Robotics power-runtime gate: from battery spec to billable uptime

Date: 2026-06-16 Agent: Finance / Charlie Status: SYNTHESIS_CANDIDATE Visibility: PUBLIC Public-safety: Evidence map only. No winner ranking. No trade recommendation. No Hugo portfolio data, private rationale, paid-report excerpts, or rumors.

0. One-line answer

The stale part of the robotics knowledge base is not another demo or even another deployment headline; it is the denominator that converts a robot into labor capacity: battery energy, charge time, autonomous docking, duty-cycle scheduling, safety certification, and achieved productive uptime. Current public evidence suggests humanoids are improving from 2-5 hour robots toward shift-compatible operations, but AMR benchmarks already show 8-14 hour runtimes and fleet-level charging orchestration, so humanoids remain S4 unless vendors disclose accepted fleet uptime, intervention rates, and ROI/payback.

1. Core question

Can a humanoid robot produce enough safe, scheduled, billable work-hours per day to matter economically, or is the industry still optimizing impressive single-cycle performance without proving duty-cycle economics?

This question matters because a 5-hour runtime claim is not the same as a 16-24 hour operational asset. The missing bridge is:

1. Battery capacity and runtime.
2. Charge time and autonomous docking.
3. Number of batteries / swaps / charging bays required per robot.
4. Productive task time vs walking / waiting / intervention / maintenance / charging time.
5. Site-level uptime across a fleet.
6. Customer ROI/payback and safety acceptance.

2. Why this is the next valuable stale-knowledge patch

Existing robotics artifacts already cover deployment evidence, customer pilots, standards, liability, data factories, dexterity benchmarks, and S4-to-S5 upgrade gates. The remaining blind spot is the operational denominator: the number of reliable robot-hours that can be scheduled into customer workflows.

Signal / noise classification:

• Signal: published runtime plus charge-time plus autonomous docking plus customer-site duty-cycle KPI plus fleet-management support.
• Signal: runtime disclosed together with safety certification path, abuse-tolerance testing, thermal design, and manufacturability.
• Signal: customer-side achieved shift duration, productive task counts, uptime distribution, charge events, intervention rate, and maintenance hours.
• Noise: standalone kWh, standalone video, “all-day” language, or “commercial-ready” claims without duty-cycle denominator.
• Noise: price-access claims without runtime, durability, service burden, or safety acceptance.

3. Evidence map: humanoids vs AMR duty-cycle benchmark

4. Reading: the hidden denominator is available work-hours, not robot count

A useful denominator is:

Available robot-hours per day = fleet robot count x calendar hours x uptime x productive-task share.

Public sources rarely disclose this chain. We can, however, classify evidence quality:

• S3 demo/spec: battery capacity, runtime, payload, speed, and product videos.
• S4 operational readiness: runtime plus autonomous docking, fleet management, safety stop systems, customer-site KPI, and manufacturing plans.
• S5 economic proof: accepted fleet robot-hours, uptime/intervention distribution, maintenance burden, charge-cycle schedule, task throughput, revenue, gross margin, ROI/payback, and repeat orders.

Current classification:

• Figure F.03 battery: S4 subsystem readiness signal because 2.3 kWh / 5h / 2kW charging / safety-certification path / cost-reduction are concrete, primary-source operating inputs.
• Unitree G1: S3/S4 access signal because US$13.5k price and 2h quick-release battery expand experimentation, but runtime does not yet support industrial shift economics.
• Agility Digit: S4 commercial-readiness signal because 4h runtime, autonomous docking, Arc charger support, and safety stack move toward deployable fleet operation.
• Locus AMRs: S5-ish benchmark for warehouse AMR category, not humanoids; 8-14h runtime and opportunity charging show what mature robots expose as product facts.

5. What would change the thesis

Upgrade humanoids toward S5 if public/customer-side sources disclose:

1. Fleet count by site and accepted robots in production use.
2. Productive robot-hours per day / week / month.
3. Uptime distribution, not just average runtime.
4. Charge events, charge time, battery swap count, battery degradation, and maintenance hours.
5. Human intervention rate per task or per hour.
6. Task throughput vs human baseline and safety incident rate.
7. Customer ROI/payback, renewal/repeat order, vendor revenue, gross margin, and service burden.

Downgrade if:

1. Runtime improves but charge scheduling requires excessive human handling.
2. Safety certification / battery abuse testing remains incomplete for customer environments.
3. Robots can work only short demo blocks or require frequent resets.
4. Published customer KPI omits downtime, intervention, and charge/maintenance time.
5. Low price expands trials but support cost or breakage prevents repeat deployment.

6. Common misconceptions

Misconception 1: “5 hours runtime means a robot can replace one shift.”

Not necessarily. A human shift includes breaks, but a robot shift includes charging, thermal limits, reset time, task waiting, failed grasps, safety stops, updates, maintenance, and supervision. The relevant metric is accepted productive work-hours, not nominal battery runtime.

Misconception 2: “Quick-release batteries solve uptime.”

Only if the site has safe swap procedures, spare inventory, trained staff or automated swap/dock infrastructure, battery health tracking, and no hidden labor burden. Otherwise battery swap moves the bottleneck from robot hardware to operations.

Misconception 3: “AMR battery specs prove humanoid feasibility.”

AMRs are the benchmark because they show what operational maturity looks like. They do not prove humanoids can match the same uptime while adding bipedal locomotion, manipulation, perception, safety, and human-shaped workspace constraints.

Misconception 4: “Battery is just a component issue.”

For humanoids, battery is a commercialization gate. It affects weight, center of mass, heat, safety certification, charging infrastructure, service model, fleet scheduling, and customer ROI.

7. Public-safe site draft section

The quiet robotics question: how many useful hours does the robot create?

Robotics has moved beyond the pure demo cycle. The next evidence gate is simpler and harder: how many safe, useful, scheduled work-hours can the robot deliver per day?

A humanoid with a 2-5 hour runtime can still be a strong engineering achievement. But for commercial deployment, runtime must connect to charging, docking, uptime, intervention rate, task throughput, maintenance, and customer ROI. That is why mature warehouse AMRs are an important benchmark: public product pages already frame 8-14 hour runtime, fast charging, and autonomous opportunity charging as basic operational facts.

The public humanoid evidence is improving. Figure says its F.03 battery reaches 2.3 kWh, 5 hours at peak performance, 2 kW fast charging, 94% energy-density improvement across three generations, and a 78% cost reduction vs F.02. Agility says Digit reaches up to 4 hours and can autonomously dock onto a charging station. Unitree’s G1 makes humanoid experimentation far cheaper, with a listed US$13.5k starting price and about 2 hours battery life.

That is S4 progress: real subsystems and deployment design are becoming visible. It is not yet S5 proof. S5 would require accepted fleet robot-hours, uptime distributions, intervention rates, maintenance cost, repeat orders, and customer ROI/payback.

8. Think Deeper questions

1. Which robot companies will first disclose productive robot-hours instead of runtime specs?
2. Does autonomous docking become a bigger moat than raw battery capacity?
3. Does low-cost humanoid access accelerate software learning enough to overcome weaker duty-cycle specs?
4. Should battery certification and thermal abuse tests be treated as leading indicators of customer acceptance?
5. In warehouse and manufacturing workflows, where is the hard cutoff between useful S4 pilot and economically meaningful S5 asset?

9. Source list

• Figure AI, “F.03 Battery Development,” 2025-07-17. 🟢 Primary. Key claims: 2.3 kWh, 5h runtime at peak performance, 2kW fast charge, 94% energy-density increase across three generations, 78% cost reduction vs F.02, UN38.3 / UL2271 certification path, 23 primary tests.
• Unitree Robotics, “Unitree G1” official product page, accessed 2026-06-16. 🟢 Primary. Key claims: price from US$13.5k, 35kg weight, 9000mAh quick-release battery, 54V 5A charger, about 2h battery life, safety/limitation caveats.
• Agility Robotics, “Agility Robotics Announces New Innovations for Market-Leading Humanoid Robot Digit,” 2025-03-31. 🟢 Primary. Key claims: up to 4h runtime, autonomous docking, CAT1 stop, Safety PLC, on-robot E-stop, FSoE, Arc charger/support/maintenance/MES/WMS/WES/PLC integrations.
• Locus Robotics, “Locus Vector” product page, accessed 2026-06-16. 🟢 Primary. Key claims: 8-10h runtime per charge, ~60min full charge, autonomous opportunity charging, up to 600lb payload.
• Locus Robotics, “Locus Origin” product page, accessed 2026-06-16. 🟢 Primary. Key claims: 14h runtime per charge, 50min full charge, autonomous opportunity charging, 80lb payload, >2x productivity claim.
• Amazon, “Amazon robotics: Meet the robots inside fulfillment centers,” accessed 2026-06-16. 🟢 Primary for >1m robots, site robotics systems, DeepFleet context, and fleet-scale denominator.
• Amazon, “Amazon unveils next-gen Proteus robot as part of €10 billion European investment,” 2026. 🟢 Primary for Proteus at 25 U.S. fulfillment centers, close to 400kg carts, STARK expansion to 15 European sites by 2027, and H1 2027 next-gen Proteus deployment plan.
• Technology.org, “Amazon's Talking Proteus Robot Heads to Europe,” 2026-06-05. 🟡 Secondary. Use only for unconfirmed Proteus charge-cycle details; do not use as primary proof.
• Charlie synthesis of power-runtime denominator and S4/S5 classification. 🟠 Estimate / synthesis.

10. Public-safety check

• No Hugo private portfolio weights, trade rationale, tax context, private channel checks, paid-report excerpts, or rumors included.
• No buy / sell / hold recommendation included.
• Public claims are source-backed and quantified where possible.
• Legal/safety/insurance implications are framed as research gates, not advice.
• Output should remain SYNTHESIS_CANDIDATE unless Hugo asks to package into site/slide/dashboard.