The Robot Economy: A Map of the Companies Building Physical Automation

People often picture the robot economy as a humanoid walking through the front door. The real expansion is broader. Robots are moving goods in warehouses, inspecting power stations, assisting surgeons, printing construction layouts, removing weeds, delivering meals and collecting data in places that are expensive or dangerous for people to reach.

The industry now contains several different businesses. Some companies build complete machines. Others build hands, motors, sensors, training data, simulation tools or software for managing fleets. A warehouse robot and a surgical robot share some technology, although their customers, regulation and economics are very different.

This map follows the money and the work. It names the companies attempting to turn physical automation into a business and places them in the part of the stack where they appear strongest.

Several forces have arrived at the same time. AI models can interpret images and instructions more effectively. Motors, batteries, sensors and computing have improved. Manufacturers have learned to produce many of these components at lower cost. Large pools of venture and corporate capital are willing to finance the difficult move from prototype to production.

Demography adds urgency. China, Japan, South Korea and much of Europe are aging. Their working-age populations are shrinking or growing slowly, while demand for manufacturing, logistics, healthcare and eldercare continues. Immigration can relieve some pressure in some countries. It does not remove the need to produce more with each worker.

This helps explain why the current robotics cycle is broader than the industrial automation of ten years ago. Buyers are looking beyond car factories toward warehouses, farms, hospitals, construction sites and care. The strongest demand appears where labour is scarce, dangerous, repetitive or difficult to retain. Demography creates the need; better AI and cheaper hardware make more of the response technically possible.

There is no single total for robotics. Research firms draw the boundary in different places. Some include autonomous vehicles and military drones. Others focus on industrial and service robots. The treatment of diversified groups such as Tesla, Alphabet, Nvidia, ABB and Hyundai changes the answer by hundreds of billions of dollars.

The annual flow of investment remains far below $1 trillion. The equity value of a reasonably defined robotics universe has moved beyond half a trillion dollars. Broader totals can exceed $1 trillion after diversified technology and automotive groups are included, though most of their value comes from businesses outside robotics.

Funding estimates also differ because the boundaries differ. Crunchbase counted $15 billion in 2025. CB Insights counted $40.7 billion. F-Prime counted $20.4 billion across the US, Europe and Israel. The range is more honest than a false point estimate.

A robot needs to perceive a scene, understand an instruction, choose an action and recover from mistakes. This has created a new group of companies building general robot intelligence. Physical Intelligence, Skild AI, Generalist, Dyna Robotics, FieldAI and Sunday Robotics are pursuing models that can learn tasks across machines or environments.

The next layer handles robot data and operations. Foxglove and Rerun help engineers inspect complex sensor data. Formant and InOrbit manage deployed fleets. Wandelbots makes industrial robots easier to program. Sereact, Covariant, Nomagic and Mujin concentrate on intelligent manipulation and warehouse work.

Noematrix belongs in this intelligence layer. It combines embodied models, physical data and force feedback. Anyverse Dynamics spans two layers: its MWA world model acts as the brain, while its K15 platform gives the model a body. Both illustrate the growing contest over the software that may control many types of robot.

Humanoids receive the most attention because factories, warehouses and homes were designed around the human body. A machine with two arms, legs or wheels and human-scale reach can use existing doors, shelves, tools and workstations.

The North American and European field includes Figure AI, Agility Robotics, Apptronik, 1X, Sanctuary AI, Boston Dynamics, NEURA Robotics, Mentee Robotics, PAL Robotics, Enchanted Tools and Pollen Robotics.

China and East Asia contain another large group: Unitree Robotics, AGIBOT, UBTECH, Fourier Intelligence, EngineAI, LimX Dynamics, Leju Robot, RobotEra, Galbot, Astribot, XPeng Robotics and Rainbow Robotics.

Agile Robots sits between industrial automation and the general-purpose field. It began with intelligent arms and an operating system and later added a humanoid platform. AGIBOT is a full-stack humanoid company, with its own machines, joints, hands and embodied software. Anyverse Dynamics combines its general robot brain with a mobile dual-arm machine.

China’s concentration of humanoid developers is deliberate. It sits on top of the world’s largest industrial-robot market and a dense supply chain for electric vehicles, batteries, motors, electronics and precision manufacturing. In 2025, China installed 354,000 industrial robots, 59 percent of the global total. That home market gives robot makers suppliers, engineers, factories and potential customers within the same industrial system.

Demography gives the programme a long-term purpose. China’s working-age population has been declining since 2012, while the share of people aged 60 and above reached 23 percent in 2025. Robots are expected to support production, logistics, agriculture and care as the labour pool ages. The policy also serves other goals: moving Chinese manufacturing up the value chain, reducing dependence on foreign technology and creating another export industry.

The party-state is managing the push through several layers. Made in China 2025 named robotics as a strategic industry. The 2023 humanoid-robot guidance set targets for innovation, mass production and practical deployment. The 15th Five-Year Plan places embodied intelligence inside the next industrial programme. Beijing and Shanghai have built public innovation centres and shared platforms such as Tiangong, Huisi Kaiwu and OpenLoong. By June 2026, more than 70 embodied-AI training grounds were operating, with over 40 more being built or planned.

This system funds both supply and demand. It supports components, models and manufacturers while organising test sites, standards, training data and early customers. That can compress years of development into a shorter cycle. It can also create duplication, weak projects and overcapacity. China’s large company count is therefore a sign of industrial depth and policy ambition, not proof that every developer will survive.

Useful work depends on contact. Grasping a soft bag, inserting a connector and lifting a metal casting require different combinations of force, touch and control.

Specialists include Shadow Robot, Tacta Systems, Touchlab, PSYONIC, Tesollo, Robotiq, OnRobot and GelSight. AGILINK, spun out of AGIBOT's hardware operation, focuses on dexterous hands and robotic end effectors.

On the factory floor, Standard Bots, Flexiv, Franka Robotics, Universal Robots, Doosan Robotics, Vention and Formic sell arms, automation systems or capacity as a service. Path Robotics automates welding. GrayMatter Robotics tackles surface finishing. Machina Labs uses robots for flexible metal forming.

Warehouses provide structured space, measurable tasks and customers that understand the cost of every movement. This has made logistics one of the most mature commercial markets.

Storage and movement are the established layer. AutoStore and Symbotic redesign the storage system itself. Exotec, Locus Robotics, Geek+, GreyOrange, HAI Robotics, Quicktron, Addverb, Brightpick, Seegrid, OTTO Motors and MiR move goods through warehouses and factories.

Picking and unloading are harder. The robot must recognise and grasp objects that vary in shape, weight and position. Dexterity, RightHand Robotics, Ambi Robotics, Plus One Robotics, Nimble, OSARO and Pickle Robot work on this less mature layer.

Beyond the warehouse, Starship Technologies, Serve Robotics, Coco Robotics and Cartken work on pavement delivery. Nuro and Gatik use road vehicles. Zipline, Wing, Manna and Flytrex take delivery into the air.

Construction sites change every day and contain dust, uneven ground and many independent contractors. This favours machines with a narrow, measurable job. Heavy equipment is the first group: Built Robotics automates machinery and solar piling, while Bedrock Robotics works on autonomous earthmoving. Layout and finishing include Dusty Robotics, Canvas, PaintJet and Civ Robotics. Their value can be measured against a known crew, schedule and rework cost.

Robotic fabrication aims at a larger change in how structures are produced. ICON and COBOD print building structures. Monumental works on masonry and AUAR on automated building production. RoboticPlus.AI combines construction robots, control software and fabrication for highly customised structures.

Agriculture splits in a similar way. Precision intervention companies such as Carbon Robotics, FarmWise, Naïo Technologies, Ecorobotix, Verdant Robotics, Aigen and FarmDroid target individual weeds or plants. Mobility and harvesting are pursued by Burro, Bonsai Robotics, Tortuga AgTech, Tevel and Guardian Agriculture. Crop type and seasonality make this a collection of specialised markets rather than one agricultural-robot market.

Food preparation has clearer indoor conditions but difficult margins. Miso Robotics, Chef Robotics, Hyphen and Picnic Works must save enough labour and waste to justify equipment, cleaning and service.

Medical robotics has a longer commercial history and a harder route to market. Regulation, clinical evidence and surgeon training matter as much as mechanical performance.

Intuitive Surgical remains the benchmark because its advantage extends beyond the machine to instruments, surgeon training, clinical acceptance and a large installed base. Newer systems must find a different procedure, price point or workflow. The field includes CMR Surgical, Moon Surgical, Medical Microinstruments, Distalmotion, Noah Medical, PROCEPT BioRobotics, Vicarious Surgical and EndoQuest Robotics.

Mobility and rehabilitation are addressed by Wandercraft, Ekso Bionics and Lifeward. Hospitals use mobile service robots from Diligent Robotics, Relay Robotics and Aethon.

Restaurants, hotels and public buildings provide a wider service market. Bear Robotics, Pudu Robotics and Richtech Robotics sell serving, delivery and cleaning machines. Their progress may say more about near-term adoption than a spectacular humanoid demonstration.

Many strong robot businesses begin where sending a person is costly, slow or dangerous.

ANYbotics sends legged robots through industrial plants. Gecko Robotics inspects critical infrastructure. Flyability builds protected drones for confined spaces. Percepto, Voliro and Exyn Technologies work on autonomous inspection and mapping.

In defence and security, the active field includes Anduril, Shield AI, Skydio, Ghost Robotics, Overland AI, Forterra, Saronic and Saildrone. Nauticus Robotics and Bedrock Ocean Exploration take autonomous systems underwater.

The same logic reaches space through Astrobotic, ispace, Lunar Outpost and GITAI. Remote operation, autonomy and durable machinery connect these companies to the wider robotics stack.

Entertainment is a smaller commercial segment with a different measure of success. The machine must move safely around guests, express personality and remain consistent with a story or character.

Dynamic Creatures is building interactive character robots for entertainment and hospitality. Engineered Arts, Furhat Robotics, Hanson Robotics, AVATARIN and Telexistence explore social presence, remote embodiment and approachable machines.

This category connects robotics with animation, games, live events and location-based attractions. Revenue may come from leasing, ticket sales, branded experiences or remote services. It also offers a public testing ground for human-robot interaction before similar machines enter homes and care settings.

These six companies are not presented as the six largest or most investable businesses in robotics. MidLincoln selected them for closer tracking because, taken together, they form a compact version of the physical-automation stack: robot intelligence, dexterous components, integrated bodies and a specialised construction application. The group provides a practical lens through which to compare where technical and commercial value may sit.

The selection is illustrative rather than a ranking. Its value lies in the contrast. A component supplier can sell across many robot brands. A model developer can try to control many kinds of body. An integrated manufacturer can optimise the whole machine. A vertical specialist can solve one customer problem more quickly. The same strategic choices appear across the global market.

The central strategic question is not simply which robot will sell. It is which layer of the system will retain pricing power once robots become more common.

Complete robot makers want to own hardware, software and fleet data. Model companies want their intelligence to work across many bodies. Component companies want to become standard suppliers. Integrators want to control the customer relationship and sell completed work rather than machinery.

Value may move as the industry matures. Early in a market, scarce components and engineering talent can command high margins. As hardware becomes more standard, software, data and distribution may matter more. In regulated or highly specialised sectors, the strongest defence may remain certification, workflow knowledge and trust rather than a general robot brain.

There may be no single winner across the whole stack. Warehouses, operating rooms, farms and construction sites have different buyers and failure costs. The valuable layer is likely to be the one that controls the hardest bottleneck in each market.

Robotics produces impressive videos. Investment returns come from machines that work for paying customers.

• Paid deployments: customer sites and recurring use carry more weight than announced partnerships.
• Utilisation: hours worked per day and tasks completed per hour show whether a robot has entered the workflow.
• Human intervention: frequent remote rescue can hide weak autonomy and expensive service labour.
• Unit economics: hardware cost, installation, maintenance and useful life determine the customer's payback.
• Gross margin: a growing fleet can destroy cash when every additional unit requires heavy engineering support.
• Manufacturing: prototypes and mass-produced machines are different businesses.
• Data: deployed fleets should generate information that improves reliability and reduces intervention.
• Customer concentration: one large pilot can create a misleading picture of commercial demand.
• Safety and regulation: medical, road-going, domestic and defence robots face very different approval and liability regimes.

These measures also explain why specialised robots may create strong businesses sooner than general humanoids. A narrow task gives the buyer a clear comparison between the cost of the machine and the cost of the existing process.

The robot economy will probably feel gradual from street level. Warehouses will add more mobile machines. Farms will automate one crop and one task at a time. Hospitals will add transport and surgical systems. Inspectors will send robots into pipes, tanks, mines and power plants. Entertainment venues will use machines as characters and hosts.

The industry already supports thousands of companies and attracts tens of billions of dollars each year. Its tracked equity value can be measured in hundreds of billions. The next stage depends on ordinary operating facts: reliability, useful hours, service cost, manufacturing quality and customer payback.

The robots are arriving through many doors. The most important ones may look too specialised to appear in a science-fiction film.

Market totals draw on the IEEE Robotics & Automation Society global funding report, F-Prime's State of Robotics 2026, Crunchbase robotics funding data, CB Insights' State of Venture 2025 and the International Federation of Robotics. Company descriptions are based on company materials and public announcements reviewed in October 2026.

China figures and policy context draw on the National Bureau of Statistics of China, the International Federation of Robotics’ 2026 figures, MIIT’s humanoid-robot guidance and the IEEE Robotics and Automation Society’s China roadmap.

Private-company valuations, order books and production figures are usually reported by companies or their investors. They should be treated as indications rather than audited market facts. Inclusion in this landscape is descriptive and does not constitute an investment recommendation.