Tech • AI • Robotics • Game
China's New Giant AI Robot Is Absolutely Terrifying (Stronger Than 10 Men)
7/10TL;DR
Robotics developers in China, Europe, and the United States are pushing humanoids and dexterous machines toward industrial use through heavy-load hydraulics, mobile manipulation, faster charging, and quicker prototyping.
KEY POINTS
China unveils 4-meter humanoid
Neodra Dynamics Robotics presented Neo1 at the 2026 China International Industry Fair with JD Industrial, positioning the machine for heavy industrial work rather than service tasks. The robot stands about 13 feet tall, weighs roughly 1.2 tons, and is described as a 4-meter-class humanoid. The company says its two arms together can handle a 500-kilogram payload.
Hydraulic-heavy design targets force and durability
Neo1 has 41 active degrees of freedom, with 39 joints driven by in-house miniature digital hydraulic linear actuators and only two forearm rotational joints using electric drives. Each actuator weighs about 1 kilogram, while the company claims continuous thrust measured in tons. The design reflects a bet that hydraulics offer better power density, shock resistance, and tolerance for harsh environments than pure-electric humanoids.
Precision claims stand out but remain unverified
Neodra says the robot can achieve repeatable positioning accuracy of plus or minus 0.005 millimeters despite its size and lifting capacity. If validated, that would suggest a machine capable of both heavy handling and delicate assembly. Independent testing is still needed to confirm real-world endurance, control, sealing, and sustained industrial performance.
Target markets include hazardous sectors
The company is aiming at the so-called four D jobs: dangerous, dirty, difficult, and dull work. Proposed uses include heavy-component assembly, port cargo handling, nuclear fuel transfer and maintenance, subsea equipment servicing, mining, demolition, and disaster response. Those remain intended applications rather than confirmed deployments.
JD Industrial partnership points to pilot programs
The alliance with JD Industrial gives Neodra access to industrial customers and supply-chain support for pilot projects. The companies plan to match the robot to specific workflows before any wider rollout. Neodra is also promoting its hybrid hydraulic-electric drive system as a broader robotics platform, not just a single humanoid product.
ETH Zurich turns a robotic hand into a mobile robot
Researchers at ETH Zurich trained an anthropomorphic hand from Wuji to support its own weight and move using its fingers as legs. The untethered system carries onboard power and computing, can crawl, steer, recover from falls, and still manipulate objects while balancing on itself. The work reframes a hand from an end effector into a compact mobile manipulator.
Learning system combines locomotion and manipulation
The hand was trained with reinforcement learning in simulation calibrated against measurements from the real device. In demonstrations, it pressed keyboard keys without vision and used overhead camera feedback to push objects toward targets. The challenge was that fingers differ in size and behavior, unlike the more uniform limbs of quadruped robots.
Teleoperation is becoming a data pipeline
Wuji is also developing the Wuji Hand 2 and Wuji Glove, shown at IROS 2026 in Pittsburgh. A human operator can wear the glove and have the robotic hand mirror movements in real time. Beyond remote control, such systems are increasingly used to collect grasping and tactile data for training embodied AI models, with support for NVIDIA Isaac Teleop.
Agility targets battery downtime as a commercial bottleneck
In the United States, Agility Robotics selected Niobolt battery technology for Digit 5, aiming for more than 20 hours of productive work in a 24-hour period. The target is a 9-minute recharge for 90 minutes of runtime, a planned 10:1 run-to-charge ratio. That would mark a major increase from the 2:1 ratio reported for Digit 4.
Digit 5 rollout and public listing advance
Agility says the improved charging profile is central to warehouse and manufacturing economics, where idle time can undermine the business case for humanoids. Digit 5 is expected to enter early access in the first half of 2027, with broader availability over the following six to nine months. Agility is also set to go public through a merger with Churchill Capital Corp. XI, with the combined company expected to trade on NASDAQ under AGLT.
Boston Dynamics speeds Atlas prototyping
Boston Dynamics says its prototype operations team can currently build an Atlas robot in about one day, with a goal of roughly one hour next year and eventually minutes for the demonstrated process. That does not equal mass-production output, but it shows a push to compress research and build cycles. In-house machining, gear cutting, rapid part revision, and close coordination between engineers and machinists are being used to speed testing, maintenance, and design iteration.
CONCLUSION
The latest robotics push is less about flashy demos than solving industrial constraints such as force, uptime, mobility, and manufacturability. Whether these systems succeed will depend on independent validation and reliable performance outside controlled demonstrations.
Ask a question