Humanoid robotics is usually presented as a competition among companies building complete robots. The current Humanoid Analytics Company Tracker shows a broader commercial contest.
The tracker contains 225 Core companies, covering humanoid manufacturers and substantial corporate humanoid programs, and 126 Adjacent companies supplying or developing dexterous hands, tactile sensors, components, AI systems, data platforms, software, and integration services. Together, those groups account for 351 in-scope entities reviewed between July 13 and July 14, 2026.
The distinction matters because building the complete robot may not be the only, or the earliest, route to commercial value. Core companies carry the difficult task of integrating mobility, manipulation, perception, autonomy, safety, manufacturing, maintenance, and customer support into one system. Adjacent companies can focus on one bottleneck and potentially sell across several robot platforms.
That does not mean Adjacent companies are commercially proven. Product pages, technical demonstrations, ecosystem positioning, and supplier announcements do not establish humanoid-related revenue or repeat demand. Humanoid Analytics does not treat funding, demos, partnerships, shipment claims, or future production targets as commercial proof without stronger customer and operating evidence.
The Humanoid Market Is Larger Than Its Robot Brands
The most visible names in humanoid robotics are generally Core companies. They design the body, select or develop the components, train the control systems, manufacture the hardware, sell the platform, and remain responsible when it fails in a customer environment.
That full-stack position offers the greatest potential control over pricing, data, customer relationships, and long-term platform economics. It also creates the greatest integration burden.
A humanoid OEM must solve multiple difficult problems at the same time. A robot that walks well but cannot manipulate variable objects has limited utility. A capable hand is not enough if perception fails, teleoperation costs remain high, batteries interrupt the workflow, or repairs require specialist technicians. Even a technically strong platform may struggle commercially if a wheeled mobile manipulator, conventional robot arm, conveyor, or process redesign can perform the task more reliably and cheaply.
Adjacent companies face a different calculation. They do not necessarily need one humanoid platform to dominate. A hand, sensor, actuator, robot foundation model, simulation environment, or integration layer may be useful across multiple OEMs and other forms of robotics.
The current tracker makes the structure visible.
| Market Layer or Segment | Tracked Entities | Share of Core and Adjacent Universe | Central Commercial Test |
|---|---|---|---|
| Core humanoid builders and corporate programs | 225 | 64.1% | Can a complete robot deliver reliable, safe, economical customer work? |
| Dexterous manipulation | 58 | 16.5% | Can the hand manipulate varied objects repeatedly without excessive cost or failure? |
| Tactile and sensing | 13 | 3.7% | Does sensing materially improve grasping, control, safety, or recovery in real workflows? |
| Humanoid components | 11 | 3.1% | Can components meet OEM requirements for cost, durability, supply, and production volume? |
| Humanoid platforms and data | 11 | 3.1% | Can data collection and training infrastructure produce transferable operating capability? |
| Humanoid platforms and AI | 6 | 1.7% | Can software generalize across tasks, environments, and robot bodies? |
| System integration | 5 | 1.4% | Can robots be installed, supported, and connected to customer workflows economically? |
| Other Adjacent segments | 22 | 6.3% | Is the humanoid connection specific, repeatable, and commercially material? |
Figures are calculated from the Humanoid Analytics Company Tracker. Six ecosystem organizations and two excluded records are not included.
The largest Adjacent category is dexterous manipulation, with 58 tracked companies. It represents 46% of the Adjacent group.
That concentration can be interpreted in two ways. It may show that the industry recognizes hands and manipulation as a central bottleneck. It may also show crowding, with many suppliers competing before humanoid OEM volumes and technical requirements have stabilized.
The tracker alone cannot determine which interpretation will prove correct.
The Picks-and-Shovels Thesis Needs Customer Evidence
Some Adjacent offerings are concrete products rather than distant concepts.
SCHUNK describes its SVH as a series-ready, five-finger servo-electric gripping hand with nine drives and electronics integrated into the wrist. That establishes product availability and an industrial supplier context. It does not establish how widely the hand is used in commercial humanoid deployments or what share of SCHUNK’s business is connected to humanoids.
Touchlab is developing customizable tactile sensing and electronic skin intended to give robots richer contact information. The technical proposition is commercially relevant because manipulation in uncontrolled environments depends on detecting contact, force, slip, and object variation. The available company material, however, does not provide comparable public evidence of humanoid-related revenue, fleet deployment, or repeat production orders.
The software and AI layer presents a similar opportunity with different risks. Physical Intelligence describes general-purpose robotic foundation models as a reusable intelligence layer that could reduce the cost and effort required to build robotic applications. Its work spans data from different robots and task types, supporting the idea that intelligence may transfer across hardware platforms. These remain company-described technical results rather than proof of broad commercial adoption.
Skild AI likewise describes an “omni-bodied” model intended to work across humanoids, quadrupeds, tabletop systems, and mobile manipulators. If cross-platform models perform reliably, an AI supplier could participate in several robotics markets rather than depend on one humanoid design. But the decisive commercial evidence would be named deployments, payment, measurable performance, renewal, and continued use by robot manufacturers or operators.
These examples illustrate why “Adjacent” should not be treated as one maturity class. It includes established industrial suppliers, early-stage technology companies, research-oriented developers, software platforms, component specialists, and entities whose humanoid connection remains limited or insufficiently verified.
The label identifies a market relationship. It does not prove commercial readiness.
Core Companies Carry More Upside and More Integration Risk
Core companies have an important strategic advantage: they control the final system and can potentially own the customer relationship, robot data, service contract, and application layer.
They also absorb failures between layers.
A customer does not purchase locomotion, tactile sensing, a foundation model, and a hand as separate technical achievements. The customer purchases a working outcome. When the robot drops an object, stops unexpectedly, requires repeated human intervention, or cannot integrate with the workflow, the Core company is responsible regardless of which component caused the problem.
That makes Core companies dependent on the quality and availability of their Adjacent ecosystem, even when they develop critical components internally.
Vertical integration may reduce dependency and improve hardware-software coordination. It can also increase capital requirements, manufacturing complexity, and development time. External sourcing may accelerate development, but it creates supplier risk, interface constraints, and the possibility that competitors can purchase similar technology.
The likely market structure is therefore not a simple choice between vertical integration and external suppliers. Core companies may internalize the components they consider strategically differentiating while purchasing standardized or mature technologies elsewhere.
The commercial question is which layers become standardized and which remain proprietary.
If hands, actuators, compute, and tactile sensors become interchangeable, value may migrate toward software, deployment data, customer integration, and service. If hardware remains highly specialized to each robot, the largest Core companies may need deeper vertical integration, limiting the addressable market for independent suppliers.
Public evidence is not yet sufficient to resolve that question.
Geography Shows Density, Not Leadership
The tracker is heavily weighted toward Asia. Of the 351 Core and Adjacent entities, 251 are based in Asia, including 169 Core and 82 Adjacent companies. China alone accounts for 144 Core and 71 Adjacent records.
North America contains 34 Core and 21 Adjacent companies, while Europe contains 19 in each group.
These counts show where publicly visible company formation and supplier activity are concentrated. They do not measure revenue, deployment maturity, technical performance, production quality, or customer adoption.
The geographic distribution is still commercially relevant. A dense regional base of OEMs, component companies, manufacturers, research institutions, and integrators can shorten development cycles and support domestic supply chains. But numerical density becomes meaningful only when it produces qualified components, repeatable manufacturing, named customers, and operating robots.
The same caution applies to changes in tracker size. A June 2026 Humanoid Analytics article referred to 121 Core builders or programs and 35 adjacent suppliers. The current tracker is substantially larger. Without a reconciled record separating newly formed companies, expanded research, reclassification, and previously missing entities, the increase should be interpreted as broader tracker coverage, not as proof that the market itself more than doubled.
What Evidence Should Appear Next
For Core companies, the next meaningful evidence remains customer operation: named users, active tasks, duration, payment, fleet size, human intervention, uptime, throughput, safety performance, service requirements, and repeat deployment.
For Adjacent companies, the evidence test is different but equally demanding.
The strongest signals would include named humanoid OEM customers, production contracts, components used in delivered robots, repeat orders, revenue attributable to humanoid robotics, qualification across multiple platforms, field-performance data, and evidence that the supplier remains necessary as OEM designs mature.
Cross-platform adoption would be particularly important. An Adjacent company supplying several unrelated OEMs would have stronger evidence of product relevance and lower dependence on one robot program. A supplier attached to a single prototype or memorandum would remain exposed to delays, redesigns, insourcing, and customer failure.
The Core and Adjacent distinction should therefore be used as a map of dependency and value capture, not as a ranking.
Core companies offer the clearest route to controlling the complete humanoid platform, but they carry the combined risks of hardware, software, manufacturing, deployment, and service. Adjacent companies may reach revenue sooner and spread risk across several platforms, but only if their technology becomes necessary, qualified, and repeatedly purchased.
The next phase of humanoid robotics will not be decided only by which robot receives the most attention. It will also be decided by which suppliers become embedded in working machines, which technologies survive integration, and which companies can show that customers return after the first trial.
Sources:
Humanoid Analytics, “Humanoid Company Tracker”: https://humanoidanalytics.com/humanoid-company-tracker/
SCHUNK, “SVH 5-Finger Servo-Electric Gripping Hand”:
https://schunk.com/de/en/gripping-systems/special-gripper/svh/c/PGR_3161
Touchlab, “Giving Touch to Robots”:
https://www.touchlab.io/
Touchlab, “Touchlab & ARIA: Unlocking Robot Dexterity With E-Skin”:
https://www.touchlab.io/post/touchlab-aria-robot-dexterity
Physical Intelligence, “The Physical Intelligence Layer”:
https://www.pi.website/blog/partner
Physical Intelligence, “π0.7: A Steerable Model With Emergent Capabilities”:
https://www.pi.website/blog/pi07
Skild AI, “Building the General-Purpose Robotic Brain”:
https://www.skild.ai/blogs/building-the-general-purpose-robotic-brain
Skild AI, “The Reindustrial Revolution: Partnering With ABB Robotics, Universal Robots, and NVIDIA”:
https://www.skild.ai/blogs/reindustrial-revolution