its latest humanoid robot, the 4NE1 Gen 4, which is designed for series production and powered by Neuron AI. The partnership aims to advance physical AI capabilities and establish an international network of training facilities.
Maxinsights raised $30 million to expand its physical AI data platform, which pays contractors to record real-world activities for robot and AI training data. The company has collected over 2 million hours of human activity, with 1.5 million hours being model-ready, ensuring exposure to real-world variations. They developed MaxLoop for footage quality monitoring and MaxGlove for capturing tactile information like grip force. This data is crucial for training models like Genesis AI, which has used Maxinsights data for its GENE-26.5 model. Maxinsights emphasizes that capturing real-world data, including edge cases, is essential for deploying robots effectively.
A wheeled humanoid robot from Yue Quan Bionics features a compact base and an articulated upper body for manipulation in confined spaces. It stands 178 cm tall, uses folding joints for reach up to 2.2 meters, and possesses 43 degrees of freedom for its arms, each capable of lifting 10 kg. The robot is equipped with onboard computing, depth cameras, LiDAR, and force sensors for environmental awareness and interaction. Its primary functions include sorting goods and assembling products within constrained work areas. The robot is commercially available and is designed for workflows emphasizing reach and compact positioning.
CYBORG H02 is a robotic hand from Cyborg Robotics, designed with five fingers and a tendon mechanism inspired by the human hand. It weighs 750 grams and has dimensions of 248 × 114 × 52 millimeters. Six electric motors drive the tendon architecture, defining its actuation and joint design. The hand received a Humanoid.Guide Hand Score of 7/10, with a Strength rating of 4/5 and a Degrees of Freedom rating of 3/5. The product is listed as in production with an advertised price of USD 4,000.
Mornine is a full-size bipedal humanoid developed by AiMOGA, a Chery Group subsidiary, for commercial environments. It measures 167 cm tall, weighs 70 kg, and features an aluminum alloy structure with electric joint actuators. The robot has a maximum speed of 3.6 km/h and can sort goods and avoid obstacles, with a 3 kg strength capacity. It is currently in production in China and is priced around US$41,400.
TORCH HAND 2 is a robotic hand from TorchKernel Robotics, rated 6 out of 10 on the Humanoid.Guide Hand Score. It excels in dexterity with a 4 out of 5 rating for degrees of freedom but has a lower 2 out of 5 rating for strength. The hand measures 200mm by 105mm by 50mm and weighs 0.48 kilograms, priced at USD 6,500. This balance suggests the hand is more focused on dexterity than gripping power. The product is currently in production, offering a commercially available option for robotic systems.
NE01 is a 145 cm tall, 40 kg humanoid robot developed by GENISOM AI for research, commercial performance, and tourism. It features an aluminum alloy structure with integrated joint modules, capable of moving at 10 km/h and exerting up to 180 Nm of torque. The robot has verified capabilities in obstacle avoidance and sorting goods, but its navigation and manipulation skills are rated very low (1 out of 5). Despite its design intent for mass production, NE01 is currently considered a preliminary platform with demonstrated, limited functionality.
Batteries for Xpeng's Iron robot have entered small batch production, though the robot itself is not yet in formal production. Xpeng announced a dedicated solid-state battery design tailored for the robot's limited space, emphasizing safety features like impact resistance. However, the specific battery chemistry remains ambiguous, with reports suggesting a solid liquid hybrid or semi-solid state was used instead of all solid-state. Xpeng plans to achieve mass production of Iron by the end of 2026, with initial commercial roles including in-store assistance.