PM Portfolio Manager Reports

From Portfolio Manager reports

Humanoid Robotics

Humanoid robotics becomes investable when trusted-paper evidence turns into paid fleet evidence: robots must handle human-designed factory, warehouse, service, or household tasks with disclosed uptime, intervention rates, task success, safety cases, gross margin, and repeat orders. The papers make the opportunity visible; this page tracks the bottlenecks that decide whether pilots become fleets and whether public proxies can convert that adoption into revenue, margin, cash flow, or source-backed disclosure.

Mental model: humanoids are a pilot-to-fleet conversion chain. A manual task has to become data, control, movement, uptime, and cost proof before the theme reaches public-company economics.

  1. Task Labor gap defines the job Factories, warehouses, logistics sites, and service operators identify manual work that is hard to staff or automate with fixed equipment. Proof: budgeted task, wage benchmark, vacancy evidence, or throughput pain.
  2. Form Human space stays unchanged The humanoid body matters when two arms, hands, reach, stairs, carts, doors, shelves, and existing tools avoid facility redesign. Proof: robot uses current workflow without a custom cell.
  3. Pilot Real sites create data Small deployments collect video, tactile, force, failure, recovery, and operator-intervention logs that simulation cannot fully replace. Proof: paid pilots, logged hours, task diversity, and repeat customers.
  4. Control Software closes the loop VLA models, low-level control, simulation, teleoperation, and exception handling turn instructions and sensor data into safe motion. Proof: lower intervention rate and higher task completion.
  5. Body Hardware survives work Actuators, hands, sensors, compute, batteries, wiring, connectors, and structures must move precisely without failing or overheating. Proof: service life, repair time, runtime, and warranty data.
  6. Fleet Unit economics decide scale A robot must lower cost per pick, load, shift, task, or site while creating repeat orders and supplier shipments. Proof: fleet expansion, gross margin, service attach, and cash conversion.
Source boundary: this paper-led readiness page uses the local Citrini humanoid primer and Roland Berger 2026 humanoid report. Durable company, sector, and macro claims still route to Physical AI And Defense Autonomy, Robotics, and the selected knowledge lanes in the Source Trail.

Current read

The two papers agree that humanoids are moving from promotional demos toward controlled industrial pilots, but the investable test is still paid deployment quality. Treat the large market scenarios as context until real fleets show task success, uptime, safety, maintenance cost, and repeat orders.

  • Industrial and logistics sites come first: factories, warehouses, machine loading, kitting, shelf picking, and internal transport are the nearest evidence window because tasks are bounded and labor substitution is measurable.
  • Software and data are the gating layer: fleet logs, teleoperation records, failure recovery, and sim-to-real validation decide whether robots become more reliable after each deployment.
  • Actuators and hands carry the hardware proof burden: torque density, reducer cost, thermal behavior, tactile sensing, hand lifespan, cable routing, and serviceability decide whether a humanoid can work useful shifts.
  • Public exposure is mostly proxy exposure: leading humanoid OEMs are private, China-listed, or buried inside larger companies, so public baskets mainly route through components, automation, semiconductors, sensing, integration, and speculative service robotics.
  • Do not confuse buyer validation with supplier economics: AMZN, GXO, TSLA, BMW, and similar operators are important proof sites, but supplier revenue needs orders, shipments, service attach, margin, and cash flow.

Value Chain Map

Card returns load from the local report API route /api/themes/humanoid-robotics/node-return-buckets?as_of=latest. The proxy baskets are mapped from existing robotics and Physical AI coverage; they are timing and provenance context, not proof that humanoid revenue is already material.

1

Motion, Actuators, Hands, And Precision Components core hardware bottleneck

Joint motion, reducers, drives, bearings, hands, and precision parts decide whether the robot can move safely at industrial cost.

Role in theme

Controls the cost, torque, durability, dexterity, heat, vibration, and repair profile that turns a humanoid body into usable labor.

What this is

Integrated actuators, motors, reducers, drives, torque sensors, bearings, pneumatic or hydraulic components, dispensing, testing, hands, grippers, and tactile modules.

Economic lever

Robot unit ramps become component content, co-development awards, replacement parts, service demand, factory utilization, price-cost spread, and margin.

Watch items

Actuator cost reduction, torque density, hand service life, warranty claims, reducer lead times, thermal limits, customer qualification, and disclosed robot content.

Proxy return basket: RRX, TKR, NOVT, AME, PH, NDSN, IEX, ITT, DOV, FTV

Filed or source-only monitors without core chart promotion: Nabtesco, Harmonic Drive, SMC, Yaskawa, Fanuc, THK, Nidec, Schunk, Robotiq, OnRobot, Shadow Robot.

2

Training Data, Simulation, Teleoperation, And VLA Control software readiness gate

Fleet logs, teleoperation, simulation, and robot-control models decide whether pilots improve or stay stuck in demos.

Role in theme

Turns real-world sensor, action, failure, and recovery data into task policies that lower intervention rates and expand the work envelope.

What this is

VLA and VLM stacks, robot foundation models, simulation pipelines, synthetic data, teleoperation consoles, fleet logging, task planning, and low-level control.

Economic lever

Useful software becomes robot subscription value, platform lock-in, lower support cost, more deployed tasks, higher utilization, and better gross margin.

Watch items

Logged operating hours, task success, intervention rate, sim-to-real transfer, operator-to-robot ratio, model latency, and named software partnerships.

Proxy return basket: NVDA, TSLA, SERV, TER, AMZN

Caveat: this is a proxy-only basket because most humanoid control stacks and fleet logs are private or embedded inside larger platforms. Open the ranked node page.

3

Edge Compute, Power, Connectivity, And Robot Semiconductors robot nervous system

On-device compute, analog, power, safety, connectivity, and motor-control silicon let robots sense, decide, and move locally.

Role in theme

Supplies low-latency inference, motor control, power management, sensing interfaces, connectivity, safety circuits, and thermal-aware compute inside the robot.

What this is

Edge AI modules, GPUs, embedded processors, analog chips, power semiconductors, motor drivers, connectivity silicon, and industrial safety controllers.

Economic lever

Design wins convert into content per robot, module shipments, software ecosystem attachment, inventory recovery, gross margin, and longer platform relationships.

Watch items

Robot design wins, disclosed content per unit, industrial channel inventory, edge module shipments, power draw, heat, latency, and data-center attribution risk.

Ranked return basket: NVDA, TXN, NXPI, QCOM, MPWR, ADI, ON, MCHP, STM, AMBA, AMD, AVGO

Direct core: NVDA, TXN, NXPI, QCOM. Proof-heavy direct names: MPWR, ADI, ON, MCHP, STM. Specialist: AMBA. Adjacent AI infrastructure monitors: AMD, AVGO. Source-only watches: Infineon, Renesas.

Open the ranked node page.

4

Perception, Tactile Sensing, And Localization sensing proof

Machine vision and scaled 3D sensing have the clearest public evidence; tactile and e-skin remain the source gap.

Role in theme

Lets a humanoid understand objects, people, force, pose, space, tools, shelves, flexible parts, and task completion in real time.

What this is

Cameras, machine vision, lidar, radar, sonar, IMUs, tactile sensors, e-skin, localization software, inspection systems, and safety sensing.

Economic lever

Robot adoption becomes sensor shipments, design wins, software attach, product gross margin, production SOPs, and replacement demand.

Watch items

Production design wins, tactile sensor integration, sensor ASPs, safety certification, customer breadth, gross margin, and cash burn for thin-conversion names.

Ranked return basket: CGNX, HSAI, TDY, OUST, AMBA, CODA, AEVA, LINK, INVZ, ARBE, ODYS, MSAI, MVIS, LIDR

Core: CGNX, HSAI, TDY, OUST, AMBA. Specialty watch: CODA. Tactile watch: LINK. Source-only monitors: Keyence, Omron, SICK, Hikrobot, tactile-sensor specialists.

Open the ranked node page.

5

Brownfield Factory And Warehouse Deployment first paid sites

Controlled factories and warehouses are the first place to test whether humanoids beat fixed automation on variable manual work.

Role in theme

Converts the humanoid thesis into site economics through machine loading, conveyor loading, kitting, shelf picking, simple transport, and integration.

What this is

Factory automation, warehouse automation, controls, machine vision, robot-cell equipment, test automation, logistics software, integration, and support.

Economic lever

Paid pilots become orders, site expansions, RPO conversion, service revenue, software attach, cost-per-task savings, margin, and cash conversion.

Watch items

10-to-50 unit pilots, repeat sites, cost per pick or load, uptime, integration time, safety incidents, customer concentration, and project margin.

Ranked supplier basket: SYM, ROK, CGNX, ZBRA, TER, EMR, HON

Buyer validation only: AMZN, GXO, TSLA, BMW, Mercedes-Benz, Foxconn. Open the ranked node page.

6

Energy, Charging, Thermal, Wiring, And Structure proxy-heavy shift readiness

Runtime, charging, heat, weight, wiring, connectors, materials, and repair time decide whether a robot can work useful shifts.

Role in theme

Controls the practical workday: battery life, charge cycles, thermal limits, cable routing through joints, EMI shielding, body weight, and modular repair.

What this is

BMS, power electronics, thermal systems, connectors, wiring harnesses, bearings, structural materials, machined parts, modular panels, and serviceable assemblies.

Economic lever

Better runtime and serviceability support higher utilization, fewer warranty claims, replacement parts, service contracts, and manufacturable robot platforms.

Watch items

Runtime moving beyond short shifts, fast-charge degradation, heat dissipation, repair time, connector failures, body weight, material cost, and field service data.

Ranked return basket: TEL, APH, MPWR, ON, TXN, ADI, AME, TKR, NXPI, STM

Core wiring: TEL, APH. Power and analog: MPWR, ON, TXN, ADI. Structure and serviceability: AME, TKR. Proof-heavy control: NXPI, STM. Adjacent monitors: ITT, PH, POWI, VICR, ALGM. Source-only monitors: CATL, Panasonic, Samsung SDI, LG Energy Solution, Murata, Victrex, Solvay, Arkema.

Open the ranked node page.

7

Humanoid OEM Platforms And Regional Ecosystems platform monitor

Finished-robot platforms own fleet data and customer relationships, but most direct leaders remain private, non-U.S., or embedded in larger companies.

Role in theme

Controls the finished robot, task roadmap, fleet logs, software updates, support model, leasing or RaaS pricing, and regional supply-chain choice.

What this is

Humanoid OEMs, robot operating systems, fleet services, regional manufacturing stacks, data partnerships, service robotics, and speculative public platform routes.

Economic lever

Paid fleets become unit sales, leasing revenue, RaaS subscriptions, service attach, software upgrades, lower support cost, and proprietary fleet data.

Watch items

Named customers, paid unit count, fleet hours, expansion orders, cost per hour, gross margin, cash burn, share count, and regional supply-chain restrictions.

Watch-only public basket: TSLA, SERV, RR, PDYN, KSCP

Private or source-only monitors: Unitree, Figure, Agility Robotics, 1X, Apptronik, Boston Dynamics, UBTECH, Xiaomi, XPeng, Toyota, Honda, Hyundai, Doosan, Kawasaki, FANUC, Yaskawa, Siemens, Keyence. Open the ranked node page.

Watch Items

Area What confirms What weakens or invalidates Watch next
01Pilot-to-fleet conversionDemos need paid expansion
What confirms

Paid pilots move from 10-to-50 unit tests into repeat site orders with disclosed uptime, task success, and customer expansion.

What weakens or invalidates

Demo videos, LOIs, or unpaid trials remain the main evidence, or customers do not expand after initial tests.

Watch next
  • Paid units
  • Fleet hours
  • Repeat sites
  • Customer names
02Factory and warehouse tasksHumanoids must beat alternatives
What confirms

Robots handle kitting, machine loading, picking, conveyor work, or transport at lower cost per task than labor or fixed automation.

What weakens or invalidates

AGVs, robot arms, or machine-vision cells solve the same task with lower capex, simpler integration, and higher uptime.

Watch next
  • Cost per task
  • Exception rate
  • Integration time
  • Safety incidents
03Software and dataFleet learning must lower intervention
What confirms

Fleet logs, teleoperation data, and real-world failure recovery improve task reliability and reduce manual intervention.

What weakens or invalidates

Sim-to-real gaps persist, remote operators remain required for routine work, or data access stays fragmented across pilots.

Watch next
  • Intervention rate
  • Task success
  • Logged hours
  • Operator ratio
04Motion, hands, and service lifeHardware has to survive shifts
What confirms

Actuator cost falls while torque density, hand durability, thermal performance, repair time, and service life improve together.

What weakens or invalidates

Hands fail quickly, reducers stay high-cost, warranty claims rise, or maintenance burden erases labor savings.

Watch next
  • Actuator cost
  • Hand lifespan
  • Warranty claims
  • Supplier lead times
05Semis and sensingDesign wins need robot revenue
What confirms

Supplier filings disclose robot design wins, content per unit, module shipments, sensor ASPs, product margin, and customer breadth.

What weakens or invalidates

Revenue is still explained by data-center AI, auto, smartphones, or broad industrial recovery rather than robot shipments.

Watch next
  • Robot design wins
  • Module shipments
  • Sensor ASPs
  • Gross margin
06Safety and regulationWork near humans needs rules
What confirms

Standards, safety cases, workplace certifications, and insurance frameworks allow repeat deployments without reducing productivity.

What weakens or invalidates

Liability remains unclear, force or speed limits reduce useful work, or certification timelines delay deployments.

Watch next
  • ISO progress
  • Safety cases
  • Insurance terms
  • Incident reports
07Speculative public platformsOption value needs filing proof
What confirms

Fleet revenue, gross margin, cash runway, retained customers, and share count improve in the same filing cycle.

What weakens or invalidates

Launch news arrives without revenue, dilution accelerates, customer churn rises, or unit economics stay negative.

Watch next
  • Revenue per robot
  • Gross margin
  • Cash burn
  • Share count

Source Trail

Route Use it for Primary links
Paper baseline Humanoid-specific market framing, cost curve, physical AI, VLA/VLM control, pilot deployments, component bottlenecks, labor scarcity, China/West scaling curves, and deployment sequence. Citrini humanoid primer; Roland Berger 2026 extraction; Roland Berger PDF
Canonical boundary Deployed-machine proof framework, defense autonomy, field autonomy, physical AI routing, funding quality, cash conversion, and source gaps. Physical AI And Defense Autonomy; Physical AI report page
Broad robotics map Existing report-layer nodes for factory automation, warehouse deployment, sensing, motion components, edge semis, and speculative service robotics. Robotics; Factory Automation And Robot OEM Proxies; Warehouse Automation And Logistics Deployment; Sensing, Vision, And Perception Stack; Robot Motion, Controls, And Precision Components; Edge Compute And Robot Semiconductors; Emerging Service And Speculative Robotics
Sector mechanics Industrial order conversion, automation capex, funding and tariff friction, semiconductors, AI infrastructure, edge compute, electronics components, and export-control caveats. Industrials; Information Technology
Companion themes AI capex, component scarcity, external funding needs, sourcing geography, export controls, and high-beta market-fragility overlays. AI Capex Cycle; Memory And Component Scarcity; Funding Dependency; Trade And Sourcing; Market Fragility
Selected security lanes Company-level setup, proof burden, catalysts, and risk routing for filed names used in the proxy baskets. A filed security lane is useful, but it is not required for return-basket inclusion when local price coverage exists. TSLA, NVDA, AME, TKR, RRX, CGNX, TER, SYM, ZBRA, ROK, TXN, MPWR, ON, TEL, MCHP, NOVT, APH, RR, SERV, PDYN, KSCP
Source gaps and monitors Direct humanoid OEMs, hands, tactile sensors, reducers, and many China/Japan/Europe suppliers remain private, non-U.S., or lack local OHLC coverage. Price-covered names without filed lanes route through the generic security page until research is created. ON and TEL now have filed security lanes. Unitree, Figure, Agility Robotics, 1X, Apptronik, Boston Dynamics, UBTECH, Xiaomi, XPeng, Harmonic Drive, Nabtesco, THK, Nidec, SMC, Schunk, Robotiq, Keyence, Omron, Infineon, Renesas, and price-covered names without filed lanes stay on monitor routing until research is created.
Discovery provenance Parent return grids and selected-security charts use the local report API and discovery-backed daily OHLC where coverage exists. report.json; API route /api/themes/humanoid-robotics/node-return-buckets?as_of=latest; selected-security charts route to /api/securities/TICKER/chart.

Categories