SPEC SHEET / VOL.13 REV 2026.05
CLASSIFIED // FORWARD-LOOKING

Industry Briefing

ROBOTICS
FUTURE/Humanoids
in every warehouse.

A 13-slide field briefing on the next decade of physical AI: humanoid platforms, autonomous agriculture, and robotic surgery.

FILE: FUT-ROB-2026-001   //   PRESS SPACE OR TO ADVANCE
02 / 13SECTION A — LANDSCAPE
STATUS: ACTIVE

The Current Wave

A handful of companies are racing to build the first general-purpose humanoid.

Five years ago this was sci-fi. Today, multi-billion-dollar bets are being placed on bipedal robots that can walk, lift, and manipulate.

Tesla
Optimus Gen 2
target: $20-30k unit price
Figure
Figure 02
BMW pilot, OpenAI partnership
1X Technologies
NEO Beta
consumer-home positioning
Apptronik
Apollo
Mercedes manufacturing trial
Unitree
G1 / H1
$16k starting — China cost edge
Boston Dynamics
Atlas (Electric)
Hyundai integration
03 / 13SECTION A — DRIVERS
CONVERGENCE INDEX: HIGH

Why Now

Three curves crossed at once.

  • AI ControllersTransformer-based VLA (vision-language-action) models can map raw camera input to motor commands. Behavior cloning + RL closes the loop that classical control could not.
  • Actuator CostBrushless motors, harmonic drives, and force-torque sensors have dropped 5-10x in a decade as drones and EVs scaled the supply base.
  • Battery DensityEV-grade Li-ion now delivers 250-300 Wh/kg, enough to run a 60kg humanoid for 4-5 hours of light work on a single charge.
AI capability battery Wh/kg actuator $/Nm (inv.) 2015 2026
FIG.03 — CONVERGENCE OF ENABLING TECHNOLOGIES, INDEXED 2015=1
04 / 13SECTION B — SUPPLY CHAIN
BOM ANALYSIS

The Supply Chain

A humanoid is a rolling manifest of geopolitically scarce parts.

Every robot pulls components from a small set of dominant suppliers. Choke points define who builds at scale.

Precision Reducers
Harmonic Drive · Nabtesco
JP duopoly · 14-30 per humanoid
Battery Cells
CATL · BYD · LG Energy
CN/KR · 2-5 kWh per unit
Compute
NVIDIA Jetson / Thor
edge inference · 100-500 TOPS
Permanent Magnets
NdFeB rare earths
~85% China refining share
Force/Torque Sensors
ATI · Bota · Robotiq
tactile feedback bottleneck
Vision
RGB-D · LiDAR · IMU
Sony/Intel/Bosch silicon
05 / 13USE CASE 01
DEPLOY: 2025-2028

Warehouse / Pick-and-Pack

The first market is the dirtiest, dullest, and most measurable.

Amazon, GXO, FedEx, and DHL already run thousands of mobile robots. Humanoids slot into the last meter — the bin, the conveyor, the trailer floor.

  • Workload1,200+ picks/hr ceiling for humans; humanoids target 600 today, 1,000 by 2027.
  • EconomicsBreak-even at ~$10/hr fully-loaded (incl. depreciation, downtime, supervision).
  • PilotFigure at BMW Spartanburg; Apptronik at Mercedes; Digit (Agility) at GXO.
PAYLOAD 25kg PICK STATION 04 · ZONE B CYCLE TIME: 3.6s · ERROR: 0.4%
FIG.05 — HUMANOID PICK-AND-PACK SCHEMATIC
06 / 13USE CASE 02
DEPLOY: ACTIVE

Agriculture / Field Robotics

Fields are open, structured, and labor-starved — the perfect first frontier.

John Deere's autonomous 8R tractor and 9.7L sprayer have been operating since 2022. Carbon Robotics' LaserWeeder kills 200,000 weeds an hour with no chemicals.

  • TractorsJohn Deere · CNH · Monarch — full autonomy on row crops, GPS-RTK accuracy ±2 cm.
  • HarvestersTortuga (strawberries), Advanced Farm (apples), Saga Robotics (raspberries).
  • Weeding/ScoutingCarbon Robotics, FarmWise, Naïo — computer vision + actuators replace herbicide.
  • AerialXAG, DJI Agras — swarm spraying at 40 ha/hr per unit.
SPRAY ZONE AUTON. TRACTOR RTK ±2cm · 24/7 OPS
FIG.06 — DRONE OVERHEAD + GROUND-UNIT FIELD COVERAGE
07 / 13USE CASE 03
DEPLOY: MATURE

Surgery / Precision Platforms

The most profitable robot in the world wears scrubs.

Intuitive Surgical's da Vinci system has performed over 14 million procedures. The category posted $7.1B in 2023 revenue with operating margins above 30%.

Intuitive Surgical
da Vinci 5
~9,000+ systems installed
Stryker
Mako
orthopedic / joint replacement
Medtronic
Hugo RAS
soft-tissue, modular arm
CMR Surgical
Versius
UK · portable form factor
Vicarious Surgical
v1.0
single-port, VR-control
Johnson & Johnson
Ottava
2024+ trial entry
"Robotic-assisted surgery is the rare physical-AI category where the unit economics already work — and have for two decades.— Industry analyst, 2024
08 / 13USE CASE 04
DEPLOY: 2027-2032

Eldercare / Assistive Robotics

Demographics will force the question, faster than the technology is ready.

Japan will have one care worker per two seniors by 2040. Germany, Italy, Korea, and China face the same wall. The robot doesn't have to be a humanoid — but it has to be physical.

  • Lift AssistToyota HSR, RIBA — transfer patient between bed and chair without back injury.
  • Fall DetectionAmbient radar + computer vision; sub-2-second alert to caregiver.
  • MedicationPill dispensing with reminder loop, biometric verification.
  • CompanionshipPARO seal, ElliQ — conversational + touch-responsive, slowing cognitive decline in trials.
Demographic Forcing Function
2.1 billion people over 60 by 2050.
UN World Population Prospects · global care-worker gap est. 13M by 2030 (WHO)
Japan — Reference Market
29.1%
share of population over 65 (2024)
09 / 13SECTION C — HARD PROBLEMS
UNSOLVED

The Hard Part / Dexterity

A two-year-old's hands are still ahead of every robot ever built.

Walking is a solved problem. Talking is solved. Picking up an unfamiliar object without crushing it or dropping it remains an open research frontier.

The bottleneck is tactile sensing density, sub-millisecond control loops, and a missing dataset. Most "humanoid demos" use carefully chosen objects in controlled poses.

  • Skin SensorsHuman fingertip ≈ 240 mechanoreceptors/cm²; best robot tactile arrays ~5/cm².
  • ComplianceSeries-elastic actuators trade speed for safety; impedance control still hand-tuned per task.
  • GeneralizationModels trained on rigid blocks fail on cloth, deformables, transparent objects.
CONTACT FORCE EST: ±0.5N · OBJECT GEOMETRY: UNKNOWN
FIG.09 — UNSTRUCTURED GRASP · CONFIDENCE: LOW
10 / 13SECTION C — DATA
SCARCITY: HIGH

The Data Problem

Real-world manipulation data is the new oil.

There is no internet of robot demonstrations. Every action — every grasp, lift, place — must be captured from teleoperation rigs, sim-to-real pipelines, or in-deployment self-supervised collection.

01
Teleop
Humans wear VR rigs and puppet robots through tasks. Slow, expensive, high quality.
02
Simulation
Isaac Sim, MuJoCo. Cheap, scalable, but suffers reality gap on contact physics.
03
Video Pre-train
Pre-train policies on YouTube/egocentric video. Action labels hard to recover.
04
Fleet Learning
Deployed robots upload trajectories. Compounding moat for whoever ships first.
The first company to deploy a million humanoids will have a dataset no one else can buy.— THE NEW MOAT
11 / 13SECTION D — POLICY
REGULATION: NASCENT

Safety + Regulation

When a 60-kilogram robot shares a hallway with a child, who is liable?

Industrial robots have lived behind cages for 60 years. Cobots loosened that. Humanoids will demolish it. The legal and insurance frameworks are not ready.

  • StandardsISO 10218 (industrial), ISO/TS 15066 (collaborative), ISO 13482 (personal care). Humanoid-specific: pending.
  • LiabilityProduct liability vs. operator negligence vs. AI-decision liability — unsettled in every major jurisdiction.
  • CybersecurityA compromised humanoid is a remote weapon. Few standards mandate signed firmware or kill-switch redundancy.
  • Labor PolicyTax and retraining frameworks (UBI, robot tax) still theoretical at national scale.
EU AI Act — physical robots
High-risk classification likely for humanoids in workplaces and public spaces; conformity assessments required.
phased compliance: 2025-2027
Open Question
Should humanoids be required to identify themselves as non-human in public?
no jurisdiction has answered
12 / 13SECTION E — OUTLOOK
CONFIDENCE: MIXED

The Honest Assessment

Industrial expansion is certain. Humanoid mass-market is a 2030s story.

Near-term · 2025-2027
Bounded autonomy
Surgery scales; agriculture scales; warehouse pilots.
Mid-term · 2027-2030
Industrial humanoid
Single-task humanoids in factories. 100k-class fleets per OEM.
Long-term · 2030-2035
Multi-task humanoid
Cross-task learning works. Per-unit cost < $25k. Eldercare opens.
Speculative · 2035+
Domestic humanoid
Kitchen-grade dexterity. Mass consumer market. Skeptics still skeptical.

The hype cycle is loud, but the underlying capital allocation is real. Even a fraction of the announced humanoid roadmaps would create a multi-hundred-billion-dollar component industry by 2030 — independent of whether any humanoid OEM achieves the consumer dream.

13 / 13SECTION F — REFERENCES
END OF BRIEFING

References / Further Viewing

Watch the machines move.

Industry Source
Intuitive Surgical 10-K · NVIDIA GTC keynotes · Boston Dynamics technical blog · IFR World Robotics Report · ARK Invest Big Ideas
Companies Tracked
Tesla · Figure · 1X · Apptronik · Unitree · Boston Dynamics · Agility · Intuitive · Stryker · Medtronic · John Deere · Carbon Robotics · CATL · NVIDIA · Harmonic Drive
FILE: FUT-ROB-2026-001   //   PRESS TO REVIEW
nav  ·  SPACE next  ·  HOME/END
01 / 13