1X Technologies
Last reviewed
Apr 26, 2026
Sources
25 citations
Review status
Source-backed
Revision
v6 · 4,211 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Apr 26, 2026
Sources
25 citations
Review status
Source-backed
Revision
v6 · 4,211 words
Add missing citations, update stale details, or suggest a clearer explanation.
1X Technologies (formerly Halodi Robotics) is a Norwegian-American artificial intelligence and robotics company focused on building general-purpose humanoid robots for home and commercial use. Founded in 2014 by Bernt Øivind Børnich in Moss, Norway, the company develops bipedal and wheeled humanoid platforms powered by end-to-end neural networks, with the goal of creating robots capable of performing any kind of work autonomously. 1X is headquartered in Palo Alto, California, and maintains manufacturing operations in both Hayward, California, and Moss, Norway.
The company is best known for two robot platforms: EVE, a wheeled humanoid deployed in security and logistics roles, and NEO, a bipedal humanoid designed for household use. 1X gained significant attention in 2023 when OpenAI led its Series A2 funding round, making it one of the first humanoid robot companies to receive backing from a major AI lab. As of early 2026, 1X employs roughly 965 people and has raised over $136 million in venture capital.
1X Technologies was founded in May 2014 under the name Halodi Robotics by Bernt Øivind Børnich, Phuong Nguyen, Jørgen Sundell, and Pål Løken in Moss, Norway. Børnich, who holds a Bachelor's degree in Robotics and Nanoelectronics from the University of Oslo, decided at age 11 that he wanted to build humanoid robots. Nguyen, the company's original CTO, earned a Ph.D. in Robotics from the University of Ulsan and previously worked in intensive care at Oslo University Hospital.
During its early years, the company focused on developing proprietary motors and cable-driven transmission systems designed for safe physical interaction between robots and humans. In 2018, Halodi introduced the Revo1, which the company described as the world's highest torque-to-weight direct-drive servo motor. This brushless direct current (BLDC) motor, paired with low gear ratio cable drives, became the foundation of the company's actuation technology and was inspired by the mechanics of human tendons.
Also in 2018, Halodi released its first humanoid robot, EVE, a wheeled platform with 25 degrees of freedom designed for logistics, security, and healthcare applications. EVE represented the company's initial effort to commercialize humanoid robotics for enterprise customers.
In April 2021, Halodi Robotics secured a strategic investment from ADT Commercial, the commercial security division of ADT Inc., to develop robots that could complement physical security solutions. This partnership led to the company's first major commercial contract.
In March 2022, Halodi Robotics signed an agreement to deliver 140 EVE robots to ADT Commercial. This represented the largest single order of humanoid robots in the industry at that time, with an estimated annual leasing revenue of approximately 70 million Norwegian kroner (roughly $7 million). Each third-generation EVE unit cost about 1 million Norwegian kroner ($100,000) to produce, and ADT leased each unit for approximately 500,000 kroner per year.
In late 2022, the company rebranded from Halodi Robotics to 1X Technologies, signaling a strategic shift from industrial pilots to consumer home deployment. The new name reflected the company's broader ambition: to build robots at a 1:1 ratio with the human population. This pivot was driven by Børnich's belief that robots must live and learn alongside humans to achieve true intelligence, a philosophy that would become central to 1X's product strategy.
In March 2023, 1X raised $23.5 million in a Series A2 funding round led by the OpenAI Startup Fund, with participation from Tiger Global, Sandwater, Alliance Ventures, and Skagerak Capital. The round valued the company at approximately $210 million post-money. This investment marked one of OpenAI's earliest bets on physical AI and embodied intelligence, bringing significant industry attention to 1X.
In January 2024, the company closed a $100 million Series B round led by EQT Ventures, with additional investment from Samsung NEXT, the Nistad Group, and existing backers Sandwater and Skagerak Capital. This brought 1X's total funding to approximately $136.5 million and positioned the company to accelerate development of its bipedal humanoid robot, NEO.
On August 30, 2024, 1X introduced NEO Beta, the company's first full bipedal humanoid robot. The 66-pound robot was designed specifically for safe home environments and represented a major hardware departure from the wheeled EVE platform.
On February 21, 2025, 1X unveiled NEO Gamma, a refined version incorporating feedback from early testing. The Gamma iteration featured a sleeker design aesthetic, reduced noise levels, improved locomotion, and enhanced AI capabilities for home environments.
In January 2025, 1X acquired Kind Humanoid, a Palo Alto-based robotics startup that had been working with renowned designer Yves Behar on a humanoid called Mona. The acquisition brought Kind Humanoid's founder, Christoph Kohstall (formerly a robotics researcher at Google), into the 1X team, along with design talent from the Behar collaboration.
On October 28, 2025, 1X opened pre-orders for the consumer version of NEO, marketing it as the world's first consumer-ready humanoid robot. Early Access units were priced at $20,000, with a subscription option of $499 per month available at a later date. The company announced that U.S. deliveries would begin in 2026, with international expansion starting in 2027.
In December 2025, 1X announced a strategic partnership with EQT, one of the world's largest private equity firms and an existing investor. Under the agreement, 1X and EQT plan to facilitate the rollout of up to 10,000 humanoid robots across EQT's approximately 300 portfolio companies between 2026 and 2030, targeting use cases in manufacturing, warehousing, logistics, facility operations, and healthcare.
In September 2025, reports indicated that 1X was seeking to raise up to $1 billion in new funding at a targeted valuation of at least $10 billion, representing more than a 12x increase from the Series B valuation.
EVE is 1X's wheeled humanoid robot, first introduced in 2018 and designed primarily for enterprise applications in security, logistics, and healthcare.
| Specification | Details |
|---|---|
| Type | Wheeled humanoid |
| Height | 6 feet 2 inches (188 cm) |
| Weight | 192 pounds (87 kg) |
| Degrees of Freedom | 25 |
| Top Speed | 9 mph (14.5 km/h) |
| Payload Capacity | 33 pounds (15 kg) |
| Battery Life | Up to 4 hours |
| Actuation | Revo1 tendon-driven motors |
| Navigation | Multi-terrain wheels; can open doors, take elevators |
EVE operates autonomously by default, using artificial intelligence to navigate workspaces, patrol facilities, open doors with different handle types, and identify people and objects at a distance. The robot's multi-terrain wheels allow it to navigate corners and ride elevators. ADT Commercial has leased EVE units for security applications, where the robot patrols buildings during night shifts and checks employee identification badges.
EVE served as 1X's primary revenue-generating product through its enterprise leasing model. In 2022, EVE generated approximately $7 million in leasing revenue through the ADT contract. The robot also provided a critical platform for 1X to develop and refine its AI and actuation technologies before transitioning to the more ambitious bipedal NEO platform.
NEO is 1X's bipedal humanoid robot, designed for household use and positioned as a general-purpose home assistant. The robot has gone through several hardware iterations: NEO Beta (August 2024), NEO Gamma (February 2025), and the consumer-ready NEO (October 2025).
| Specification | Details |
|---|---|
| Type | Bipedal humanoid |
| Height | 5 feet 6 inches (167 cm) |
| Weight | 66 pounds (30 kg) |
| Total Degrees of Freedom | 75 |
| Hand Degrees of Freedom | 22 per hand |
| Lifting Capacity | 154 pounds (70 kg) |
| Carrying Capacity | 55 pounds (25 kg) |
| Arm Payload | 18 pounds (8 kg) per arm |
| Battery | 842 Wh |
| Battery Life | Approximately 4 hours |
| Charge Time | Approximately 2 hours |
| Noise Level | 22 dB |
| Actuation | Patented Tendon Drive |
| Safety | 3D lattice polymer soft body; pinch-proof joints; HIC < 250 |
| Onboard Compute | NVIDIA embedded GPU |
| Price (Early Access) | $20,000 |
| Subscription Option | $499/month |
NEO's body is wrapped head-to-toe in custom 3D lattice polymer structures, making the robot's exterior entirely soft and pinch-proof. All joints are covered and inaccessible from outside, preventing accidental pinching or crushing injuries. The robot uses 1X's patented Tendon Drive, which the company claims employs the highest torque-density motors available to power tendon-based transmissions. This design produces gentle, compliant movements that are inherently safe around people. NEO's Head Injury Criterion (HIC) rating is below 250, and its noise level of 22 dB is quieter than a modern refrigerator.
The NEO Gamma iteration introduced a knit suit and shoes that are 3D-printed from nylon using Japanese Shemiseki machinery, giving the robot a minimalist, home-friendly aesthetic. The design also includes "Emotive Ear Rings" intended to improve nonverbal communication between the robot and its users.
NEO is designed to perform household tasks including cleaning, lifting, organizing, unloading dishwashers, watering plants, and retrieving objects. The robot can walk with a natural human gait and arm swings, squat down to pick things up from the ground, sit in chairs, and maintain balance throughout these activities. Its 22-degree-of-freedom hands enable human-level dexterity for grasping a wide variety of objects.
NEO supports natural language interaction through voice commands, processed by offboard speech-to-text large language models that convert spoken instructions into command vectors for the onboard AI system.
During the initial Early Access phase in 2026, NEO operates using a hybrid teleoperation model rather than full autonomy. Human teleoperators (referred to as "1X Experts") can remotely assist the robot with complex or unfamiliar tasks using VR headsets, viewing live camera feeds from NEO and controlling the robot in real time. This teleoperation occurs only with explicit user consent, and all interactions are encrypted. 1X has described this as an intentional human-in-the-loop design that simultaneously provides utility to customers and generates training data to improve the robot's autonomous capabilities over time.
The foundation of 1X's hardware technology is the Revo1 motor, developed in 2018. The Revo1 is a high-torque brushless direct current (BLDC) motor that uses a cable-driven differential transmission rather than the conventional approach of high-ratio harmonic gears. This design, inspired by human tendon mechanics, achieves what 1X describes as the highest torque-to-weight ratio of any direct-drive motor in the world. The low-gearing, low-friction design produces compliant, quiet movements that are safe for close human interaction.
The Revo1 technology evolved into 1X's patented Tendon Drive, which powers both the EVE and NEO platforms. The Tendon Drive's compliance and backdrivability allow NEO to absorb impacts safely, making it well-suited for operation in household environments where contact with people, pets, and furniture is expected.
Redwood is 1X's primary AI model for controlling the NEO robot. It is a 160-million-parameter vision-language transformer that serves as NEO's onboard "brain," jointly handling perception, navigation, and manipulation.
Redwood processes three types of input simultaneously: pre-trained language embeddings (from voice commands), vision tokens from a vision transformer, and proprioception data (joint positions and forces). These inputs pass through transformer blocks to create a latent representation, which is then decoded into physical actions using a diffusion policy.
The model runs entirely on NEO's embedded NVIDIA GPU at approximately 5 Hz, eliminating dependence on cloud infrastructure or external computing resources.
Redwood is among the first vision-language-action models (VLAs) to jointly control locomotion and manipulation, enabling behaviors such as bracing against walls while reaching for objects, or leaning into tasks that require whole-body coordination. The model supports mobile bimanual manipulation, meaning NEO can navigate a space while simultaneously using both hands to interact with objects.
Redwood learns from both successful and failed task attempts, not only from successful demonstrations. Additional "cognitive" prediction targets help ground visual representations, improving the model's ability to generalize to environments and objects not seen during training.
Demonstrated tasks include retrieving objects, opening doors, navigating homes, grasping diverse items, and performing multi-contact manipulation. 1X formally appointed Mohi Khansari as Head of Robot Learning and chief architect of Redwood after the departure of Eric Jang, who served as VP of AI from 2022 to 2025.
The Redwood Mobility system is a reinforcement learning-based locomotion controller that gives NEO its full range of physical movements. The system uses a two-stage approach: a high-level kinematic planner generates target trajectories based on human motion-capture data (defining what natural movement looks like), and a low-level RL controller figures out the physics of tracking that trajectory while maintaining balance.
The mobility controller runs at 100 Hz and enables NEO to walk in any direction with a natural human gait and arm swings, sit down and stand up, kneel, get up from the floor, and climb stairs using stereo vision. All of these skills were learned using reinforcement learning from human motion-capture data.
The controller is trained entirely in simulation across thousands of randomized physics parameters (friction, mass, sensor noise) so that it transfers robustly to the real world despite the simulation being only an approximation of real physics.
The 1X World Model (1XWM) is a generative AI system that enables NEO to learn new tasks by watching video rather than requiring explicit programming or teleoperated demonstrations. Announced in January 2026, the World Model represents 1X's approach to scaling robot intelligence through video pretraining.
The 1XWM consists of two core components:
The system operates as a two-stage pipeline: the World Model generates a video showing the intended future, the IDM extracts the necessary action trajectory from that video, and the robot executes the sequence in the real world.
The World Model backbone follows a multi-stage training strategy:
The training pipeline also uses "caption upsampling," where vision-language models generate detailed visual descriptions for training data to improve the model's ability to follow text prompts accurately.
A key feature of the 1XWM is its self-learning capability. The system creates a flywheel where exploration, evaluation, and policy refinement are driven by NEO's own deployment experience rather than being limited to expert demonstrations. By leveraging the understanding of world dynamics inherent in internet-scale video, the model can generalize to novel objects, motions, and tasks without pre-training on large-scale robot-specific data.
Test-time compute scaling also improves performance: generating multiple parallel rollouts (up to 8) and selecting the best one increases task success rates. For example, generating 8 options improved performance on a tissue-pulling task from 30% to 45% success.
The team has identified several current limitations, including weak 3D spatial grounding from monocular pretraining, occasional physically implausible video generations, a 5-second planning horizon that constrains longer tasks, and the need for closed-loop replanning to handle drift and partial observability.
In March 2025, 1X and NVIDIA announced a research collaboration at NVIDIA's GTC 2025 keynote. The teams worked together to integrate NVIDIA's GR00T N1 foundation model with NEO Gamma. In a demonstration during Jensen Huang's keynote, NEO performed a dish-loading task autonomously using a single end-to-end neural network based on GR00T N1.
To enable the collaboration, 1X created a dataset API for NVIDIA to access data collected from 1X offices and employee homes, along with an inference SDK to serve model predictions at a continuous 5 Hz vision-action loop using either NEO's onboard GPU or an offboard GPU. According to CEO Børnich, NVIDIA's GR00T N1 provided a significant boost to NEO's reasoning and skill capabilities, with the model being developed and deployed over the course of a single week at a 1X employee's home.
| Name | Role | Background |
|---|---|---|
| Bernt Øivind Børnich | Founder and CEO | B.Sc. in Robotics and Nanoelectronics, University of Oslo; previously at Data Respons, Inmeta Crayon, Bring Dialogue |
| Mohi Khansari | Head of Robot Learning | Chief architect of Redwood AI |
| Jorge Milburn | VP of Sales | Former global growth lead at Tesla |
| Per Selvaag | VP of Design | Former lead designer at BMW |
| Dar Sleeper | VP of Growth | Former Tesla Cybertruck product manager |
| Vikram Kothari | VP of Operations | Former SpaceX Dragon/Starship supply chain manager; previously led Microsoft hardware supply chain |
| Christoph Kohstall | Robotics Researcher | Former Google robotics researcher; joined through Kind Humanoid acquisition |
Eric Jang, a former Google Brain researcher, joined 1X in April 2022 as VP of AI and was instrumental in steering the company's transition from industrial to consumer robotics. During his tenure, Jang oversaw the development of Redwood AI, the World Model, and the broader AI research agenda. He departed the company in 2025, with Mohi Khansari succeeding him as Head of Robot Learning.
1X was originally headquartered in Moss, Norway, where it built its initial engineering team and manufacturing facility. The Moss facility handles actuator manufacturing, robot assembly, and quality testing for both EVE and NEO, following quality practices adapted from automotive industry standards including Design Failure Mode and Effects Analysis (DFMEA).
In July 2025, 1X announced the consolidation of its Moss, Norway, and Sunnyvale, California, teams into a new global headquarters in Palo Alto, California. The 80,000-square-foot facility can seat 400 people and was designed to accelerate product development by co-locating the company's AI, engineering, and business teams.
The company maintains three office locations: the Palo Alto headquarters, the Moss manufacturing and engineering campus in Norway, and an office in Oslo, Norway.
1X has outlined an ambitious manufacturing roadmap for NEO:
| Year | Production Target |
|---|---|
| 2025 | Thousands of units |
| 2026 | Tens of thousands of units |
| 2027 | Hundreds of thousands of units |
| 2028 | Millions of units |
The company has stated that the NEO manufacturing process was designed from the outset to be more efficient than EVE production, with vertical integration of component manufacturing to support rapid scaling.
| Round | Date | Amount | Lead Investor | Notable Participants | Valuation |
|---|---|---|---|---|---|
| Early funding | April 2017 | Undisclosed | N/A | Norwegian investors | N/A |
| Series A2 | March 2023 | $23.5 million | OpenAI Startup Fund | Tiger Global, Sandwater, Alliance Ventures, Skagerak Capital | ~$210M post-money |
| Series B | January 2024 | $100 million | EQT Ventures | Samsung NEXT, Nistad Group, Sandwater, Skagerak Capital | ~$820M |
| Reported Series C | September 2025 | Up to $1 billion (seeking) | TBD | N/A | $10B+ (targeted) |
Total confirmed funding through Series B stands at approximately $136.5 million. The OpenAI-led Series A2 was notable as one of OpenAI's first investments in a physical robotics company, reflecting the AI lab's interest in embodied AI and the physical manifestation of artificial general intelligence.
1X operates in the rapidly growing humanoid robot market, which Goldman Sachs projects could reach $38 billion by 2035. The company competes with several well-funded rivals, each taking distinct technological and commercial approaches.
| Company | Robot | Primary Focus | Key Technology | Total Funding | Status |
|---|---|---|---|---|---|
| 1X Technologies | NEO | Home/consumer | Tendon-drive; end-to-end neural networks | ~$136.5M | Pre-orders open; U.S. deliveries 2026 |
| Tesla | Optimus | Internal/industrial | Actuator-based; leverages vehicle AI data | Internal | No external customers yet |
| Figure AI | Figure 02/03 | Industrial/commercial | Actuator-based; BMW factory deployment | ~$1.9B | Deployed at BMW; negotiating with UPS |
| Boston Dynamics | Atlas (electric) | Research/industrial | Hydraulic/electric hybrid | Hyundai-owned | Factory testing with Hyundai |
| Agility Robotics | Digit | Logistics/warehousing | Purpose-built for logistics | N/A | RaaS agreement with GXO Logistics |
1X differentiates itself through several strategic choices. Its consumer-first approach contrasts with competitors like Figure AI and Agility Robotics, which have pursued industrial deployments first. The company's tendon-based actuation system produces quieter, more compliant movements compared to traditional actuator-based designs, making NEO better suited for home environments where safe human-robot interaction is essential.
1X's AI strategy also differs from competitors. While Tesla leverages data from its fleet of millions of vehicles and Figure AI focuses on factory teleoperation data, 1X is building a pipeline that combines home teleoperation data, internet-scale video pretraining, and reinforcement learning in simulation to train its models.
Several significant challenges face 1X as it scales production and deployment:
Manufacturing scaling: CEO Børnich has acknowledged the difficulty of transitioning from low-volume to high-volume production, noting that "something kind of magical happens when you go from tens of thousands to hundreds of thousands of units" and that many companies fail at this stage.
AI limitations: As of mid-2025, 1X leadership candidly stated that "all our current AIs completely fail at most reasoning tasks," highlighting the gap between current capabilities and the vision of fully autonomous home robots.
Teleoperation dependency: The initial reliance on human teleoperators raises privacy concerns, as remote operators view live camera feeds from inside customers' homes. While 1X has implemented encryption and consent mechanisms, consumer acceptance of this model remains uncertain.
Cost competitiveness: Current bill-of-materials costs for humanoid robots range from $10,000 to $300,000 depending on configuration. Achieving mass-market affordability while maintaining quality and safety presents a major engineering and manufacturing challenge.
Competition: The humanoid robotics space is attracting enormous investment, with global patent filings mentioning "humanoid" quadrupling since 2019. Chinese manufacturers are emerging as potential low-cost competitors, and well-resourced companies like Tesla and Figure AI are pursuing aggressive development timelines.
The broader humanoid robotics market is projected for significant growth. Goldman Sachs estimates the market could reach $38 billion by 2035, with robot shipments increasing fourfold to 1.4 million units. Morgan Stanley projects the U.S. market alone could reach a cumulative installed base of 8 million units by 2040 and 63 million by 2050, generating roughly $240 billion in annual revenue by 2040.
1X's EQT partnership, with its plan for up to 10,000 robots across 300 portfolio companies by 2030, represents one of the largest commercial commitments in the humanoid robotics industry. Combined with the consumer Early Access program launching in 2026, the company is pursuing a dual-market strategy targeting both home and enterprise customers simultaneously.