1X EVE
Last reviewed
Apr 12, 2026
Sources
29 citations
Review status
Source-backed
Revision
v3 · 4,877 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Apr 12, 2026
Sources
29 citations
Review status
Source-backed
Revision
v3 · 4,877 words
Add missing citations, update stale details, or suggest a clearer explanation.
| 1X EVE | |
|---|---|
| General information | |
| Manufacturer | 1X Technologies |
| Country of origin | Norway |
| Year introduced | 2018 (prototype); 2022 (commercial deployment) |
| Status | In production (industrial) |
| Type | Wheeled humanoid android |
| Height | 188 cm (6 ft 2 in) |
| Weight | 87 kg (192 lb) |
| Degrees of freedom | 25 |
| Max speed | 4 m/s (14.4 km/h / 9 mph) |
| Payload capacity | 15 kg (33 lb) |
| Battery | 1.05 kWh lithium-ion |
| Runtime | 4 to 6 hours |
| Actuation | Revo1 quasi direct-drive |
| Compute | Intel i7 (control) + NVIDIA Jetson Xavier (AI) |
| Website | 1x.tech/eve |
1X EVE is a wheeled humanoid robot developed by 1X Technologies (formerly Halodi Robotics), a Norwegian-American robotics company founded in 2014. EVE was the company's first commercial android platform, designed for security, logistics, and healthcare applications. Standing 188 cm tall and weighing 87 kg, EVE uses a two-wheeled self-balancing base rather than bipedal legs, allowing it to travel at speeds up to 9 mph while carrying loads of up to 15 kg. The robot was first publicly demonstrated in July 2017 and entered commercial deployment in 2022.
EVE is notable for being one of the first humanoid robots to achieve large-scale commercial deployment. In March 2022, ADT Commercial (now Everon) placed what was described as the world's largest single order for humanoid robots, contracting 140 EVE units for autonomous security patrols in commercial buildings across the United States.[1] Beyond security, EVE has been deployed in logistics, retail, and healthcare pilot programs across Europe and the United States.
The robot's real-world operational experience provided 1X Technologies with critical data and engineering lessons that directly informed the development of 1X NEO, the company's bipedal humanoid designed for home environments. EVE continues to serve in industrial and enterprise roles where its wheeled design offers practical advantages over legged locomotion.
1X Technologies was founded in May 2014 in Moss, Norway, by roboticist Bernt Oivind Bornich alongside co-founders Phuong Nguyen, Jorgen Sundell, and Pal Loken.[2] Originally named Halodi Robotics, the company set out to develop safe, general-purpose robots capable of operating alongside humans in shared environments. Bornich, who studied Robotics and Nanoelectronics at the University of Oslo, had been fascinated by electromechanics since childhood and believed that creating truly useful robots required a fundamentally different approach to actuation and control.[3]
From its inception, the company rejected the conventional approach of using high-ratio harmonic gear drives, which were standard in industrial robotics at the time. These traditional actuators introduced significant friction and reflected inertia, making them poorly suited for robots that needed to interact safely with humans. Instead, the Halodi team pursued a novel design philosophy centered on low-gear-ratio transmissions and direct-drive motors, drawing inspiration from the way biological tendons transmit force in the human body.
The company's foundational technology breakthrough came with the development of the Revo1 motor, a high-torque brushless DC (BLDC) servo motor paired with a cable-driven differential transmission. First demonstrated publicly at ICRA 2018 (the IEEE International Conference on Robotics and Automation), the Revo1 was described by the company as the world's highest torque-to-weight direct drive servo motor at the time of its development.[4]
The Revo1's design was inspired by human tendon mechanics. Rather than coupling motors directly to joints through rigid gearboxes, the system uses low-friction cable transmissions to deliver force, similar to how tendons connect muscles to bones. This approach yields several engineering advantages:
This motor technology became the hardware foundation for EVE and established 1X's competitive differentiation in the humanoid robotics market. The quasi direct-drive approach would later evolve into the Tendon Drive system used in the NEO Beta and NEO Gamma.
Development of EVE began shortly after the company's founding in 2014, with the engineering team iterating on motor designs, control architectures, and mechanical configurations. The first functional prototype was publicly demonstrated in July 2017, performing simple warehouse and kitchen tasks to showcase the robot's manipulation and navigation capabilities.[5] By January 2019, an improved version was shown autonomously navigating warehouse environments on two wheels while sorting packages, validating the wheeled humanoid concept for real-world logistics applications.
The decision to build EVE as a wheeled robot rather than a bipedal walker was a deliberate strategic choice. By mounting a humanoid torso on a self-balancing wheeled base, 1X bypassed the enormous challenge of stable and efficient bipedal locomotion, which remained an unsolved problem at the time for commercial applications. Wheels offered greater reliability, faster movement, longer battery life, and simpler control. This pragmatic design allowed the company to begin deploying robots in real customer environments years before competitors with legged designs, turning commercial facilities into data-collection platforms for training the company's artificial intelligence systems.[6]
EVE has a humanoid upper body mounted on a self-balancing two-wheeled base, with a small caster wheel at the rear for additional stability. The robot's overall form factor is roughly human-sized, standing approximately 188 cm (6 ft 2 in) tall and weighing 87 kg (192 lb) including the battery. The wheeled base allows EVE to navigate smoothly on flat indoor surfaces, take elevators, traverse ramps, and maneuver around corners in commercial and industrial environments.
The robot's exterior uses soft, organically inspired mechanics designed to minimize injury risk during physical contact with humans. All pinch points in the joints are eliminated or enclosed, and the outer surfaces incorporate padding to absorb incidental contact forces. The head unit, designed by Norwegian industrial design firm Eker Design, features dual high-resolution AMOLED displays that render animated facial expressions for intuitive face-to-face communication with human operators and bystanders.[7]
EVE has 25 degrees of freedom (DOF) distributed across its body, enabling a wide range of manipulation and mobility tasks:
| Body region | Degrees of freedom | Description |
|---|---|---|
| Arms | 7 DOF per arm (14 total) | Full reach and dexterity for manipulation tasks |
| Wheeled leg assembly | 6 DOF | Self-balancing locomotion and terrain adaptation |
| Neck | 1 DOF | Head pan/tilt for gaze tracking and perception |
| Hands | 1 DOF per hand (2 total) | Gripper-style end effectors for grasping |
| Wheels | 1 DOF per wheel (2 total) | Independent drive for differential steering |
The 7-DOF arms provide the kinematic reach needed for tasks such as opening doors, retrieving objects from shelves at various heights, handling packages, and squatting down to access floor-level storage. Each arm can handle a payload of up to 8 kg independently, with a combined lift capacity of approximately 15 kg.[8]
All of EVE's joints are powered by the proprietary Revo1 quasi direct-drive actuators. These consist of high-torque-density brushless DC motors paired with low-gear-ratio cable-driven transmissions. The quasi direct-drive approach places EVE in a distinct category from robots that use either fully direct-drive motors (which sacrifice torque for simplicity) or high-ratio harmonic drives (which sacrifice backdrivability for torque).
Key actuation specifications include:
| Parameter | Value |
|---|---|
| Motor type | Brushless DC (BLDC) with cable-driven transmission |
| Gear ratio | Low ratio (quasi direct-drive) |
| Backdrivability | 95% |
| Torque accuracy | 2% |
| Backlash | Near zero |
| Friction | Extremely low |
The 95% backdrivability rating is a critical safety feature. If a human or object collides with EVE's arm, the arm gives way rather than rigidly resisting, dramatically reducing impact forces. The 2% torque accuracy allows EVE to perform delicate manipulation tasks that require precise force control, such as handling fragile objects or operating door handles.[9]
EVE's perception system is built around a multi-camera array developed in partnership with Immervision, a Canadian optics and vision technology company. The system uses three 187-degree panomorph lenses arranged to provide:
| Sensor | Details |
|---|---|
| Front cameras | 2x high-resolution HDR with panomorph lenses |
| Rear camera | 1x HDR with panomorph lens |
| Depth sensing | 3D depth camera in the head unit |
| Audio | Stereo microphones + speaker system |
| Inertial measurement | IMU for balance and orientation |
| Field of view | 360-degree panoramic (combined) |
The high-resolution HDR cameras provide robust perception across varying lighting conditions, which is particularly important for security patrol applications where the robot operates in dimly lit buildings at night. The depth-sensing 3D camera in the head enables precise distance estimation for obstacle avoidance and object manipulation. The perception stack supports simultaneous localization and mapping (SLAM) for autonomous navigation, enabling EVE to build and maintain spatial maps of its operating environment.[10]
EVE uses a dual-processor computing architecture that separates real-time motor control from higher-level AI inference:
| Component | Processor | Function |
|---|---|---|
| Real-time control | Intel Core i7 | Joint control, balance, safety monitoring |
| AI inference | NVIDIA Jetson Xavier | Computer vision, neural networks, path planning |
The Intel i7 handles the deterministic, low-latency tasks required for stable balancing and responsive motor control. The NVIDIA Jetson Xavier module runs the higher-level perception and planning algorithms, including the vision-based neural network that governs autonomous behavior. The software stack is built on a custom Linux-based operating system and incorporates ROS 2 (Robot Operating System 2), OpenCV, and a mix of Python, C++, and Java applications.[11]
EVE is powered by a 1.05 kWh lithium-ion battery pack that provides between 4 and 6 hours of continuous operation depending on workload intensity. Security patrol operations, which involve relatively steady movement with occasional stops, tend toward the longer end of the runtime range. More physically demanding tasks involving frequent arm manipulation and heavy lifting reduce the runtime. The battery can be recharged in approximately one hour, and EVE is capable of autonomously navigating to its charging station and self-docking when battery levels drop below a threshold.[12]
| Specification | Value | |---|---|---| | Height | 188 cm (6 ft 2 in) | | Weight | 87 kg (192 lb) | | Degrees of freedom | 25 | | Maximum speed | 4 m/s (14.4 km/h / 9 mph) | | Payload capacity | 15 kg (33 lb) | | Per-arm payload | 8 kg (17.6 lb) | | Battery capacity | 1.05 kWh (lithium-ion) | | Runtime | 4 to 6 hours | | Recharge time | ~1 hour | | Actuation | Revo1 quasi direct-drive | | Backdrivability | 95% | | Control processor | Intel Core i7 | | AI processor | NVIDIA Jetson Xavier | | Vision system | 3x HDR cameras (360-degree coverage) | | Safety rating | HIC < 250 | | Operating system | Custom Linux (ROS 2, OpenCV) |
EVE's autonomous behavior is governed by a single vision-based neural network that operates at 10 Hz, processing camera feeds and generating control commands for driving, arm movements, gripper operation, torso adjustments, and head positioning. This end-to-end approach means that a single learned model handles all aspects of the robot's physical behavior, rather than relying on separate hand-engineered modules for navigation, manipulation, and obstacle avoidance.[13]
In a widely publicized demonstration released in late 2023, 1X showcased 30 EVE robots performing autonomous warehouse and facility tasks. The company stated that every behavior shown in the video was controlled entirely by neural networks, with no teleoperation, scripted trajectories, video edits, or speed alterations. The robots demonstrated tasks including picking items from various heights, placing objects into containers, opening doors independently, navigating to charging stations, and self-charging through a squatting docking maneuver.[14]
1X developed EVE's autonomous capabilities through an iterative machine learning process that combines imitation learning, teleoperation data, and progressive fine-tuning:
This hierarchical training approach allows 1X to deploy EVE in new environments relatively quickly while maintaining robust general-purpose capabilities. The company described this system as "a data engine for solving general-purpose mobile manipulation tasks in a completely end-to-end manner."[16]
EVE supports a shared autonomy operating model in which the robot handles routine tasks autonomously while human operators can intervene remotely for complex or unforeseen situations. An operator wearing a VR headset can "teleport" into any EVE unit, seeing through its cameras and controlling its body in real time. This telepresence capability serves dual purposes: it provides a fallback for situations beyond the robot's autonomous capabilities, and every teleoperation session generates training data that improves the AI models.
A single human operator can manage up to 15 EVE units simultaneously in a supervisory role, intervening only when a robot encounters a situation it cannot handle autonomously. This ratio is a key part of EVE's commercial value proposition, as it allows a small team of human operators to oversee a large fleet of robots across multiple facilities.[17]
In March 2023, OpenAI led a $23.5 million Series A2 investment in 1X Technologies, marking one of the first major investments by the OpenAI Startup Fund into a robotics company.[18] The partnership went beyond financial investment. 1X began integrating its foundational embodied intelligence models with OpenAI's large language models for high-level planning and natural language instruction comprehension. This integration allowed EVE to process voice commands and translate them into physical actions, bridging the gap between human language and robot behavior.
The OpenAI relationship significantly accelerated EVE's autonomy progress. By combining 1X's embodied learning models (trained on real-world robot interaction data) with OpenAI's language and reasoning capabilities, the system achieved more flexible and generalizable task execution than either approach could deliver independently.
EVE's most significant commercial deployment came through a partnership with ADT Commercial (later rebranded as Everon), one of the largest security services providers in the United States. The relationship began in 2020 when ADT invested in Halodi Robotics and the two companies began co-developing autonomous security patrol capabilities.[19]
In March 2022, ADT Commercial signed an agreement to purchase 140 EVE units for deployment in autonomous night security patrols at commercial buildings across the United States. This order was described at the time as the world's largest single order for humanoid robots.[1] Each EVE unit cost approximately 1 million Norwegian kroner (roughly $100,000 USD) to manufacture, while ADT leased each robot for about 500,000 kroner ($50,000 USD) per year, giving the contract an estimated annual value of approximately 70 million kroner ($7 million USD).[20]
The security deployment showcased several of EVE's core capabilities:
At CES 2023 in Las Vegas, ADT Commercial unveiled its EvoGuard brand of intelligent autonomous guarding solutions, featuring EVE as the humanoid robot component alongside the Tando autonomous indoor drone (developed with Israel-based Indoor Robotics). 1X CEO Bernt Bornich demonstrated EVE's patrol capabilities live at ADT's booth. EvoGuard was positioned as a response to high turnover rates and ongoing labor shortages in the commercial security industry, offering 24/7 surveillance capability without the staffing challenges of human guards.[21]
Reports from various sources indicate that between 150 and 250 EVE robots were ultimately deployed or contracted through the ADT/Everon partnership for commercial building security across the United States.
EVE was deployed in healthcare settings, most notably at Sunnaas Hospital in Norway, one of the country's leading rehabilitation hospitals. In this environment, EVE assisted with tasks such as meal delivery, patient monitoring, and facility support operations. The hospital deployment provided valuable data on human-robot interaction in sensitive healthcare contexts, where safety and reliability are paramount.[22]
The Belgian elderly care organization I-Mens also conducted paid pilot programs with EVE, exploring the robot's potential for assisting care workers with routine tasks in residential care facilities.
Several other organizations conducted commercial pilot programs with EVE:
These deployments expanded EVE's operational data set beyond security applications, contributing to the breadth of the company's training data for embodied AI development.
Norwegian universities gained access to EVE units for academic research purposes, allowing researchers to study human-robot interaction, autonomous navigation, and manipulation in controlled settings. Additionally, the robot was used by organizations including the U.S. Department of Defense for building surveillance applications, where fleets of up to 10 EVE robots could autonomously patrol large facility areas with a single centralized human operator available for VR intervention when needed.[23]
In November 2024, EVE gained unexpected mainstream attention when popular Twitch streamer Kai Cenat featured an EVE unit during his Mafiathon 2 livestream, stating he had purchased the robot for approximately $24,000.[24]
EVE manufacturing was conducted at two facilities: the company's original site in Moss, Norway, and a facility in Dallas, Texas. The dual-location production strategy allowed 1X to serve both European and North American customers while maintaining proximity to its engineering teams on both continents.
The third-generation production version of EVE cost approximately $100,000 per unit to manufacture, reflecting the complexity of the Revo1 actuator system and the integrated perception and computing hardware. This unit cost positioned EVE as an enterprise-grade product rather than a consumer device, with the ADT leasing model (approximately $50,000 per year per unit) providing a recurring revenue stream for 1X.[20]
The company's manufacturing goal during the 2022 to 2023 period was 140 units to fulfill the ADT contract, with additional units produced for healthcare, logistics, and research customers. This production experience gave 1X valuable lessons in scaling humanoid robot manufacturing, lessons that would inform the company's later investment in a dedicated in-house manufacturing facility in Hayward, California, for NEO production.
Safety was a central design priority for EVE, given that the robot operates in shared environments alongside human workers, patients, and the general public. The Revo1 quasi direct-drive actuator system provides the hardware foundation for safe operation, but 1X implemented multiple additional layers of safety:
EVE served as far more than a standalone commercial product for 1X Technologies. It functioned as a critical proving ground for the hardware, software, and operational strategies that would later be applied to 1X NEO, the company's bipedal humanoid designed for home environments.
The Revo1 quasi direct-drive technology validated in EVE became the conceptual ancestor of NEO's Tendon Drive actuation system. While NEO uses a more advanced tendon-based transmission rather than EVE's cable-driven system, the core principles of low-gear-ratio actuation, high backdrivability, and safe force control carried directly from one platform to the other. The engineering team's years of experience with Revo1 reliability, maintenance, and performance characteristics informed every aspect of NEO's actuator design.[26]
Perhaps EVE's most important contribution to NEO was the massive dataset of real-world operational data it generated. By deploying hundreds of EVE units in commercial facilities across multiple industries and countries, 1X accumulated an enormous volume of sensor recordings, teleoperation sessions, and autonomous operation logs. This data was used to train the foundational AI models that would later be adapted for NEO's Redwood AI system.
The iterative training pipeline developed for EVE (base model training, specialized fine-tuning, site-specific adaptation) established the methodology that 1X continues to use for NEO. The company's approach of treating deployed robots as continuous data-collection platforms, where every autonomous action and every teleoperation session contributes to model improvement, was proven at scale with EVE before being applied to the home robotics domain.[27]
Running a fleet of humanoid robots in real commercial facilities taught 1X practical lessons that cannot be learned in a laboratory:
While EVE's wheeled design proved effective for flat indoor environments, 1X recognized that a home robot would need to handle stairs, navigate cluttered rooms, and access the full range of spaces designed for human bodies. The transition from wheels to legs in NEO was a strategic decision driven by the target environment. Bipedal locomotion allows NEO to shuffle sideways through narrow passages, brace against surfaces for leverage, lean for balance during heavy lifting, and access multi-level homes without requiring ramps or elevators.
1X CEO Bernt Bornich has described EVE as the company's "workhorse" that gathered "the mountains of data needed to teach an AI how to exist outside a laboratory," enabling the company to approach the far more challenging problem of bipedal home robotics with a foundation of real-world operational experience that few competitors possess.[28]
As of 2025, EVE remains in production and continues to operate in deployed enterprise environments. The robot is marketed as EVE Industrial, reflecting its positioning for commercial and institutional applications. Existing deployments in security, logistics, and healthcare continue to function, and the robot's operational data still feeds into 1X's AI training pipelines.
However, 1X's primary strategic focus has shifted to the NEO platform. The company's $100 million Series B funding round in January 2024, led by EQT Ventures with participation from Samsung Next and the OpenAI Startup Fund, was directed primarily toward scaling NEO production and AI development.[29] The opening of 1X's new 80,000-square-foot global headquarters in Palo Alto, California, in July 2025 further signaled the company's emphasis on consumer robotics over industrial deployments.
EVE is no longer being actively marketed to new customers, though it remains available and supported for existing deployments. Its legacy is best measured not in unit sales but in the operational knowledge, training data, and engineering heritage it provided to the NEO program.
| Feature | EVE | NEO Beta | NEO Gamma |
|---|---|---|---|
| Year introduced | 2018 | 2024 | 2025 |
| Locomotion | Wheeled (self-balancing) | Bipedal | Bipedal |
| Height | 188 cm (6 ft 2 in) | 165 cm (5 ft 5 in) | 167 cm (5 ft 6 in) |
| Weight | 87 kg (192 lb) | 30 kg (66 lb) | 30 kg (66 lb) |
| Degrees of freedom | 25 | ~75 | 75 |
| Hand DOF | 1 per hand (gripper) | 22 per hand | 22 per hand |
| Max speed | 14.4 km/h (9 mph) | 12 km/h (7.5 mph) | 12 km/h (7.5 mph) |
| Battery life | 4 to 6 hours | 2 to 4 hours | Up to 5.5 hours |
| Noise level | Not specified | ~32 dB | ~22 dB |
| Actuation | Revo1 quasi direct-drive | Tendon Drive | Tendon Drive |
| Target market | Enterprise / industrial | Home / consumer | Home / consumer |
| AI compute | Intel i7 + NVIDIA Xavier | Not disclosed | 1X NEO Cortex (NVIDIA Jetson Thor) |