RobotEra — Tsinghua-incubated humanoid robots from Beijing

• Aug 2023
  Founded
• Beijing, China
  Headquarters
• L7 (2025)
  Latest flagship
• $200M+
  Funding raised
• China
  Country
• Private
  Status

About RobotEra

Beijing Xingdong Jiyuan Technology Co., Ltd. — better known internationally as RobotEra and by its Chinese name 星动纪元 (Xingdong Jiyuan) — was founded on August 4, 2023 and incubated by the Institute for Interdisciplinary Information Sciences (IIIS) at Tsinghua University. Its founder, Chen Jianyu, is a doctoral supervisor and assistant professor at Tsinghua IIIS whose research focus spans bipedal locomotion, whole-body humanoid control, and reinforcement-learning-based policies for dynamic motion. RobotEra is widely described as the only humanoid-robotics enterprise in which Tsinghua University itself holds an equity stake, an unusual arrangement that reflects the depth of its research lineage.

The company's mission is to build a general-purpose humanoid robot that can operate in real, unstructured environments — factories, warehouses, retail floors, and eventually homes — powered by an end-to-end embodied-AI model rather than hand-coded task scripts. Within roughly a year of founding, RobotEra had unveiled its first full-size platform, the XBot-L, a 185 cm, 80 kg bipedal humanoid with 61 degrees of freedom and proprietary quasi-direct-drive actuators. That platform paved the way for the record-setting STAR1, which in October 2024 ran at up to 3.6 m/s (roughly 12.8 km/h or 8 mph) across the Gobi Desert during a widely reported 34-minute outdoor test — the fastest sustained running speed ever demonstrated by a full-size humanoid robot at the time.

RobotEra has raised more than $200 million across a rapid succession of rounds. A pre-Series A of nearly RMB 300 million closed in October 2024. A Series A of approximately RMB 500 million (about $69–70 million) followed in July 2025, co-led by CDH Investments and Haier Capital. In November 2025 the company closed a Series A+ / strategic round of nearly RMB 1 billion — approximately $140–141 million — led by Geely Capital, with participation from BAIC Capital, Alibaba, Lenovo, and the Beijing Artificial Intelligence Industry Investment Fund. Industry press has widely reported a unicorn valuation of around $1.4 billion following the 2025 rounds, and RobotEra has since claimed more than RMB 500 million (~$70 million) in commercial orders.

The team, weighted heavily toward hardware, controls, and AI engineers drawn from Tsinghua and China's broader robotics ecosystem, has grown to roughly 200 employees. RobotEra's long-term ambition is a general-purpose humanoid capable of handling the majority of physically demanding, high-turnover jobs across manufacturing, logistics, and service — a vision that places it directly alongside China's other well-funded humanoid contenders, including Unitree, XPeng Robotics, Agibot (智元), Fourier Intelligence, UBTech, and LimX Dynamics, as well as Western players such as Figure AI, Tesla Optimus, 1X, Apptronik, Agility, and Boston Dynamics.

Technology stack

RobotEra's technical thesis, inherited directly from its Tsinghua IIIS research lineage, is that a competitive general-purpose humanoid must vertically integrate mechatronics, perception, and an end-to-end embodied-AI foundation model. The company designs and manufactures more than 95% of its own hardware — joints, motors, sensors, control boards — and trains its own VLA model on data collected from its fleet.

Perception

RobotEra platforms combine multi-camera vision, LiDAR, and time-of-flight (ToF) modules with force-torque sensing at the wrists and high-resolution 3D tactile sensors in the XHAND1 fingertips. This fusion of vision, depth, and touch is what allows L7 and STAR1 to perform contact-rich manipulation — bin picking, deformable-object handling, and chopstick-level fine motor tasks — in unstructured factory, warehouse, and service environments.

ERA-42 Vision-Language-Action model

ERA-42 is RobotEra's in-house Vision-Language-Action architecture. It maps natural-language instructions and multi-modal sensor input directly to whole-body actuator commands, replacing traditional code-based task programming. RobotEra has also publicized Ctrl-World, a generative world model developed with Stanford University collaborators that lets robots "rehearse" tasks in an imagination space before executing them physically — a technique that materially improves sample efficiency during policy training.

Actuators, power, and compute

L7 and STAR1 use RobotEra's self-developed brushless DC (BLDC) joint motors — quasi-direct-drive on the earlier XBot-L, and fully self-actuated with peak joint torques up to 400 Nm at joint speeds up to 25 rad/s on STAR1 and L7. There are no hydraulics anywhere in the stack, which simplifies serviceability and improves energy efficiency for extended outdoor runs like the Gobi Desert test. Onboard compute is understood to include GPU-class acceleration to run the ERA-42 model locally, so the robot does not depend on cloud connectivity for real-time perception and control on the factory floor.

Power management is one of the hardest subsystems in a humanoid of this class. Each joint's servo controller has to deliver hundreds of watts of instantaneous torque without thermally throttling, which drives the use of high-efficiency wide-bandgap semiconductors (SiC and GaN) in the motor-driver stages, tightly matched precision MLCC capacitors and thin-film resistors on every gate-drive and current-sense channel, and dedicated battery-management ICs for cell balancing, thermal control, and multi-stage protection on the pack. On a 55-DoF platform like L7, those subsystems collectively define how long the robot can actually run between charges — and how quickly its supply chain can respond when any one of them goes onto allocation.

The robotics supply chain — where Hybrid Electronics fits in

A modern humanoid robot is one of the most component-dense products ever built. A single RobotEra-class L7 or STAR1 platform depends on thousands of individual electronic components: precision BLDC motor drivers, high-torque quasi-direct-drive servo modules, IMUs and 6-DoF force-torque sensors, LiDAR and time-of-flight modules, high-resolution 3D tactile sensors, GPU compute boards, battery-management ICs, wide-bandgap SiC/GaN power semiconductors, high-speed board-to-board connectors, precision thin-film resistors, and thousands of MLCC capacitors distributed across every power and signal rail.

Building that kind of hardware puts enormous pressure on the electronic-components supply chain. Prototype pilots and low-volume production runs frequently hit allocation, EOL notices, and long lead times on the exact parts they need, and R&D and field-service teams routinely have to support older platforms that use semiconductors no longer in mainstream distribution. During the current humanoid robotics and general-purpose robot AI boom, that pressure is amplified: dozens of well-funded teams — including RobotEra, Unitree, XPeng Robotics, Agibot, Fourier Intelligence, and their Western peers — are chasing the same short list of high-torque BLDC drivers, precision IMUs, GPU compute modules, and automotive-grade power semiconductors, and franchise stock on many key parts is now measured in months of lead time rather than weeks.

Hybrid Electronics has been sourcing electronic components for industrial and embedded systems since 1995. For robotics teams — humanoid, mobile, industrial, and research — we specialize in finding the parts you need when mainstream distribution can't:

Trademark and affiliation notice. RobotEra®, 星动纪元, XBot-L, STAR1, L7, Q5, XHAND1, and ERA-42 are trademarks of Beijing Xingdong Jiyuan Technology Co., Ltd. Hybrid Electronics is an independent electronic-components sourcing company and is not affiliated with, endorsed by, or authorized by Beijing Xingdong Jiyuan Technology Co., Ltd. All information on this page is compiled from public sources for educational and reference purposes.