Overview
An evidence-based strategic comparison built from approved Research Packs. It explains how companies choose technology routes, commercial entry points, and deployment boundaries. It is not a ranking.
No company in this six-entity set has public evidence sufficient for Scale Deployment.
These six companies are not ranked on one track; they represent different technology, customer, and commercialization boundaries.
A dynamic-control and commercial-robotics incumbent cautiously moving an electric humanoid, Atlas, toward industrial pilots.
An AI-first route combining learned control, data collection, and a manufacturing-scale narrative for general-purpose humanoids.
A vertically integrated, vision/AI and manufacturing-led route using internal factories as the current validation channel.
An accessible hardware and developer/research ecosystem route that broadens humanoid use.
A modular hardware, industrial-partner pilot, and real-world data route for the Apollo family.
A dynamic-mobility, Arc-software, and RoboFab manufacturing route embedding Digit in logistics workflows.
A common factual frame, with route, advantage, and constraint drawn from the Research Pack Bimanual analysis.
| Company | Primary product | Strategic route | Target market | Commercial stage | Key advantage / Main constraint |
|---|---|---|---|---|---|
| Boston DynamicsUnited States · Waltham, Massachusetts · Founded 1992 | Atlas | A dynamic-control and commercial-robotics incumbent cautiously moving an electric humanoid, Atlas, toward industrial pilots. | Industrial manufacturing, logistics, and high-value robotic workflows | Pilot | Long-running depth in dynamic motion, perception, and field robotics. Customer operation, duration, and scale evidence for Atlas remain limited. |
| Figure AIUnited States · San Jose, California · Founded 2022 | Figure 03 | An AI-first route combining learned control, data collection, and a manufacturing-scale narrative for general-purpose humanoids. | Automotive manufacturing, logistics, and general industrial tasks | Prototype | Learned systems such as Helix and BMW-linked industrial evidence form a clear AI-platform route. Figure 03 duration, intervention rate, and completed deployment evidence remain undisclosed. |
| Tesla OptimusUnited States · Austin, Texas · Founded 2003 | Optimus | A vertically integrated, vision/AI and manufacturing-led route using internal factories as the current validation channel. | Tesla internal manufacturing; external industrial customers not established | Prototype | Potential synergy across manufacturing, compute, vision, and data assets. No independently supported external customer Pilot or Deployment evidence. |
| Unitree RoboticsChina · Hangzhou, Zhejiang · Founded 2016 | Unitree H1 | An accessible hardware and developer/research ecosystem route that broadens humanoid use. | Research, education, developers, and early applications | Prototype | Hardware accessibility, cost/open-development signals, and research use. No named customer site or sustained H1 operation is publicly established. |
| ApptronikUnited States · Austin, Texas · Founded 2016 | Apollo 2 | A modular hardware, industrial-partner pilot, and real-world data route for the Apollo family. | Manufacturing, logistics, and industrial material handling | Pilot | A dense partner network connecting product, manufacturing validation, and field pilots. Apollo 2 customer-generation mapping, sites, and sustained-operation data remain limited. |
| Agility RoboticsUnited States · Salem / RoboFab; some historical materials said Corvallis · Founded 2015 | Digit | A dynamic-mobility, Arc-software, and RoboFab manufacturing route embedding Digit in logistics workflows. | Warehouse logistics and manufacturing material handling | Deployment | Field evidence in one GXO workflow, combined with software and manufacturing systems. Autonomy, runtime, intervention, and unit-count data beyond the GXO workflow are not publicly disclosed. |
Product lifecycle status is kept separate from Commercial Maturity; unknown fields stay unknown.
Electric Atlas · Current product
Third generation · Current product
Optimus Gen 1–Gen 2; Gen 3 is the latest public boundary · Prototype / next generation announced
H1 family · Current product / research platform
Second-generation Apollo · Current product
Industrial Digit family · Current family; new version announced
Commercial Maturity is a normalized product-and-evidence result, not a value judgment about a company.
| Product | Commercial stage | Evidence boundary | Customer / Site | Application |
|---|---|---|---|---|
| Atlas Boston Dynamics | Pilot | Hyundai customer-pilot evidence supports Pilot; broad Deployment is not established. | Hyundai Customer pilot / field testing; site boundary is only partly public | Industrial material handling and manufacturing workflows |
| Figure 03 Figure AI | Prototype | Current-generation evidence supports Prototype; Figure 02 BMW records are not copied to Figure 03. | BMW evidence is tied to Figure 02; Figure 03 customer operation is unconfirmed Spartanburg project context; Figure 03 site boundary undisclosed | Manufacturing task demonstrations and industrial projects |
| Optimus Tesla Optimus | Prototype | Internal operation and public demonstrations support Prototype, not external Deployment. | Tesla internal Internal office, lab, and factory contexts | Internal testing and public demonstrations |
| Unitree H1 Unitree Robotics | Prototype | Product channels and academic use support Prototype; availability is not Deployment. | Research institutions and developers; named H1 deployment customers undisclosed Research-use contexts; specific H1 sites undisclosed | Research, teaching, and development |
| Apollo 2 Apptronik | Pilot | Partner pilots and company-reported Apollo 2 operation support Pilot; not Scale Deployment. | Mercedes-Benz, GXO, Jabil, and Apollo 2 Robot Park context Customer/partner sites; operating details partly undisclosed | Manufacturing, logistics, and material-handling pilots |
| Digit Agility Robotics | Deployment | GXO live-site evidence and the 100,000-tote milestone support bounded Deployment, not Scale Deployment. | GXO (one confirmed workflow) Flowery Branch, Georgia | Warehouse tote handling |
Eight dimensions are carried from the approved Research Packs; undisclosed areas are not filled by inference.
Dynamic electric bipedal humanoid
ConfirmedWhole-body dynamic balance and agile gait
ConfirmedWhole-body bimanual manipulation; generalization boundary undisclosed
ReportedVision and proprioception; full architecture not disclosed
ReportedControlled autonomy and human collaboration; scaled autonomy unknown
ReportedControl and learned behaviors; method details limited
ReportedLong-running robotics data and partner-data route
ReportedSimulation used for control validation; coverage not fully disclosed
ReportedStandard bipedal humanoid platform
ConfirmedBipedal walking and whole-body coordination
ConfirmedLearned bimanual manipulation; generalization boundary unknown
ReportedVision-led multimodal perception
ReportedHierarchical VLA/control stack; sustained autonomy undisclosed
ReportedLayered and end-to-end learned control
ReportedReal-task data collection and industrial data loop
ReportedSimulation and real-world data used together; details partly undisclosed
ReportedStandard bipedal humanoid
ConfirmedBipedal walking and balance
ConfirmedHand and whole-body manipulation demonstrations; generalization unknown
ReportedVision-led; sensor details partly disclosed
ReportedTesla internal autonomy narrative; sustained-operation boundary unknown
ReportedLearned control and large-scale AI infrastructure route
ReportedInternal operation/video data and potential Tesla-ecosystem reuse
ReportedTraining and validation infrastructure not fully disclosed
ReportedFull-size bipedal humanoid
ConfirmedBipedal walking and dynamic balance
ConfirmedBimanual/whole-body manipulation; customer generalization unknown
ReportedDepth/LiDAR/vision sensor combination
ReportedResearch/developer control stack; general autonomy unknown
ReportedImitation and reinforcement learning are publicly discussed
ReportedResearch-task data and developer ecosystem
ReportedAcademic sim-to-real and simulation support; scope not fully disclosed
ReportedModular bipedal humanoid for industrial workflows
ConfirmedBipedal mobility and load handling
ConfirmedForce-controlled industrial manipulation
ReportedMultimodal vision/force sensing; details limited
ReportedPilot autonomy; site-operation boundary unknown
ReportedArtemis/partner AI and real-task learning
ReportedPartner-site and Robot Park data collection
ReportedSimulation/digital environments with real-world data; details incomplete
ReportedBipedal humanoid for industrial material handling
ConfirmedDynamic bipedal mobility
ConfirmedIndustrial tote grasping and handling
ReportedVision and proprioception; full stack undisclosed
ReportedArc and task autonomy on site; general autonomy unknown
ReportedReinforcement learning/whole-body control and demonstrations
ReportedWorkflow data and Arc operations data
ReportedSimulation/RL for motion and task training; details limited
ReportedThe following passages are Bimanual comparative analysis, not company statements.
Premium robotics hardware, services, and workflow integration, with Spot and Stretch as commercial reference points.
A route centered on engineering maturity and industrial reliability rather than the fastest AI-platform expansion.
It connects humanoid research demonstrations to the operating constraints of an established industrial robotics business.
A combination of hardware, AI/control platform, and customer deployment model; profitability and fleet continuity are not established.
Connects an AI-first data loop to industrial partnerships, while current product evidence remains Prototype.
Represents competition around sustained data and model iteration beyond one-off humanoid demonstrations.
A long-horizon hardware/software option inside Tesla; external revenue model remains unclear.
The strongest manufacturing option value, but public commercial evidence remains Prototype.
Tests whether a large manufacturing company can absorb humanoids into its own production system.
Direct hardware sales and a developer/research ecosystem; service and fleet economics undisclosed.
Differentiates through accessibility and iteration rather than capital-intensive industrial deployment.
Moves humanoid hardware from a few company labs into a broader research and developer network.
Hardware, partner integration, pilots, manufacturing collaboration, and future deployment services.
Sits between industrial partner pilots and manufacturing scale, with evidence closer to field use than a pure prototype.
Shows how humanoids can enter workflows through manufacturers, logistics operators, and integrators.
Digit hardware, Arc software, RaaS/orders, workflow integration, and manufacturing.
Has the clearest bounded Deployment evidence in this set, limited to the GXO workflow.
Makes deployment legible as a task, site, and operating record rather than a corporate-scale slogan.
Research Packs, Claim Ledgers, and Source Libraries form the evidence layer. A commercial stage moves up only when product, customer, site, and operating evidence align; analysis remains visibly separate from company facts.
This MVP includes six core entities that passed the Landscape release gate. Commercial maturity is normalized at product and evidence level; a demo, pilot, product availability, manufacturing capacity, or plan does not automatically become Deployment.
Source links come from the current CMS Source entities and represent the public evidence used for this comparison.
Hyundai customer-pilot evidence supports Pilot; broad Deployment is not established.
Current-generation evidence supports Prototype; Figure 02 BMW records are not copied to Figure 03.
Internal operation and public demonstrations support Prototype, not external Deployment.
Product channels and academic use support Prototype; availability is not Deployment.
Partner pilots and company-reported Apollo 2 operation support Pilot; not Scale Deployment.
GXO live-site evidence and the 100,000-tote milestone support bounded Deployment, not Scale Deployment.
Continue from the Landscape into maintained topic knowledge pages.