ANYbotics AG is a Swiss robotics company that designs, builds, and deploys autonomous quadrupedal robots for industrial inspection. Headquartered in Zurich, Switzerland, the company was founded in 2016 as a spin-off from ETH Zurich's Robotic Systems Lab. ANYbotics is best known for its ANYmal line of four-legged robots, which are used to perform automated inspections in sectors such as oil and gas, chemicals, power generation, metals, and mining. The company has raised over $150 million in total funding and, as of late 2025, employs roughly 200 people with offices in Zurich and San Francisco.
The technology behind ANYbotics traces back to 2009, when researchers at ETH Zurich's Robotic Systems Lab began developing four-legged robots. The first prototype, called ALoF, was a proof-of-concept machine that could walk and self-right on sand dunes but was slow, rigid, and not practical for real-world deployment. ALoF demonstrated that legged locomotion was feasible for field applications, motivating continued research into more capable designs.
In 2012, the team built StarlETH, a quadruped fitted with spring-loaded joints that could hop, jump, and climb. StarlETH represented a leap forward in dynamic locomotion because its compliant joints, which used springs in series with the motors, allowed the robot to absorb impacts and store energy. This approach, known as series-elastic actuation, proved essential for stable and robust walking on uneven ground and served as a direct precursor to the actuator technology used in ANYmal.
The first ANYmal prototype was completed in 2015 under the leadership of Marco Hutter, who heads the Robotic Systems Lab. This prototype introduced the custom series-elastic actuator module (later branded ANYdrive) that would become a defining feature of the platform. The research was supported by multiple programs including NCCR Robotics, the EU ECHORD++ project, the Branco Weiss Fellowship, and ETH Pioneer Fellowships.
ANYbotics AG was formally incorporated in 2016 to commercialize the ANYmal technology. The founding team consisted of eight roboticists: Péter Fankhauser, Marco Hutter, Christian Gehring, Samuel Bachmann, Remo Diethelm, Hanspeter Fässler, Andreas Lauber, and Roland Siegwart. The company was incubated through the Wyss Zurich translational research program, which bridges the gap between academic research and commercial products at ETH Zurich and the University of Zurich.
Péter Fankhauser, who holds a doctorate in robotics from ETH Zurich focused on perception and motion planning for legged robots, serves as co-founder and CEO. During the company's early years, he led business development and marketing activities before taking on the full CEO role as the organization scaled. Marco Hutter, an Associate Professor at ETH Zurich and Branco Weiss Fellow, serves as co-founder and Vice-Chairman. His research interests span the development of novel machines and actuation concepts together with control, planning, and learning algorithms for locomotion and manipulation. Roland Siegwart, a full Professor of Autonomous Systems at ETH Zurich, founding Co-Director of Wyss Zurich, and recipient of the IEEE RAS Pioneer Award, is also among the co-founders.
ANYbotics recorded its first sales in 2017. Early deployments focused on offshore oil and gas platforms and other challenging environments, including sewer inspections, that demonstrated the robot's ability to operate where human access was difficult or dangerous.
In 2019, the company unveiled ANYmal C, a major redesign optimized for rugged industrial environments. ANYmal C featured a streamlined, weather-sealed body with an IP67 rating, a 50-kilogram chassis, 360-degree obstacle detection, and autonomous docking for recharging. The redesigned actuators offered a 40% higher torque-to-weight ratio compared to the previous generation.
The DARPA Subterranean Challenge in September 2021 brought the platform significant visibility. Team CERBERUS, an international consortium led by the University of Nevada, Reno in collaboration with ETH Zurich and other institutions, deployed four ANYmal C robots in the final event. The team won the competition and its $2 million grand prize. During the final mission, all four ANYmals operated for nearly one hour, autonomously navigating tunnel, urban, and cave environments covering over 1,700 meters without a single fall. The robots traversed rocky and slippery surfaces, high steps, steep inclines, and dense smoke, demonstrating why the top six teams in the competition all employed legged robots.
ANYbotics opened commercial sales of ANYmal D in late 2022 and received the IEEE Robotics and Automation Award for Product Innovation that same year, with the citation describing ANYmal as "a versatile quadrupedal robot with high mobility for inspection and maintenance." In March 2022, the company also announced ANYmal X, the world's first Ex-proof legged robot designed for Zone 1 explosive atmospheres.
In November 2024, ANYbotics opened an office in San Francisco, in collaboration with the Swiss Business Hub USA and under the patronage of the Swiss Consul General, to support its expansion into North America. By December 2024, the company had deployed nearly 200 robots conducting thousands of weekly inspections at energy and industrial sites worldwide.
In February 2025, ANYbotics launched Data Navigator, an asset management platform that transforms inspection data into actionable insights for preventive maintenance. In June 2025, the company introduced a Gas Leak and Presence Detection solution combining modular gas detectors with a 360-degree acoustic imaging payload. First customer deliveries of ANYmal X are planned for 2026, and a Series C funding round targeting broader growth and new industries could follow in the same year.
ANYmal is an electrically driven, four-legged robot designed for autonomous operation in complex, multi-level industrial facilities. The name stands for "Autonomous Navigating Your MAchine for Locomotion." Each ANYmal has 12 degrees of freedom (three actuated joints per leg) and uses custom series-elastic actuators called ANYdrives. These actuators incorporate high-torque motors, harmonic drive gears, and a rotational spring in series, enabling precise torque control, impact absorption, and energy storage. Each ANYdrive delivers up to 40 Nm of torque at a speed of 12 rad/s with a position accuracy of 0.025 degrees and a torque resolution of 8 mNm. The spring element in each actuator mimics the way forces are handled by biological muscles and tendons, making the robot more robust by absorbing sharp impacts and temporarily storing energy to improve peak power output.
The robot runs Ubuntu Linux with a ROS-based software framework distributed across three independent computers connected via an internal network: one handles locomotion control, one manages navigation and autonomy, and one runs inspection algorithms. Navigation relies on 360-degree LiDAR (with up to 100-meter range), six depth cameras for obstacle detection, and wide-angle cameras for teleoperation. The inspection payload, mounted on a pan-tilt unit with +/- 165 degrees horizontal and +/- 90 degrees vertical range, includes a 20x optical zoom camera capable of 4K video, a thermal camera covering -20 to 500 degrees Celsius, an ultrasonic microphone, and LED lighting. The body features a Kevlar belly panel and foam impact absorption to survive falls exceeding 0.5 meters, along with four Optoforce tactile sensors in the feet for terrain feedback.
Since the original 2015 prototype, ANYbotics has released several generations of the ANYmal platform. Each version introduced significant improvements in hardware, sensing, software, and industrial readiness.
| Version | Year | Key Features |
|---|---|---|
| ALoF (precursor) | 2009 | First ETH Zurich quadruped; proof of concept for walking and self-righting |
| StarlETH (precursor) | 2012 | Spring-loaded joints enabling running, jumping, and climbing |
| ANYmal Alpha | 2015/2016 | First ANYmal; modular ANYdrive actuators, LiDAR navigation, ~30 kg weight, 650 Wh battery, 2 to 4 hour runtime |
| ANYmal Beth | 2017 | Ruggedized design, waterproof housing for field inspections |
| ANYmal B | 2018 | Upgraded actuators and vision systems; first offshore deployment |
| ANYmal C | 2019 | IP67-rated, 50 kg, 360-degree vision, autonomous docking, 40% higher actuator torque-to-weight ratio, 2+ hour battery life, 1 m/s speed |
| ANYmal D | 2021 | Fourth-generation; 55 kg, 932 Wh battery, 15 kg payload, dual i7 CPUs, complete autonomous inspection system |
| ANYmal X | 2023 (announced) | ATEX/IECEx Zone 1 certified for explosive atmospheres; 60.1 kg; IP67; integrated gas detection; first deliveries planned 2026 |
ANYmal D is the standard commercial model for non-hazardous industrial environments.
| Specification | Value |
|---|---|
| Length | 93 cm |
| Width | 53 cm |
| Height (standing) | 89 cm |
| Weight | ~55 kg |
| Speed | Up to 1.3 m/s (4.68 km/h) |
| Degrees of Freedom | 12 |
| Battery | 932 Wh lithium-ion |
| Runtime | 90 to 120 minutes |
| Charge Time | 3 hours (100 minutes to 70%) |
| Payload Capacity | Up to 15 kg |
| IP Rating | IP67 (dust-tight and waterproof to 1 m submersion) |
| Max Stair Angle | 45 degrees |
| Max Slope | 30 degrees |
| Step Height | Up to 35 cm |
| Gap Crossing | Up to 30 cm |
| Computing | Dual Intel Core i7 (6-core), 8 GB RAM each, two 240 GB SSDs |
| Connectivity | Wi-Fi 802.11ac (2.4/5 GHz), optional 4G LTE |
| Operating System | Ubuntu Linux |
ANYmal X is the explosion-proof variant engineered for hazardous environments in the oil and gas and chemical industries. It is the world's first Ex-certified legged robot.
| Specification | Value |
|---|---|
| Length | 891 mm |
| Width | 651 mm |
| Height (standing) | 872 mm |
| Height (lying down) | 440 mm |
| Weight (with inspection payload and gas sensor) | 60.1 kg |
| Speed | Up to 1.0 m/s |
| Runtime | 60 to 120 minutes |
| Charge Time | 3 hours (100 minutes to 70%) |
| IP Rating | IP67 |
| Explosion-Proof Certification | ATEX and IECEx, Zone 1 |
| Safety Standard | ISO 13849-1:2015 Category 3 PL c |
| Inspection Payload Weight | 6 kg |
| Pan-Tilt Range | Pan: +/- 165 degrees; Tilt: -90 to +90 degrees |
| Zoom Camera | 20x optical, 4K resolution |
| Thermal Camera | -20 to 500 degrees Celsius |
| Gas Sensors | Up to 2 modular sensors (compatible with Honeywell SensePoint and Drager PIR 3000) |
| Detectable Gases | Methane, carbon monoxide, hydrogen sulfide, and others |
| Obstacle Detection | 0.3 to 3 m range, depth cameras with 87.3 x 58.1 x 95.3 degree FOV |
| Environment Mapping | 360-degree LiDAR |
| Connectivity | Wi-Fi 802.11ac wave2 (2.4/5 GHz) |
| Operating System | Ubuntu 20.04 |
ANYmal's locomotion capabilities are rooted in research at ETH Zurich's Robotic Systems Lab on reinforcement learning (RL) for quadrupedal robots. This body of work has produced several landmark publications and represents one of the most successful examples of sim-to-real transfer in legged robotics.
Published in Science Robotics in 2020, this paper from the Robotic Systems Lab presented a controller for blind quadrupedal locomotion trained entirely through reinforcement learning in simulation and then deployed zero-shot on real hardware. The controller used only proprioceptive feedback (joint positions, velocities, and torques) without any vision input. The training process ran inside a physics simulator where the virtual robot was exposed to randomized terrain, physical parameters, and disturbances. Because the simulated conditions were diverse enough, the resulting policy generalized to real-world scenarios it had never encountered during training. The robot was successfully tested on steep mountain trails, creeks with running water, mud, thick vegetation, loose rubble, snow-covered hills, and damp forest floors, all without any fine-tuning on real data.
A follow-up paper published in Science Robotics in 2022 extended the approach to incorporate exteroceptive perception, meaning the robot could now use visual and depth information rather than relying on proprioception alone. The team first trained a teacher policy with access to privileged ground-truth simulation data using RL, then distilled it into a student policy that learned to reconstruct the environment state from noisy real-world sensor observations. This teacher-student distillation framework addressed the sim-to-real gap for perception, which is harder to bridge than the gap for proprioceptive control because real-world cameras and LiDAR produce noisy, incomplete data that differs from clean simulation data. Four ANYmals using this controller explored over 1,700 meters during the DARPA Subterranean Challenge finals, navigating tunnel, urban, and cave courses without a single fall.
Subsequent work from the lab has explored distributional reinforcement learning for risk-aware locomotion, allowing the robot to modulate its behavior based on a risk parameter when climbing steps of varying heights. Rather than always playing it safe or always being aggressive, the robot can adjust its risk tolerance depending on the situation. In 2024, researchers demonstrated ANYmal climbing ladders, a task that requires precise coordination of all four legs on narrow rungs. In September 2025, researchers demonstrated ANYmal D playing badminton using reinforcement learning, showcasing the platform's agility in tasks requiring fast reaction times and dynamic balance. The research pipeline from ETH Zurich continues to feed improvements back into the commercial ANYbotics software stack through the close relationship between the university lab and the company.
ANYmal performs defined inspection missions without human intervention or an internet connection. The robot continuously maps its environment using stereo cameras and LiDAR to build a 3D representation of its surroundings, localizes itself within the map, and plans optimal paths through complex, multi-level facilities. It can traverse industrial stairs, narrow corridors (as narrow as 60 cm), ramps, and uneven surfaces. The software was built on open-source tools including the Grid Map library for managing multi-layer 2D environmental data and the Free Gait framework for defining specific robot motions. Autonomous docking and recharging allow the robot to operate continuously across multiple missions without human involvement.
Launched in February 2025, Data Navigator is ANYbotics' centralized asset management platform. It aggregates inspection data collected by ANYmal, including visual, thermal, acoustic, and gas readings, and provides trend monitoring, anomaly detection, and asset health assessments. The platform supports on-premise, cloud, and air-gapped deployments, and integrates with enterprise systems such as SAP and IBM Maximo. It requires no specialized training for maintenance teams to use. During pilot programs with companies including Grace, Outokumpu, and DSM-Firmenich, Data Navigator analyzed over 2,500 inspections in two weeks, with on-premise integrations becoming operational within hours.
Introduced in June 2025, ANYbotics' Gas Leak and Presence Detection solution combines modular gas detectors with a 360-degree acoustic imaging camera. The acoustic payload can detect a range of common industrial gas leaks (steam, compressed air, toxic gases, hydrocarbons) and identify partial discharge events and mechanical anomalies. Users can independently swap in different gas detector modules for gases including oxygen, hydrocarbons, and ammonia, then calibrate them on-site. A single undetected gas leak can cost a facility over $57,000 annually in addition to safety risks. ANYbotics offers this capability as part of its robotics-as-a-service (RaaS) subscription, with pricing starting below $1,000 per month.
ANYbotics focuses on automating routine inspections in industrial settings where conditions are dangerous, remote, or difficult for human workers. ANYmal performs tasks such as automated readouts of analog instruments (gauges, liquid levels, lever positions), thermography and vibration analysis of rotating equipment, gas concentration monitoring, acoustic anomaly detection, and 3D scanning of infrastructure.
ANYbotics counts major industrial companies among its customers and partners.
| Category | Organizations |
|---|---|
| Oil and Gas | BP, Equinor, Petrobras, PETRONAS, Shell, SLB |
| Chemicals | BASF, DSM-Firmenich |
| Metals and Mining | Outokumpu, Novelis, Vale |
| Energy and Power | Siemens Energy, GE Vernova |
| Technology Partners | AWS, SAP, NVIDIA, Yokogawa |
| Industrial Services | SLB, Equans, Kanadevia Inova |
ANYbotics partners with Zollner Elektronik AG for production of the ANYmal robot. Zollner's facility is located in Hombrechtikon, Switzerland, approximately 30 kilometers from ANYbotics' Zurich headquarters. Zollner handles PCB manufacturing, component pre-assembly, and final robot assembly, managing the entire procurement process for a product with more than 400 components. The modular production setup uses independent production islands that can be scaled individually, allowing capacity to increase beyond 1,000 units per year. Zollner, a global electronics manufacturing services company, was chosen for its ability to handle both low-volume prototype runs and high-volume production scaling.
ANYbotics has raised over $150 million across multiple funding rounds.
| Round | Date | Amount | Lead Investors | Notable Participants |
|---|---|---|---|---|
| Grant | January 2018 | Undisclosed | European Union | EU funding |
| Series A | December 2020 | ~$22 million | Swisscom Ventures | ACE & Company, EquityPitcher Ventures |
| Series B | May 2023 | $50 million | Walden Catalyst, NGP Capital | Bessemer Venture Partners, Aramco Ventures, Swisscom Ventures, Swisscanto, EquityPitcher Ventures, Verve Ventures |
| Series B Extension | December 2024 | $60 million | Qualcomm Ventures, Supernova Invest | Bessemer Venture Partners, NGP Capital, Swisscanto, Swisscom Ventures, TDK Ventures, Walden Catalyst |
| Strategic Investment | September 2025 | Undisclosed | Climate Investment | Joining existing investors including Aramco Ventures, Bessemer, NGP Capital |
The December 2024 extension brought the total Series B to $110 million and was directed at driving expansion into the United States market. Following the September 2025 strategic investment from Climate Investment, ANYbotics' total funding exceeded $150 million. Since closing its initial Series B in May 2023, the company reported that its sales had doubled.
ANYbotics operates through two primary channels. The first is direct robot sales, where customers purchase ANYmal units outright. The second is a robotics-as-a-service (RaaS) model offered through financing partners, providing customers with predictable recurring costs and allowing them to deploy robots without large upfront capital expenditure. CEO Péter Fankhauser has emphasized that ANYbotics' value proposition centers on solving customers' data problems rather than simply selling robotic hardware, stating: "Our clients don't want to buy a robot to play around; they have a data problem."
The company has also signed a pledge against weaponizing its robots, reflecting a commitment to keeping the technology focused on civilian industrial applications.
ANYbotics competes in the growing market for autonomous quadrupedal industrial inspection robots. The quadruped market is currently dominated by Boston Dynamics and Unitree Robotics, with ANYbotics and DEEP Robotics occupying more specialized positions.
| Company | Robot | Headquarters | Key Differentiator |
|---|---|---|---|
| Boston Dynamics | Spot | Waltham, Massachusetts, USA | Largest installed base; broad ecosystem and SDK; priced at ~$74,500 |
| Unitree Robotics | B2, Go2 | Hangzhou, China | Low-cost platforms starting at ~$1,600; strong payload capacity; ROS2 compatible |
| DEEP Robotics | Lynx | Hangzhou, China | Chinese industrial inspection focus |
| ANYbotics | ANYmal D/X | Zurich, Switzerland | Only ATEX/IECEx Zone 1 certified legged robot; vertically integrated inspection software |
ANYbotics differentiates itself through its focus on hazardous industrial environments, particularly with ANYmal X as the only Ex-certified legged robot on the market. The company's vertically integrated approach, controlling both hardware and software for end-to-end inspection workflows, positions it for enterprise deployments where reliability and regulatory compliance are critical. Boston Dynamics' Spot has a larger overall installed base and a broader application ecosystem, but ANYmal targets a more specialized niche in heavy industry. Unitree offers dramatically lower price points, making its robots accessible for education and research, but its software ecosystem is less mature for autonomous industrial inspection. When asked about humanoid robot alternatives, Fankhauser has noted that "a bird and an elephant are good at different things," arguing that form factors should be matched to specific task requirements rather than pursuing a one-size-fits-all approach.
As of late 2025, ANYbotics estimates that its current fleet represents less than 1% of the addressable market, with thousands of industrial plants worldwide that could benefit from autonomous robotic inspection.