CtrlK
| Developer | Techman Robot |
| Type | Wheeled humanoid robot |
| Country of origin | Taiwan |
| Parent company | Quanta Computer |
| Unveiled | August 20, 2025 |
| Status | Prototype (pilot testing H2 2026) |
| Height | ~167 cm (estimated) |
| Degrees of freedom | 22+ |
| Mobility | Wheeled base |
| Compute platform | NVIDIA Jetson Thor |
| AI model | Vision-Language-Action (VLA) |
| End effectors | SCHUNK precision grippers, quick-change system |
| Target industries | Semiconductor, electronics, automotive |
| Website | tm-robot.com |
TM Xplore I (also written as TM Xplore 1) is a wheeled humanoid robot developed by Techman Robot, a Taiwanese collaborative robot manufacturer and subsidiary of Quanta Computer. Unveiled on August 20, 2025, at the Taiwan Automation Intelligence and Robot Show at Taipei's Nangang Exhibition Center, the TM Xplore I is the first humanoid robot produced by a Taiwanese company. The platform combines a humanoid upper body featuring more than 22 articulated joints with a wheeled mobile base, a design choice that prioritizes stability and energy efficiency over bipedal locomotion for factory and warehouse environments.
The robot is built on NVIDIA's Jetson Thor computing module and integrates a Vision-Language-Action (VLA) multimodal model, enabling it to process visual and linguistic inputs for real-time decision-making. Techman leverages its decade of experience in AI-powered cobots with built-in vision systems to position the TM Xplore I as a practical industrial platform rather than a research demonstration. The company plans to begin pilot testing with automotive and semiconductor customers in the second half of 2026, with patent filings and commercialization efforts proceeding in parallel.
Techman Robot Inc. traces its origins to 2012, when Shi-chi Ho, then General Manager of Quanta Storage Inc. (a subsidiary of Quanta Computer), established a robotics laboratory as an internal business division. The lab developed the first collaborative robot with a fully integrated vision system: the TM5. This prototype was unveiled at the International Robot Exhibition (iREX) in Tokyo in 2015, and Techman was formally incorporated as an independent subsidiary that same year.[1][2]
The first commercial TM5 cobots shipped at the end of 2016. By 2018, Techman had risen to become the world's second-largest collaborative robot brand by unit volume, trailing only Denmark's Universal Robots.[2] The company's competitive advantage lies in its built-in AI vision system, which enables functions such as shape matching, barcode and QR code reading, color recognition, and optical character recognition (OCR) directly on the robot arm, without requiring external vision hardware.
Techman is headquartered in Hwa Ya Technology Park in Taoyuan, Taiwan, and employs over 400 staff. The company maintains offices in Shanghai, Shenzhen, Chongqing, Busan, and Alblasserdam (Netherlands), with distribution partners across the United States, Europe, China, Japan, South Korea, and Southeast Asia.[2] Chairman Shi-chi Ho oversees strategic direction, while Haw Chen serves as CEO and Scott Huang as Chief Operating Officer.[3]
Quanta Computer, Techman's parent company, is the world's largest laptop manufacturer and a major contract electronics producer. Founded by Barry Lam in 1988, Quanta manufactures hardware for companies including Apple, Meta, and Tesla. This corporate relationship provides Techman with access to advanced supply chains for semiconductors, motors, sensors, and batteries, as well as deep manufacturing expertise.[4] Quanta Cloud Technology (QCT), another Quanta subsidiary, collaborated with Techman on the TM Xplore I's AI infrastructure at NVIDIA GTC 2026.[5]
In 2018, Techman signed a strategic alliance memorandum of understanding with Omron Corporation, a Japanese industrial automation company. Omron subsequently acquired a 10% equity stake in Techman in October 2021, and Techman's TM series cobots have since been sold globally as co-branded products through Omron's distribution network.[6] Omron accounted for approximately 23% of Techman's revenue in 2025.[7]
Techman is also an official NVIDIA robotics partner, integrating NVIDIA's Jetson computing platforms across its cobot and humanoid product lines.[8]
Techman Robot went public on the Taipei Exchange (TPEx) on September 26, 2025, under the ticker symbol 4585. The listing marked the first IPO within the Quanta Group in over two decades, following Quanta Computer's own listing in 1999. Chairman Barry Lam personally backed the offering.[9]
| Metric | 2023 | 2024 | 2025 |
|---|---|---|---|
| Revenue (NT$) | NT$1.26 billion | NT$1.48 billion | NT$1.82 billion |
| Year-over-year growth | -- | 17.5% | 23% |
| Net profit (NT$) | NT$11.18 million | NT$93.78 million | NT$136.37 million |
| Gross margin | -- | -- | 51% |
| Earnings per share (NT$) | -- | NT$1.04 | NT$1.46 |
Revenue in 2025 reached NT$1.82 billion (approximately US$56.7 million), a 23% increase over 2024. Net profit grew 45% year-over-year to NT$136.37 million, and gross margin held steady at 51%. The company's full-factory automation offerings grew particularly fast, rising from NT$66 million in 2024 to NT$280 million in 2025. In its 2025 results, Techman noted that industry leader Universal Robots experienced a 20% revenue decline and South Korea's Doosan Robotics saw a 40% decline, underscoring Techman's relative outperformance.[7][10]
Techman's 2025 revenue breakdown by customer type was: direct customers and system integrators (42%), distributors (28%), Omron (23%), Quanta Computer (6%), and others (1%). Geographically, China accounted for 29% of revenue, followed by Taiwan (19%), Europe (18%), Southeast Asia (12%), Japan (11%), the United States (7%), and South Korea (4%).[7]
Before entering the humanoid robotics market, Techman Robot built its reputation on the TM Robot series, which the company describes as the world's first collaborative robots with built-in AI vision. The cobot line has evolved through multiple generations.
The original TM Robot series includes the TM5, TM12, and TM14 models, offering payloads from 4 kg to 14 kg and reaches from 700 mm to 1,300 mm. All models integrate a camera and vision system directly into the robot's wrist, eliminating the need for external vision equipment. Six-axis articulation provides versatile motion across applications including pick-and-place, machine tending, assembly, inspection, and palletizing.
The enhanced AI Cobot S Series (introduced circa 2023-2024) includes the TM5S, TM7S, TM12S, TM14S, TM25S, and TM30S models. The S Series features improved joint speeds (up to 450 degrees per second on the sixth axis), higher repeatability (plus or minus 0.03 mm), and a built-in 5-megapixel autofocus camera. Payloads range from 5 kg (TM5S) to 30 kg (TM30S), with the TM30S capable of handling up to 35 kg in palletizing configurations.[11]
| Model | Payload | Reach | Key Applications |
|---|---|---|---|
| TM5S | 5 kg | 700/900 mm | 3D bin picking, assembly, PCB handling, inspection |
| TM7S | 7 kg | 700/900 mm | Assembly, labeling, pick-and-place |
| TM12S | 12 kg | 1,300 mm | Conveyor tracking, machine tending, welding |
| TM14S | 14 kg | 1,100 mm | Palletizing, packaging, polishing |
| TM25S | 25 kg | 1,902 mm | High-volume pick-and-place, machine tending |
| TM30S | 30 kg (35 kg palletizing) | 1,702 mm | Heavy palletizing, packaging, machine tending |
The company also offers the TM Manager software platform for fleet management and the TM Palletizing Operator for automated palletizing workflows. This existing ecosystem of vision-equipped cobots provides the technological foundation for the TM Xplore I humanoid, particularly in AI vision, precision control, and industrial safety compliance.[11]
Techman began the TM Xplore I project in 2024, driven by what CEO Haw Chen described as the global momentum in humanoid robotics, championed in particular by NVIDIA CEO Jensen Huang's advocacy for physical AI.[9] The company recognized that its existing expertise in AI vision, motion control, and collaborative safety could translate directly into a humanoid platform designed for industrial environments. Rather than building a bipedal research robot, Techman elected to develop a wheeled humanoid that could be deployed practically in the near term.
The project began with internal prototyping at Techman's facilities in Taiwan, drawing on Quanta Computer's manufacturing infrastructure and supply chain. Development focused on combining Techman's proprietary "See, Think, Act" AI framework with NVIDIA's latest computing hardware and foundation models.
The TM Xplore I was publicly unveiled on August 20, 2025, at the Taiwan Automation Intelligence and Robot Show at Nangang Exhibition Center in Taipei. The demonstration was covered by the Associated Press and drew attention as Taiwan's first domestically developed humanoid robot.[12] At the event, Vice President William Wang stated that the company planned to launch the robot in the second half of 2026, pending smooth mass production progress. Initial testing would occur at Techman's own manufacturing plants in New Taipei City before customer delivery.[8]
Demonstrations at the show highlighted the robot's quick-change end-effector system, integrated inspection cameras, and SCHUNK precision grippers. The robot performed inspection, sorting, and handling tasks, showcasing its ability to transition between different tools and operational modes without extended downtime.[13]
Techman Robot presented an updated version of the TM Xplore I at NVIDIA's GPU Technology Conference (GTC) 2026 in San Jose on March 18, 2026. The presentation was a joint effort with Quanta Cloud Technology (QCT) and NVIDIA, positioning the robot within a broader Physical AI development pipeline.[5]
At GTC 2026, Techman announced a collaboration with j-mex, a motion capture technology company. The partnership integrates j-mex's Moxi wearable motion capture suit and VR interface with the TM Xplore I, allowing human operators wearing a VR headset and motion capture suit to control the robot through real-time body movements. This teleoperation approach generates high-quality training data by directly capturing human motion, which is then used to improve the robot's autonomous performance through a "sim-to-real" development process. The system uses inertial measurement unit (IMU)-based motion capture to convert human movements into structured training data.[14][15]
At the GTC booth, the TM Xplore I demonstrated its ability to interact with visitors and deliver items to designated locations autonomously.[16]
Scott Huang, Techman's COO, stated at the event: "Future automation will no longer be limited to monotonous, repetitive tasks, but will fully advance toward deep, intelligent collaboration equipped with perception and thinking capabilities."[14]
The TM Xplore I adopts a wheeled humanoid configuration: a human-like upper body with two articulated arms mounted on a mobile wheeled base. This design diverges from the bipedal approach pursued by companies such as Tesla (with Optimus), Figure AI (with Figure 02), and Unitree Robotics (with the H1). Techman's rationale centers on practical considerations for industrial deployment: wheeled platforms provide superior stability during precision manipulation tasks, better energy efficiency, and more reliable navigation across flat factory floors. The wheeled base eliminates the balancing challenges inherent to bipedal locomotion, allowing the robot to focus its computational and mechanical resources on dexterous upper-body manipulation.[8][12]
The robot's overall form resembles a sleek white humanoid torso, with a head module containing advanced sensor arrays and vision systems. Its aesthetic balances industrial functionality with a recognizable humanoid silhouette designed for collaborative environments where workers interact directly with the robot.[17]
The upper body features more than 22 articulated joints, enabling fluid, human-like manipulation across complex tasks. The dual-arm configuration allows coordinated bimanual operations such as grasping, sorting, assembly, and inspection. The arms inherit precision control technology from Techman's cobot line, refined over a decade of development.[12][13]
The robot integrates a quick-change end-effector system that allows operators to swap between different tool heads, including SCHUNK precision grippers, inspection cameras, and other attachments, without extended downtime. This modularity makes the platform adaptable to different tasks within the same shift, a requirement for high-mix, low-volume production environments common in semiconductor and electronics manufacturing.[13]
The TM Xplore I carries Techman's proprietary TM AI Vision cameras for environment and object detection, supplemented by proximity sensors, tactile sensors for safe human-robot interaction, and motion sensors for accurate arm and base control. The vision system builds on the built-in camera technology that distinguishes Techman's cobot line, enabling capabilities such as object recognition, barcode reading, and visual inspection without external hardware.[18]
The TM Xplore I is powered by the NVIDIA Jetson Thor module, a computing platform designed for high-performance edge AI in robotics applications. Jetson Thor provides the processing capacity needed for real-time multimodal sensor fusion, generative AI inference, and autonomous navigation. The platform enables the robot to run complex AI models on-device without requiring a cloud connection for time-critical operations.[5][16]
QCT provides supporting infrastructure through its QCT Dev. Kit for Physical AI, which integrates optimized server hardware (specifically the QuantaGrid D75E-4U, an NVIDIA RTX PRO Blackwell GPU server based on NVIDIA MGX architecture) with NVIDIA frameworks and robot foundation models. This full-stack approach streamlines the data generation, model training, and real-world deployment pipeline.[5]
Techman describes the TM Xplore I's cognitive architecture through its core "See, Think, Act" technology framework:
This framework extends Techman's existing cobot AI capabilities into a more general-purpose humanoid context, where the robot must handle a wider variety of tasks and environments than a fixed-position cobot arm.[5][14]
The TM Xplore I leverages several components of NVIDIA's robotics software ecosystem:
| NVIDIA Technology | Function |
|---|---|
| Isaac GR00T | Robot foundation model for improved dexterity and autonomous manipulation |
| Isaac Sim | Simulation and testing of robot behavior in realistic industrial scenarios |
| FoundationStereo | Deep learning-based depth perception and navigation accuracy |
| Cosmos | Open world foundation models for synthetic data generation |
| Jetson Thor | Edge AI inference module for onboard computing |
The integration of Isaac GR00T is particularly significant, as it provides pre-trained manipulation skills that Techman can fine-tune for specific industrial applications. Isaac Sim allows developers to test and validate robot behavior in digital twin environments before physical deployment, reducing development time and risk. FoundationStereo enhances the robot's spatial awareness for navigation in cluttered factory environments.[5][16]
The VLA multimodal model is a central element of the TM Xplore I's intelligence. By combining visual perception, natural language understanding, and action generation within a single model, the robot can interpret high-level instructions (spoken or text-based), assess the visual scene, and generate appropriate motor actions. This capability allows the robot to adapt to new tasks without extensive reprogramming, learning instead from human demonstrations, simulated environments, and synthetic data.[5][12]
The VLA approach represents a broader industry trend toward foundation models for robotics, with similar architectures being adopted by Figure AI (Helix), Google DeepMind (RT-2), and others.
The partnership with j-mex provides the TM Xplore I with a human-in-the-loop training methodology. Operators wearing j-mex's Moxi motion capture suit and a VR headset can perform tasks (such as dual-arm grasping, object sorting, and precision placement) while the robot mirrors their movements in real time. The system captures detailed motion data using IMU-based tracking, which is then processed into training datasets for the robot's AI models.[14][15]
This approach offers two advantages. First, it generates high-quality training data that reflects real human manipulation strategies, which can be more effective than purely scripted or random exploration approaches. Second, it enables a "sim-to-real" transfer pipeline: robots can first be trained in controlled simulated environments using NVIDIA Isaac Sim, then refined with real-world data captured through teleoperation, progressively improving performance in dynamic physical settings.[14]
Techman has not released a comprehensive specification sheet for the TM Xplore I, which remains in the prototype and pilot-testing phase. The following specifications are compiled from official announcements, press coverage, and database listings.
| Category | Specification | Details |
|---|---|---|
| Physical | Form factor | Wheeled humanoid (upper body + mobile base) |
| Physical | Estimated height | ~167 cm |
| Physical | Design | White humanoid torso with sensor head module |
| Manipulation | Degrees of freedom | 22+ articulated joints |
| Manipulation | End effectors | SCHUNK precision grippers, quick-change system |
| Manipulation | Fingers per hand | 5 |
| Mobility | Locomotion type | Wheeled base |
| Mobility | Estimated max speed | 0.83 m/s (3.0 km/h) |
| Computing | Primary compute | NVIDIA Jetson Thor |
| Computing | AI model | Vision-Language-Action (VLA) multimodal |
| Computing | Software | Linux-based, closed source |
| Sensors | Vision | TM AI Vision cameras (object detection, inspection) |
| Sensors | Additional | Proximity, tactile, and motion sensors |
| Connectivity | Wireless | Wi-Fi, Bluetooth |
| Connectivity | Network | Factory/facility network integration |
| Power | Estimated battery life | ~2 hours |
Note: Several specifications (exact weight, payload capacity, operating temperature range, and ingress protection rating) have not been officially disclosed as of early 2026. The height and speed estimates are drawn from third-party aggregator databases and have not been confirmed by Techman.[18]
The TM Xplore I is designed for industrial environments where flexibility, precision, and human-robot collaboration are required. Techman has identified several primary application domains.
Semiconductor fabrication requires extreme precision and contamination control. The TM Xplore I's integrated vision system and dexterous manipulation are aimed at tasks such as wafer handling, component inspection, and equipment loading and unloading. Chipmakers accounted for approximately 25% of Techman's H1 2025 cobot sales, and the company views the humanoid as an extension of these existing customer relationships.[8]
High-mix, high-precision electronics production lines demand robots that can adapt to frequent product changeovers. The quick-change end-effector system and VLA-based learning allow the TM Xplore I to switch between tasks (inspection, assembly, sorting) without manual reprogramming, making it suitable for small-batch and dynamic production environments.[13]
Automotive factories require robots capable of handling diverse parts, performing quality inspections, and navigating between workstations. The wheeled base provides mobility across the factory floor, while the articulated upper body enables manipulation tasks at different stations. Vice President William Wang noted in March 2026 that the TM Xplore I would begin pilot testing with automotive customers in H2 2026.[7]
The robot's wheeled mobility and dual-arm manipulation make it suitable for warehouse environments, where it can pick, sort, and transport items between locations. The autonomous navigation capabilities powered by NVIDIA's FoundationStereo and Isaac GR00T enable the robot to operate in semi-structured environments with dynamic obstacles.[4][16]
The TM Xplore I's integrated inspection cameras and precision grippers position it for metrology applications, including dimensional measurement, surface inspection, and calibration tasks. The robot can physically access difficult-to-reach inspection points and transition between measurement equipment and automated optical inspection zones, serving as a mobile link in quality control workflows.[13]
| Date | Milestone |
|---|---|
| 2024 | TM Xplore I project initiated |
| August 20, 2025 | Prototype unveiled at Taiwan Automation Intelligence and Robot Show |
| September 26, 2025 | Techman Robot IPO on Taipei Exchange (TPEx: 4585) |
| March 18, 2026 | Updated prototype and j-mex training system shown at NVIDIA GTC 2026 |
| 2026 | Patent filings and commercialization preparations |
| H2 2026 | Pilot testing with automotive and semiconductor customers |
| 2027 (earliest) | Potential bipedal variant commercialization |
Techman has indicated that wheeled humanoid platforms are its near-term priority due to their proven stability and readiness for industrial deployment. A bipedal version is under consideration but is not expected to reach commercialization before 2027 at the earliest. The company plans to offer configurable solutions, with upper-body-only or full-body configurations varying by customer requirements.[7]
The TM Xplore I's debut places it within a growing Taiwanese robotics ecosystem that has attracted attention from global investors and industry analysts. Taiwan's position as a leading manufacturer of semiconductors, electronics, and precision machinery provides a natural foundation for robotics innovation.
Several major Taiwanese companies have entered or expanded their robotics efforts:
| Company | Robotics Activity |
|---|---|
| Quanta Computer / Techman Robot | Cobots (TM series), TM Xplore I humanoid |
| Foxconn (Hon Hai) | Nurabot healthcare humanoid (with Kawasaki), factory automation |
| HIWIN Technologies | Precision components (ball screws, harmonic reducers, bearings) for humanoid supply chain |
| Hota Industrial | Precision gears and drivetrain components for robots |
| TSMC | Semiconductor fabrication supporting AI chips for robotics |
Four Taiwanese firms (TSMC, Foxconn, HIWIN, and Hota) were included in Morgan Stanley's "Humanoid 100" list of companies expected to benefit from advances in humanoid robotics.[19] HIWIN's robotics business is projected to exceed 10% of group revenue in 2026, driven by demand for critical components such as ball screws, harmonic reducers, and crossed roller bearings used in humanoid platforms.[19]
Taiwan is ranked among the world's top three manufacturers and exporters of rectangular single-axis robots, service cleaning robots, and collaborative robots.[20] NVIDIA CEO Jensen Huang has identified Taiwan as a potential hub for humanoid robot assembly, although supply chain gaps in areas such as advanced actuators remain a challenge.[21]
The TM Xplore I enters a rapidly expanding humanoid robotics market. Its wheeled design positions it differently from most competitors, which pursue bipedal locomotion.
| Company | Robot | Form Factor | Status (Early 2026) |
|---|---|---|---|
| Techman Robot | TM Xplore I | Wheeled humanoid | Prototype; pilot testing H2 2026 |
| Figure AI | Figure 02 / Figure 03 | Bipedal | Commercial (limited deployment) |
| Tesla | Optimus | Bipedal | Prototyping |
| Agility Robotics | Digit | Bipedal | Commercial pilots (Amazon) |
| Unitree Robotics | H1 / G1 | Bipedal | Commercial |
| Dobot | Atom | Wheeled humanoid | Mass production begun |
| Boston Dynamics | Atlas (electric) | Bipedal | Commercial launch 2026 |
| Apptronik | Apollo | Bipedal | Pilot deployments |
Techman's most direct competitor may be Chinese manufacturer Dobot (Shenzhen Yuejiang Technology), which began mass production of its Atom wheeled humanoid in 2025. Both companies share a collaborative robot heritage and have opted for wheeled bases to prioritize industrial reliability over bipedal agility.[4]
Techman's key competitive differentiators include its decade-long track record in vision-equipped cobots, the established Omron global distribution channel, Quanta Computer's manufacturing scale and supply chain, and deep integration with NVIDIA's Physical AI ecosystem. The company's publicly listed status (TPEx: 4585) also provides financial transparency uncommon among humanoid robotics startups.
However, Techman faces challenges. Its humanoid is still in the prototype stage while competitors such as Figure AI and Agility Robotics have already achieved commercial deployments. The wheeled design, while practical for factories, limits the robot's versatility compared to bipedal platforms that can navigate stairs, uneven terrain, and environments designed for human bipedal movement.
Techman Robot has outlined an ambitious roadmap for its humanoid program. President Haw Chen has emphasized that the company's AI vision integration capabilities, honed through years of cobot development, provide a differentiation that extends naturally into humanoid robotics.[7] The company projects that Asia-Pacific markets, particularly China's semiconductor equipment demand and the integration of cobots with autonomous mobile robots in wafer fabrication facilities, will drive near-term sales growth.[7]
The TM Xplore I represents the first step in what Techman envisions as a broader humanoid product line. Wheeled platforms are the initial focus for their reliability and industrial readiness, but the company has acknowledged plans for bipedal variants as the technology matures. Configurations will vary by customer needs, ranging from upper-body-only solutions to full-body humanoid platforms.[7]
With its IPO completed, a strong cobot business generating over NT$1.8 billion in annual revenue, and partnerships with NVIDIA, Omron, and QCT, Techman is positioned as a credible entrant in the industrial humanoid market. Whether the company can translate its cobot expertise into humanoid commercial success will depend on the pilot deployments planned for the second half of 2026.