Payload vs Runtime: The Humanoid Trade-off Nobody Lists

Kepler K2 and Galbot G1 are the only shipping humanoids we track rated for a full 8-hour shift on one charge. They get there with 2,330 Wh and 1,440 Wh batteries in bodies of 75 and 92.5 kg. Atlas carries twice the K2's payload for half the runtime. Agility's Digit 5 goes the other way entirely: 90 minutes per charge and a 9-minute recharge. Payload and runtime pull against each other on every humanoid, and the spec sheet almost never shows the cost of either.

All figures below are from our catalog, last verified on 26 September 2026, and are manufacturer ratings. Of the 26 humanoids we list as shipping, 13 do not publish either a payload or a runtime figure, and 19 do not publish a charge time. The comparison covers the ones that do.

Eleven robots, four numbers

| Robot | Payload | Runtime | Body mass | Battery | Wh per kg of body | Charge time | |---|---|---|---|---|---|---| | Atlas (Electric) | 30 kg | 240 min | 90 kg | not published | not published | self-swapping battery | | 1X NEO | 25 kg | 240 min | 30 kg | 842 Wh | 28.1 | 24 min | | Figure 03 | 20 kg | 300 min | 60 kg | not published | not published | not published | | Digit | 16 kg | 240 min | 65 kg | not published | not published | not published | | Kepler K2 | 15 kg | 480 min | 75 kg | 2,330 Wh | 31.1 | 60 min | | AgiBot A2 | 15 kg | 120 min | 69 kg | 700 Wh | 10.1 | not published | | Galbot G1 | 10 kg | 480 min | 92.5 kg | 1,440 Wh | 15.6 | 210 min | | Unitree H2 | 7 kg | 180 min | 70 kg | 972 Wh | 13.9 | not published | | PAL TALOS | 6 kg | 90 min | 95 kg | 1,080 Wh | 11.4 | not published | | Unitree G1 | 2 kg | 120 min | 35 kg | 421 Wh | 12.0 | 105 min | | Digit 5 (2027) | 22.7 kg | 90 min | 129 kg | not published | not published | 9 min |

Read the table by column and the pattern is clear. The four robots with the highest payload (Atlas, NEO, Digit 5, Figure 03) all run five hours or less. The two robots that run eight hours carry 15 kg and 10 kg.

What eight hours costs

An 8-hour runtime is the number procurement teams ask for first, because it maps to one shift. Two robots claim it, and both pay for it in mass.

Kepler K2 carries a 2,330 Wh battery, the largest among the shipping robots in the table. That works out to 31 Wh per kilogram of body, three times the AgiBot A2 and more than twice the Unitree H2. Its average draw over 8 hours is about 290 W. Payload is 15 kg, half of Atlas. Galbot G1 gets the same 480 minutes from a smaller 1,440 Wh pack, an average draw of about 180 W. The reason is its body: Galbot G1 is a semi-humanoid on a wheeled base. It does not spend energy balancing on two legs. It is also the heaviest robot here after TALOS and Digit 5, at 92.5 kg, and carries 10 kg. The K2 vs Galbot G1 comparison shows the two approaches side by side.

At the other end, PAL TALOS has a 1,080 Wh battery and lasts 90 minutes, an average draw near 720 W. TALOS is a torque-controlled research platform with torque sensing in every joint and a 2 kHz control loop. Energy efficiency was never the design goal, and the number shows it.

The lesson is that battery size alone predicts little. Locomotion type, control approach and duty profile decide how far a watt-hour goes.

Charge time is the missing column

Runtime without charge time says little about availability, and 19 of 26 shipping humanoids do not publish a charge time. Where both numbers exist, the share of the day a robot can spend working varies widely.

Unitree G1 runs 120 minutes and charges for 105. With one battery it works about 53% of the time. Galbot G1 runs 480 minutes and charges for 210, or about 70%. Kepler K2 runs 480 and charges in 60, about 89%. 1X NEO runs 240 minutes and charges in 24, about 91%.

Digit 5 lands on the same 91% with a very different profile: 90 minutes of work, 9 minutes on the charger. We covered that design choice in Digit 5 vs Digit. A short cycle keeps the battery small and the robot lighter per kilogram of payload than a long-runtime design would be. The cost moves elsewhere: chargers near every work cell and a fleet schedule built around docking every hour and a half.

Battery swapping is the third route. Boston Dynamics says Atlas walks to a station and swaps its own battery, with 4 hours per pack. UBTECH's Walker S2 has a hot-swappable battery. With swapping, runtime per pack matters less than swap time and the number of spare packs, and neither maker publishes those.

Payload per kilogram of robot

The other ratio worth computing is payload divided by body mass. It shows how much of the robot's weight turns into useful lifting.

1X NEO stands out at 0.83: 25 kg of rated payload on a 30 kg body. That is a manufacturer claim for a home robot that, as of mid-2026, had no independently verified customer delivery. Treat it as a design target until third-party tests appear. The Atlas vs NEO comparison sets it against the heaviest lifter in the table.

Among industrial robots, Atlas and Figure 03 both reach 0.33. Digit is at 0.25, AgiBot A2 at 0.22 and Kepler K2 at 0.20. Galbot G1 and Unitree H2 are at about 0.1, and TALOS and Unitree G1 at about 0.06.

Digit 5 drops to 0.18 from Digit's 0.25. Agility added 64 kg of body to gain 6.7 kg of payload. Some of that mass likely goes into the structure needed to work next to people without a fence, but Agility has not itemised it.

Payload times hours: one number for a shift

Multiplying payload by runtime gives a rough measure of how much lifting capacity a robot delivers per charge. It ignores speed, reach and cycle time, but it puts long-runtime and high-payload designs on one scale.

Atlas and Kepler K2 tie at 120 kg-hours: 30 kg for 4 hours, 15 kg for 8. NEO and Figure 03 follow at 100. Galbot G1 reaches 80, Digit 64, AgiBot A2 30, Digit 5 about 34 per charge, Unitree H2 21, TALOS 9 and Unitree G1 4.

The tie between Atlas and K2 is the useful result. Two very different robots deliver the same capacity per charge, one by lifting more and one by running longer. The right choice depends on the task. If the heaviest item in the process is 25 kg, the K2 cannot do the job however long it runs. If the process is light and continuous, Atlas needs a battery swap halfway through every shift.

The K2 vs AgiBot A2 comparison shows the same effect at equal payload: both carry 15 kg, and the K2 runs four times as long on a battery 3.3 times larger.

How to use these numbers when buying

Start from the heaviest object in the process, not from the runtime. A robot that cannot lift the heaviest tote needs a person next to it, and then runtime does not matter.

Then ask for three things the spec sheet usually omits: the charge time, the workload used to measure runtime, and whether the battery can be swapped and by whom. A 240-minute figure measured walking empty and a 240-minute figure measured carrying 15 kg are different robots.

Finally, turn runtime and charge time into availability per day and price that into the contract. Our warehouse TCO breakdown shows how availability changes cost per task, and the explore section collects our cost and comparison tools.