Robotics energy / charging utilization gate — battery specs are not productive shift economics

Date: 2026-06-18 Owner: Finance / Charlie AGT-002 Status: RESEARCH_ONLY Visibility: PUBLIC Output target: none Public-safety flag: yes. No Hugo private portfolio data, no trade recommendation, no private channel checks, no paid-report excerpts. Source discipline: 🟢 primary / 🟡 secondary / 🟠 estimate / 🔴 guess

0. One-line answer

The freshest high-value robotics stale gap is the energy / charging utilization gate: humanoid vendors are now disclosing larger packs, fast charging, wireless or autonomous docking, quick-release batteries and safety certifications, but S5 economics still require a harder denominator — productive robot-hours per 24h, charge scheduling, battery degradation, charger/fleet ratio, duty-cycle utilization, service burden and customer ROI. Figure’s 2.3 kWh / 5h / 2 kW fast-charge stack, Agility Digit’s up-to-4h autonomous-docking update, Unitree H2/H2 Plus’s 0.972 kWh / about-3h quick-release battery, and Tesla’s lack of Optimus battery/runtime disclosure together show why “runtime spec” is S3/S4 evidence, not S5 proof. 🟢/🟠

1. Core question

When a humanoid robot enters a warehouse, factory, store, home, or public-adjacent commercial site, can its battery and charging system turn robot capability into a repeatable productive shift — or does the fleet lose too much time to charging, docking, swaps, thermal limits, battery degradation, maintenance, safety review and human supervision?

This is an investment-research evidence gate, not engineering, certification, legal, tax, or deployment advice.

2. Why this is additive, not duplicate

Existing robotics artifacts already cover:

runtime / uptime / intervention as S4-to-S5 gates,
maintenance and service burden,
safety / standards / conformity,
Figure / Tesla / Unitree / Agility company evidence curves,
service-robot denominators and RaaS adoption.

This artifact isolates the energy system as a fleet-utilization bottleneck. A runtime number is only one input. The commercial denominator is a full duty-cycle equation:

productive robot-hours = scheduled task hours - charging/docking/swap time - maintenance/reset time - safety stop time - remote/intervention time - unplanned downtime.

A humanoid can look capable in a single shift and still be economically weak if energy replenishment, battery wear, charger footprint, support staffing or thermal/safety limits prevent high utilization across a fleet.

3. Evidence table

Evidence layer	Source-backed fact	Quantified / dated anchor	What it changes	Source grade	Signal grade
Figure F.03 pack	Figure says F.03 battery is in-house, integrated into the torso, and designed for runtime, fast charge, safety, reliability, cost and scale.	2.3 kWh; 5 hours runtime at peak performance; 2 kW fast charge with active cooling; 94% energy-density increase across generations; 78% cost reduction vs F.02; post dated 2025-07-17.	Strong S4 subsystem evidence: battery is becoming part of humanoid product architecture and manufacturing strategy, not a generic component.	🟢 Figure official	S4 energy-stack signal
Figure battery safety	Figure says F.03 targets UN and UL certification, is in process for UN38.3 and UL2271, and was designed around 23 primary tests. Figure 03 page later says the battery has achieved UN38.3 certification.	UN38.3 achieved per Figure 03 post dated 2025-10-09; UL2271 in-process per battery post.	Energy system is also a safety/conformity gate; battery proof is not just runtime.	🟢 Figure official	S4 safety bridge, not S5 economics
Figure wireless charging	Figure 03 supports wireless inductive charging through coils in the feet, 2 kW wireless charging, and wireless data offload; Figure says this enables near-continuous operation if the robot can return to the charging mat during use-case breaks.	2 kW wireless charging; 10 Gbps mmWave data offload; post dated 2025-10-09.	The key question shifts from pack size to duty-cycle orchestration: can robots opportunistically charge without killing utilization?	🟢 Figure official	S4 fleet-utilization signal
Agility Digit battery/docking	Agility announced expanded battery capabilities that run more efficiently and last up to four hours, plus autonomous docking onto a charging station.	Up to 4h; autonomous docking; announcement dated 2025-03-31.	Autonomous docking is more important than the headline runtime if it reduces human swap/charging labor and improves fleet scheduling.	🟢 Agility official	S4 operations-readiness signal
Agility safety/integration context	Same Agility announcement lists CAT1 stop, Safety PLC, on-robot E-stop, wireless teach pendant with E-stop, FSoE, Arc platform integrations, charger support, remote monitoring/support/maintenance, and MES/WMS/WES/PLC integration.	2025-03-31; customer/warehouse/manufacturing positioning.	Charging cannot be evaluated separately from safety, fleet management, workcell EMS and support burden.	🟢 Agility official	S4 integration signal
Unitree H2 / H2 Plus battery	Unitree lists H2/H2 Plus with 15Ah / 0.972 kWh battery, max 75.6V, smart quick-release battery, charger included, about 3h battery life, and 70kg weight with battery.	H2 price US$29,900; H2/H2 Plus 0.972 kWh; about 3h; pages accessed 2026-06-18.	Low-cost humanoids now expose a public energy denominator; quick-release may help utilization, but public sources do not disclose swap time, charger ratio, degradation, or fleet economics.	🟢 Unitree official	S3/S4 cost-access + energy-denominator signal
Tesla Optimus non-disclosure	Tesla Q1 2026 Exhibit 99.1 gives Optimus factory construction/capacity intent but no public Optimus pack capacity, runtime, charge rate, docking, battery degradation, or duty-cycle disclosure in the extracted SEC exhibit.	Q1 2026 Exhibit 99.1 filed 2026-04-22; construction / 1M Fremont designed line / 10M Texas long-term designed line already tracked in prior artifacts.	Capacity intent without energy/duty-cycle disclosure is still S4 manufacturing intent, not S5 deployment economics.	🟢 Tesla SEC for what is disclosed; 🟠 absence classification	S4 gap signal
Service-robot denominator	IFR says professional service robot sales were almost 200,000 units in 2024, with transportation/logistics 102,900 units and RaaS fleet +31%.	IFR World Robotics 2025 service-robot release dated 2025-10-07.	Mature service robots already operate around fleet logistics and utilization denominators; humanoids need similar fleet-energy disclosure before S5 claims.	🟢 IFR	Base-rate denominator
Dynamic-stability standard	ISO/CD 25785-1 covers actively controlled stability industrial mobile robots, including bipedal/quadrupedal/wheeled balancing robots that could become unstable without power.	New project 2025-05-22; Committee Draft registered 2026-05-08; CD consultation initiated 2026-05-12.	Energy loss is not only downtime; for actively stable robots, power and safety are linked.	🟢 ISO	S4 standards/safety signal

4. Quick derived ratios

These are estimates derived from vendor specs, not measured field performance. 🟠

Robot / pack	Public spec	Implied average draw	What this does and does not mean
Figure F.03	2.3 kWh / 5h	0.460 kW / 460 W	Useful sanity check for pack/runtime claim. It does not prove productive workload share, battery life, thermal stability, or customer ROI.
Unitree H2/H2 Plus	0.972 kWh / about 3h	0.324 kW / 324 W	Useful cost-access energy denominator. It does not prove industrial utilization, reliability, service burden, or paid deployment economics.
Figure vs Unitree	2.3 kWh vs 0.972 kWh; 5h vs about 3h	Figure pack is 2.37x capacity and 1.67x runtime	Differences may reflect robot design, workload assumptions, safety margin, compute, motion profile and disclosure style; do not rank companies from this alone.

5. The energy / charging S4-to-S5 gate

A robotics deployment should not upgrade toward S5 only because a vendor discloses “3h,” “4h,” or “5h” runtime. Upgrade quality improves only when several of these appear together:

Accepted fleet size by customer/site. 🟢
Productive robot-hours per robot per 24h, separated from idle/runtime spec. 🟢
Charge or swap time, docking success rate, charger-to-robot ratio, and queueing delays. 🟢
Utilization by shift and task class, including human supervision/intervention. 🟢
Battery cycle-life / degradation under field duty cycle, not only lab rating. 🟢/🟡
Thermal throttling, safety stop, emergency stop, and charging incident history. 🟢
Maintenance/service hours caused by battery, charger, BMS, connectors, wireless coils or quick-release mechanisms. 🟢/🟡
Customer ROI/payback after charging infrastructure, facility modification, support and battery replacement. 🟢
Vendor gross margin after battery warranty, spare packs, chargers, field support and replacements. 🟢

6. Signal vs noise

Signal

A vendor discloses pack energy, runtime, charge rate, docking/wireless/swap mechanism, safety certification status and manufacturing cost reduction together. Figure’s F.03 is the clearest current public example. 🟢
Autonomous docking or wireless charging is tied to actual fleet utilization: robot-hours per day, dock success, charger occupancy, charge scheduling and shift design. 🟢/🟡
Quick-release batteries are tied to swap time, spare-pack inventory, safety procedure, training and battery degradation. 🟢/🟡
Customer case studies disclose productive robot-hours and charging/intervention downtime, not only “runtime.” 🟢
Filings disclose warranty reserve, service cost, battery replacement, gross margin and charger/accessory economics as fleets scale. 🟢

Noise unless upgraded

“Can work a shift” without workload, productive-hour share, charge/swap time, duty cycle or intervention. 🟠
“Near-continuous operation” without charging schedule, dock access, queueing, uptime, and customer workflow fit. 🟠
“Fast charging” without battery degradation, thermal limits, safety certification, facility power needs and support cost. 🟠
“Quick release” without swap labor, spare inventory, connector durability and safety process. 🟠
“Designed capacity” for robot factories without the energy/duty-cycle disclosure needed for deployed fleets. 🟠

7. Stage classification

Current classification: S3/S4 evidence, not S5 economics. 🟠

Why:

Figure and Agility show that leading humanoid vendors are productizing charging/docking and battery safety into the deployment stack. 🟢
Unitree gives a public low-cost energy denominator for H2/H2 Plus, including quick-release battery and about-3h endurance. 🟢
Tesla provides strong Optimus manufacturing/capacity-intent disclosure elsewhere, but no public Optimus energy/duty-cycle disclosure in the reviewed Q1 2026 SEC update. 🟢/🟠
None of the reviewed public sources disclose the full fleet-energy economics needed for S5: productive hours, charger ratio, queueing, degradation, replacement cost, charging-related downtime, support labor, ROI/payback and gross margin after support. 🟠

8. What would change our mind

Upgrade signals

Figure, Agility, Unitree, Tesla, Apptronik, UBTECH, 1X or customers publish fleet utilization with productive robot-hours per day and charging/docking/swap downtime. 🟢
A named customer discloses robot count, charger count, shift schedule, charging queue, intervention rate and ROI/payback. 🟢
Vendors publish battery cycle-life / degradation curves under real humanoid duty cycles and connect them to warranty/support economics. 🟢/🟡
Safety/certification sources confirm battery/charging safety scope for humanoid applications and field deployments. 🟢
Filings disclose robot/accessory/service gross margin with enough detail to isolate battery/charger/service burden. 🟢

Downgrade signals

Runtime claims improve but field utilization remains undisclosed for 12-24 months. 🟠
Fast charging produces unacceptable battery degradation, thermal limits, safety events, charger downtime or replacement cost. 🟡/🟠
Fleets require high spare-pack inventory, manual swaps, or remote support that erodes labor-savings economics. 🟡/🟠
Customers choose simpler AMRs, conveyors, cobots or task-specific service robots because humanoid charging/service burden is too high for the same task. 🟡/🟠
Tesla/other OEMs continue capacity-intent disclosures without deployment duty-cycle evidence. 🟠

9. Common misconceptions

“Runtime equals productive labor substitution.”
- Correction: runtime is not productive robot-hours. Productive economics require charging, intervention, downtime, maintenance and task throughput denominators. 🟠
“Fast charging solves utilization.”
- Correction: fast charging helps only if it does not create degradation, thermal, safety, charger-queue, facility-power or support-cost problems. 🟠
“Wireless charging means near-continuous operation by default.”
- Correction: near-continuous operation depends on dock availability, schedule fit, charge power, task breaks, battery health and fleet management. 🟢/🟠
“Quick-release batteries make industrial deployment easy.”
- Correction: swaps add spare inventory, safety procedure, connector wear, labor, charging stations and tracking needs. 🟠
“Low-cost humanoid hardware proves low-cost labor substitution.”
- Correction: purchase price excludes battery replacement, charger infrastructure, support, downtime, warranty and customer workflow integration. 🟠
“Factory capacity intent proves fleet readiness.”
- Correction: production capacity is not deployed-unit utilization. Optimus-style capacity evidence remains incomplete without robot energy/duty-cycle disclosure. 🟢/🟠

10. Think Deeper questions

Which company will first disclose productive robot-hours per deployed robot per day instead of only pack/runtime specs?
Does the winning charging architecture look like Figure-style wireless opportunity charging, Agility-style autonomous docking, Unitree-style quick-release battery, or a hybrid by use case?
Does battery/charging favor vertical OEMs that can integrate pack, BMS, charger, safety and fleet software, or suppliers/integrators that optimize multi-vendor sites?
Will humanoids compete poorly against AMRs in high-utilization logistics tasks because AMRs already optimize charge scheduling and fleet duty cycles?
What will reveal S5 first: charger ratio, battery degradation, warranty reserve, service gross margin, or customer renewal after 6-12 months?

11. Public-safe site draft section

The hidden robotics denominator: energy utilization

Humanoid robotics is starting to publish better energy evidence. Figure says its F.03 battery is 2.3 kWh, supports 5 hours of runtime at peak performance, charges at 2 kW, and has reached UN38.3 certification on Figure 03. Agility says Digit’s expanded battery lasts up to four hours and can autonomously dock to a charging station. Unitree lists H2/H2 Plus at 0.972 kWh and about three hours of battery life with a smart quick-release pack.

That is progress, but it is not yet economics. A robot does not replace labor because a spec sheet says it can run for 3-5 hours. It replaces labor only if the fleet can schedule productive robot-hours around charging, docking, swaps, maintenance, safety stops and human intervention.

The next useful dashboard column is therefore not just “runtime.” It is duty-cycle utilization: productive hours per robot per day, charging downtime, charger-to-robot ratio, dock success, battery degradation, support cost, and customer ROI. Until those appear, battery specs are S3/S4 productization evidence — important, measurable, but not S5 scaled-commercial proof.

Evidence map only. No company ranking. No trade recommendation. Battery runtime is not productive shift economics.

12. Source list

Primary sources:

Figure AI, “F.03 Battery Development,” 2025-07-17. Used for 2.3 kWh, 5h runtime, 2 kW fast charge, 94% energy-density improvement, 78% cost reduction, torso integration, active cooling, UN/UL certification process and 23-test statement. 🟢 https://www.figure.ai/news/f-03-battery-development
Figure AI, “Introducing Figure 03,” 2025-10-09. Used for UN38.3 achievement, wireless inductive charging, 2 kW wireless charging, wireless data offload, near-continuous operation claim and home/fleet positioning. 🟢 https://www.figure.ai/news/introducing-figure-03
Agility Robotics, “Agility Robotics Announces New Innovations for Market-Leading Humanoid Robot Digit,” 2025-03-31. Used for up-to-four-hour battery, autonomous docking, CAT1 stop, Safety PLC, E-stop, FSoE, charger support, remote monitoring/support/maintenance and enterprise-system integration. 🟢 https://www.agilityrobotics.com/content/agility-robotics-announces-new-innovations-for-market-leading-humanoid-robot-digit
Unitree Robotics, “Unitree H2 Destiny Awakening.” Used for H2 price US$29,900, 70kg with battery, 31 joints, 0.972 kWh / 15Ah battery, max 75.6V, smart quick-release battery, charger, about 3h battery life, warranty and safety caveats. 🟢 https://www.unitree.com/mobile/H2
Unitree Robotics, “Unitree H2 Plus.” Used for 0.972 kWh / about-3h battery, smart quick-release battery, Jetson T5000, 2070 TFLOPS FP4, 70kg with battery, payload and remote emergency stop. 🟢 https://www.unitree.com/H2plus/
Tesla Q1 2026 Update, Form 8-K Exhibit 99.1 via SEC EDGAR, filed 2026-04-22. Used for Optimus construction/capacity-intent context and lack of disclosed Optimus battery/runtime/charging specifications in reviewed exhibit. 🟢 https://www.sec.gov/Archives/edgar/data/1318605/000162828026026551/exhibit991.htm
IFR, “World Robotics 2025 report – SERVICE ROBOTS,” 2025-10-07. Used for almost 200,000 professional service robots sold in 2024, transportation/logistics 102,900 units, RaaS fleet +31%, and service-robot denominator. 🟢 https://ifr.org/ifr-press-releases/news/service-robots-see-global-growth-boom
ISO/CD 25785-1, “Robotics — Safety requirements for dynamically stable industrial mobile robots,” Committee Draft. Used for actively controlled stability, bipedal/quadrupedal/wheeled balancing robot scope, power-loss instability relevance, project timeline and industrial-environment boundary. 🟢 https://www.iso.org/standard/91469.html
NIST, “Test Methods for the Evaluation of Manufacturing Mobile Manipulator Safety,” published 2016-04-20. Used as supporting official evidence that mobile-manipulator deployment requires metrics/test methods for functional safety and performance, not only capability demos. 🟢 https://www.nist.gov/publications/test-methods-evaluation-manufacturing-mobile-manipulator-safety

Estimate / synthesis:

Charlie calculations from vendor specs: Figure implied average draw 2.3 kWh / 5h = 0.460 kW; Unitree implied average draw 0.972 kWh / 3h = 0.324 kW; Figure/Unitree capacity ratio 2.37x and runtime ratio 1.67x. 🟠 Calculated 2026-06-18 with Python.
Charlie S4/S5 duty-cycle classification and signal/noise synthesis. 🟠

13. Public-safe flag

PUBLIC-safe as an evidence framework only. Do not include Hugo private portfolio data, position weights, purchase prices, tax/legal context, trade rationale, private channel checks, paid-report excerpts or rumors. Do not frame Tesla, Figure, Agility, Unitree, suppliers, customers, standards bodies or any public/private security as buy / sell / hold. Do not imply any company has proven S5 scaled-commercial economics until future primary/customer/filing sources disclose accepted fleet scale, productive robot-hours, charging downtime, battery degradation, service burden, customer ROI/payback, revenue and margin.