Dexmate — bimanual mobile manipulators and developer-friendly humanoids

• 2024
  Founded
• Santa Clara, CA
  Headquarters
• Vega
  Latest flagship (2026)
• ~$41M
  Funding raised
• USA
  Country
• Private
  Status

About Dexmate

Dexmate Inc. was founded in 2024 by a research-heavy team of PhDs drawn from MIT, UC San Diego, and Carnegie Mellon University. Tao Chen — an MIT EECS PhD from the Computer Science and Artificial Intelligence Laboratory (CSAIL) and a Conference on Robot Learning best-paper award recipient for his work on dexterous manipulation — serves as chief executive officer. Yuzhe Qin, a UC San Diego PhD focused on robot simulation and learning, is co-founder and chief technology officer, and Chongyang (Max) Wang, an MIT-trained operator with roughly a decade of industry experience, rounds out the founding team as chief operating officer. The company set out from day one to attack general-purpose robotics from a specific angle: high-DoF bimanual mobile manipulation on a wheeled base, rather than bipedal locomotion.

That strategic choice — wheels instead of legs, and dexterous dual arms as the centerpiece — is the core of Dexmate's thesis. Bipedal humanoids remain among the most expensive and least reliable classes of robot to build at scale, and the majority of practical work in warehouses, retail back rooms, and light-industrial cells does not require walking. By focusing on an omni-directional wheeled base with a foldable torso that reaches from 0.66 m folded up to 2.2 m extended, Dexmate can deliver a robot with genuinely useful vertical workspace and long shift runtime for a small fraction of the price of a comparable bipedal platform. The company shipped its first hardware prototype in roughly six months from founding — an unusually fast concept-to-product cycle for a full-size humanoid-class robot.

Funding has kept pace with execution. According to public filings and third-party data trackers, Dexmate has raised approximately $41 million in venture funding across a 2024 seed round and a 2026 strategic round, with named investors including LG Technology Ventures (the corporate venture arm of South Korea's LG Group), TSVC, Epsilon Ventures, Mana Ventures, Jinqiu Capital, and RoboStrategy. In March 2026, LG CNS — the IT-services arm of LG — announced a strategic stake in Dexmate as part of its "Robot Transformation" (RX) business, with the explicit goal of integrating Vega hardware into industrial workplace deployments. Dexmate is also an NVIDIA Inception company, and Vega received a Red Dot Design Award in 2025.

Dexmate is part of a specific competitive cohort: US and North America-anchored robotics startups spun out of elite robot-learning labs, targeting the manufacturing, logistics, and research markets. Its closest peers on strategy include Astribot (another bimanual, non-bipedal platform), Physical Intelligence (foundation-model-first robot AI), and Sanctuary AI (general-purpose bimanual manipulation with an emphasis on cognitive architecture). It also indirectly competes with the bipedal humanoid wave — Figure AI, 1X Technologies, Apptronik, Agility Robotics, and Boston Dynamics — where Dexmate's differentiator is price, developer accessibility, and the practical decision to skip legs.

Technology stack

Dexmate's platform is a vertically integrated combination of custom mechatronics, a rich sensor suite, NVIDIA edge compute, and a data-driven learning stack that trains policies in simulation and from human demonstration before deployment. Vega is explicitly positioned as a developer-friendly research platform: it ships with a Python API installable via pip, native ROS integration, and first-class support for VR/exoskeleton teleoperation using consumer headsets such as Apple Vision Pro and Meta Quest.

Perception

Vega carries a rich sensor stack: stereo RGB and RGBD cameras in the head, additional RGB cameras on the body, LiDAR for base navigation, ultrasonic proximity sensors, IMUs for body-state estimation, and 6-axis force/torque sensors for contact-rich manipulation. That combination — long-range LiDAR, stereo depth for close-range grasping, and tactile/force sensing at the wrists — is what lets Vega do bimanual, contact-rich tasks in unstructured warehouse and light-industrial environments without environment modification.

Data-driven control and teleoperation

Rather than betting a single monolithic foundation model, Dexmate publicly emphasizes a data-flywheel approach: policies are trained in simulation, refined with human teleoperation demonstrations, and then deployed on-robot. The company sponsored the $200,000 Whole-Body Control & Dexterity Challenge at ICRA 2025 and has released an open-source teleoperation device to accelerate third-party data collection. The Python-first API and open teleoperation stack are deliberate — they make Vega usable as a shared research substrate the way industrial arms once were.

Actuators, power, and compute

Vega uses electric brushless (BLDC) actuators throughout — no hydraulics — and its onboard compute runs on NVIDIA Jetson-class edge AI (AGX Orin on early units, with AGX Thor on newer configurations). The high-capacity battery pack targets 10+ hours of continuous bimanual operation and up to 20–30 hours in light-load mode, with a peripheral interface panel that exposes four USB 3.2 ports, one gigabit Ethernet, one DisplayPort, and 5 V / 12 V accessory power for integrators bolting on their own sensors and compute.

Power management is a particularly demanding subsystem for a foldable bimanual platform of this class. Each joint's servo controller must deliver hundreds of watts of instantaneous torque without thermally throttling — driving the use of high-efficiency wide-bandgap semiconductors (SiC and GaN) in the motor-driver stages and requiring tightly matched precision passives on every gate-drive and current-sense channel. The onboard battery pack integrates cell balancing, thermal management, and multi-stage protection ICs, and the arm and torso rails rely on precision MLCC capacitors and low-tolerance resistors to keep 36+ actuators, edge compute, LiDAR, ultrasonics, and the sensor stack sharing a single power tree without noise or rail sag.

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 Vega-class bimanual mobile manipulator depends on thousands of individual electronic components: precision BLDC motor drivers and high-torque servo modules across 36+ joints, IMUs and 6-DoF force-torque sensors, LiDAR and time-of-flight modules, stereo RGB and RGBD camera assemblies, ultrasonic proximity sensors, GPU compute boards (NVIDIA Jetson-class edge AI), battery-management ICs, wide-bandgap SiC/GaN power semiconductors, high-speed board-to-board connectors, precision 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, this pressure is amplified: dozens of well-funded teams 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. Dexmate®, Vega, Vega U, and Dexgripper are trademarks of Dexmate, Inc. Hybrid Electronics is an independent electronic-components sourcing company and is not affiliated with, endorsed by, or authorized by Dexmate, Inc. All information on this page is compiled from public sources for educational and reference purposes.