Robotics energy / duty-cycle gate — from demo-hours to shift economics

Date: 2026-06-18 Status: SYNTHESIS_CANDIDATE Visibility: PUBLIC Output intent: none by default Public-safety: yes — no Hugo portfolio data, no trade recommendation, no private channel checks, no paid-report excerpts.

1. One-line answer

The next high-value robotics evidence gap is the energy / duty-cycle gate: humanoids are starting to disclose battery, charging, swapping, shift-runtime, and daily-operation claims, but public evidence still supports an S3/S4 deployment-readiness signal rather than S5 scaled labor economics. 🟢 primary source claims through 2026-06-18; 🟠 Charlie stage synthesis.

2. Core question

Can a humanoid robot move from a useful task demo into sustained labor availability across real shifts without requiring too much charging downtime, spare fleet capacity, human supervision, battery handling, safety reset, or maintenance?

This matters because even a competent robot is not economically equivalent to labor unless its useful operating time, recharge / swap process, intervention rate, and support burden fit customer workflows. 🟠 framework judgment.

3. Why this is stale-gap material

Prior robotics artifacts already cover:

• factory buildout vs deployment economics;
• Figure / Unitree / Tesla / UBTECH / Agility / Apptronik evidence curves;
• teleoperation disclosure;
• safety / conformity gates;
• AMR and warehouse robotics as S5 counterfactuals;
• robotics foundation-model data flywheel.

The missing public-safe layer is duty-cycle quality: not “can the robot do the task once?” but “how much useful work can it perform per day, under what energy replenishment model, and with what disclosed uptime / intervention denominator?”

4. Evidence table — energy and shift anchors

5. Derived checks and why they matter

These derived numbers are estimates based only on disclosed figures; they are not company-reported KPIs.

1. Figure BMW throughput proxy: 90,000+ parts / 1,250+ runtime hours = about 72 parts/hour. 🟠 derived from Figure primary figures.

   • Signal: enough to reason about task-level productivity, not enough to conclude payback.
   • Missing: robot count, labor baseline, cycle-time achievement, downtime, intervention frequency, and total cost.

2. Figure BMW shift availability context: a 10-hour weekday schedule equals 50 scheduled hours/week, about 29.8% of theoretical 24/7 calendar time. 🟠 derived from 10h x 5 days / 168h.

   • Signal: the deployment was meaningful daily production-line operation.
   • Caveat: it was not a disclosed 24/7 autonomous labor replacement.

3. Unitree G1 nominal battery-energy proxy: 13-string lithium battery x 3.7V nominal x 9000mAh ≈ 433Wh; about 2h battery life implies roughly 216W average draw if the nominal pack assumption holds. 🟠 engineering estimate from Unitree disclosed battery format.

   • Signal: public hardware spec allows power-envelope thinking.
   • Caveat: actual voltage, configuration, workload, and safety margins vary; do not treat this as audited performance.

4. Unitree G1 charger power: 54V x 5A = 270W charger rating. 🟠 derived from official charger spec.

   • Signal: energy replenishment may be a major operational design question if robots are expected to run many hours/day.
   • Caveat: charge curve, battery health, swap logistics, and multi-battery workflows are undisclosed.

5. 1X “6min per hour runtime” snippet implies 0.1 charging hours per runtime hour, or 10 runtime-hours replenished per charge-hour, if interpreted literally. 🟡/🟠 because this came from search-result snippet and needs re-extraction from full order page.

   • Signal: consumer-home humanoid vendors are now marketing energy convenience as a product feature.
   • Caveat: home chores may have sparse duty cycles; this does not imply industrial continuous operation.

6. Signal vs noise

Signal

• Named customer + task + runtime + intervention KPI target, e.g. Figure BMW. 🟢
• Autonomous charging / docking / battery swap disclosed at product or deployment layer, e.g. Agility Digit, UBTECH Walker S2, 1X NEO. 🟢
• Fleet-management integration with chargers, AMRs, WMS/MES/WES/PLC, safety PLC, E-stop, FSoE, remote support, e.g. Agility Arc. 🟢
• Shift schedule and parts / task denominator disclosed, not just “robot worked in a factory.” 🟢
• Battery / charger specs that let us estimate energy envelope, e.g. Unitree G1 9000mAh, 54V 5A, about 2h. 🟢/🟠

Noise

• “24/7” marketing without field uptime, swap-station utilization, service incidents, maintenance, or customer ROI. 🟠
• Battery life in a spec table without workload definition. 🟠
• A robot video showing a task once without runtime hours, intervention count, recharge process, or support model. 🟠
• “Self-charging” claims without delivered-unit evidence, utilization, failure handling, and customer retention. 🟠
• Treating charging architecture as margin proof. 🔴 if unsupported.

7. Stage classification

Current classification: S3/S4 duty-cycle evidence, not S5 scaled labor economics.

• Unitree G1: S3 productized hardware / developer access. 🟢/🟠
• UBTECH Walker S2: S4 architecture signal because autonomous swap directly targets continuous operation. 🟢/🟠
• Agility Digit: S4 industrial deployment infrastructure because runtime, autonomous docking, safety, Arc fleet management, and AMR integration appear together. 🟢/🟠
• Figure BMW: S4 named-customer task KPI because runtime, parts handled, shift schedule, and KPI targets are disclosed. 🟢/🟠
• 1X NEO: S3/S4 home-service design because self-charge and Expert Mode are explicit, but delivered real-home metrics remain pending. 🟢/🟠

S5 would require repeat paid deployments with customer-confirmed ROI/payback, high utilization, low intervention, disclosed uptime, support cost, revenue, margin, and repeat orders. 🟠 stage rule.

8. What would change our mind

Upgrade signals:

1. A named customer discloses robot count, uptime %, productive hours/day, intervention minutes/hour, charging or swap downtime, and payback. 🟢
2. A vendor reports multi-site fleet utilization with charger / swap-station denominator, not just robot specs. 🟢
3. Humanoid vendors disclose duty-cycle economics: spare battery count, battery life degradation, maintenance intervals, field failures, service gross margin. 🟢
4. Figure-like deployments show repeated task KPI across multiple customers and shifts, with achieved intervention target. 🟢
5. Agility / UBTECH / 1X publish retained-customer or repeat-order data tied to energy / autonomy / support model. 🟢

Downgrade signals:

1. “24/7” claims remain marketing while field deployments operate only supervised day shifts. 🟠
2. Battery replacement, charging, reset, calibration, or support labor becomes the hidden cost center. 🟠
3. Customers require large spare fleets to compensate for charging / maintenance downtime. 🟠
4. Energy management works in labs but not in safety-regulated customer sites. 🟠
5. Home robots generate high support / teleoperation burden relative to paid price or subscription revenue. 🟠

9. Common misconceptions

1. “2h battery life is bad, therefore humanoids cannot commercialize.”

   • Correction: many workflows are intermittent or can use swap / dock / fleet orchestration. The question is useful duty cycle and cost, not battery life alone. 🟠

2. “Autonomous battery swap means 24/7 economics are solved.”

   • Correction: swap architecture attacks downtime, but S5 still needs station cost, field reliability, service burden, safety approval, customer utilization, and ROI. 🟢/🟠

3. “Figure ran 10-hour shifts, so it replaced a human shift economically.”

   • Correction: Figure disclosed meaningful runtime and task KPIs, but not contract value, robot count in the official post, achieved intervention result, energy model, or payback. 🟢/🟠

4. “Self-charging home robots are ready for consumer economics.”

   • Correction: self-charging is necessary but insufficient; home economics also require autonomy, safety/privacy acceptance, retention, support cost, and subscription conversion. 🟢/🟠

5. “Energy is a hardware-only problem.”

   • Correction: duty cycle is a system problem: battery, charger/swap, software scheduling, task design, fleet orchestration, safety, maintenance, and human-in-loop support. 🟠

10. Public-safe site draft section

The boring robotics question: how many useful hours per day?

Humanoid robotics is becoming easier to measure. The next useful denominator is not just price, speed, or a viral task video. It is duty cycle.

Unitree lists about 2 hours of battery life for G1. UBTECH’s Walker S2 attacks the same bottleneck with autonomous battery swapping in about 3 minutes and markets 24/7 continuous operation. Agility says Digit can run up to 4 hours and dock autonomously, while its Arc platform connects robots, chargers, AMRs, remote support, and enterprise systems. Figure’s BMW deployment gives a different type of evidence: 10-hour weekday shifts, 1,250+ runtime hours, 90,000+ parts loaded, and explicit cycle-time / accuracy / intervention KPIs.

That is progress. But it is not yet full economics. The investable question is whether these robots can deliver high useful-hours/day with low intervention, low downtime, acceptable safety, and customer-confirmed payback. Until those metrics are public, energy and duty-cycle disclosures should be treated as S3/S4 deployment-readiness evidence — stronger than demos, weaker than scaled labor economics.

Footer: Evidence map only. No winner ranking. No trade recommendation. Battery life, self-charging, shift runtime, and battery swapping are duty-cycle signals; they are not customer ROI or gross-margin proof.

11. Think Deeper questions

1. In early humanoid deployments, does the binding constraint shift from “can it do the task?” to “can it do enough useful hours/day without support labor eating the savings?”
2. Which architecture wins duty-cycle economics: larger batteries, fast charging, autonomous docking, battery swapping, spare fleets, or low-utilization task scheduling?
3. Is “24/7 humanoid labor” a realistic near-term goal, or should early deployments be benchmarked against bounded 4-10 hour task windows?
4. Who captures value from duty-cycle orchestration: OEM, fleet-management software, customer systems integrator, battery/swap supplier, or RaaS operator?
5. Should public research require vendors to disclose “productive hours per robot per day” before treating deployments as S5?

12. Source list

• Unitree G1 official product page, extracted 2026-06-18: https://www.unitree.com/g1 🟢 Used for US$13.5K price, 9000mAh smart quick-release battery, 54V 5A charger, about 2h battery life, 23-43 DOF, 35kg weight, and safety caveats.
• UBTECH Walker S2 official product page, extracted 2026-06-18: https://www.ubtrobot.com/en/humanoid/products/walker-s2 🟢 Used for autonomous battery swapping, 24/7 continuous-operation claim, dual-battery dynamic balancing, and 3-minute autonomous swap claim.
• 1X NEO official product page, extracted 2026-06-18: https://www.1x.tech/neo 🟢 Used for self-charging, Expert Mode, foundational autonomy, and human expert remote guidance framing.
• 1X order page search result, accessed 2026-06-18: https://www.1x.tech/order 🟡 Used only for 842Wh and “6min per hour runtime” snippet; needs full-page re-extraction before public charting.
• Agility Robotics official announcement, 2025-03-31, extracted 2026-06-18: https://www.agilityrobotics.com/content/agility-robotics-announces-new-innovations-for-market-leading-humanoid-robot-digit 🟢 Used for up-to-4h runtime, autonomous docking, CAT1 stop, Safety PLC, on-robot E-stop, FSoE, Arc fleet management, AMR integration, and remote support / enterprise-system integration.
• Figure AI official post, 2025-11-19, extracted 2026-06-18: https://www.figure.ai/news/production-at-bmw 🟢 Used for 11-month BMW deployment, 10-hour Monday-Friday shifts, 90,000+ parts, 1,250+ runtime hours, 30,000+ X3 vehicles, 1.2m+ steps / 200+ miles, 84s cycle-time target, >99% placement-success target, and zero-intervention target.
• Charlie derived duty-cycle calculations, 2026-06-18 🟠 Used for Figure 72 parts/hour proxy, Figure 29.8% of theoretical 24/7 schedule, Unitree nominal pack-energy proxy, Unitree charger power, and 1X charge/runtime interpretation.

13. Risks / exclusions

• Do not include Hugo private portfolio data, position weights, watchlist sizing, purchase prices, tax context, private trade rationale, private channel checks, paid-report excerpts, or rumors.
• Do not frame Unitree, UBTECH, Agility, Figure, 1X, Tesla, Apptronik, Mercedes-Benz, BMW, GXO, NVIDIA, Google, or any related public/private company as buy / sell / hold.
• Do not imply battery life, autonomous charging, battery swapping, or 10-hour shifts prove ROI, revenue, gross margin, repeat orders, uptime, or S5 scaled economics.
• Do not use the 1X 842Wh / 6min-per-hour runtime snippet in a final public chart until it is re-verified from a full primary page extraction.
• Do not wake Warrior or mark READY_FOR_WARRIOR unless Hugo explicitly approves packaging.