On the afternoon of October 21, Unitree released a major announcement, officially launching the H2-X "Kung Fu Legend" humanoid robot. Standing 182 cm tall and weighing approximately 70 kg, it features 31 degrees of freedom and a maximum torque of 360 N·m at the leg joints. It packs 2070 TOPS of computing power under local air cooling. The video demonstrated various high-difficulty movements and excellent performance across multiple task scenarios.
Unitree's robots showing off their capabilities has become standard practice in the industry, offering little novelty. However, this time, it breaks the "2-hour curse," with official claims of a maximum endurance of 8 hours on a single charge.
A few days prior, overseas Agility Robotics also released a message: Digit 5 has partnered with Nyobolt to implement an ultra-fast charging solution, achieving 90 minutes of endurance with just a 9-minute charge.
One focuses on the long endurance of the robot itself, while the other relies on a fast-charging infrastructure to boost the utilization rate. These two approaches provide distinctly different reference models for the energy strategies of humanoid robots.
01. Unitree H2-X: Packing 8 Hours into the Robot
The endurance upgrade of the H2-X marks a landmark breakthrough for domestic humanoid robots in single-unit battery life. The previous generation H2 had a battery life of about 2 to 3 hours, also failing to break through the two-hour industry bottleneck. The H2-X officially boasts a maximum endurance of 8 hours, perfectly aligning with the time window of a single-shift work schedule.
The core logic of this approach is a built-in long-endurance solution for the robot. It utilizes custom high-rate battery cells paired with Unitree's self-developed BMS and battery pack solutions. This setup accommodates the pulse discharge requirements for explosive movements like aerial kicks and heavy lifting, while simultaneously supporting prolonged continuous power supply. The team has also been continuously optimizing power consumption across the motors, reducers, and whole-machine motion control, extending the available endurance without compromising the bipedal center of gravity balance.
However, there is a boundary that the industry needs to view rationally: the 8-hour figure applies to light-load operating conditions. If high-load tasks such as continuous load-bearing and intense physical exertion are performed, the overall power consumption will increase significantly, and the endurance will drop accordingly.
Its target application scenarios are very clear: environments where charging docks do not need to be deployed everywhere and the robot needs to operate independently outdoors, such as campus security, field inspections, and emergency rescue. With a single charge, the robot can independently complete a full shift without needing to frequently return to a charging station.
The trade-off stems from physical constraints. Achieving longer endurance requires increasing battery capacity, which in turn increases battery weight. This directly impacts the robot's overall balance, joint load, thermal dissipation, and cost. The space for counterweights in bipedal humanoid robots is already highly constrained, meaning that simply stacking batteries to gain endurance hits a clear physical ceiling.
02. Digit 5 + Nyobolt Fast Charging: Boosting Utilization Rate Through Infrastructure
Contrary to Unitree's approach, Agility's Digit 5 does not focus on extending single-charge endurance. Its single full-charge runtime is only 90 minutes. However, paired with the Nyobolt ultra-fast charging system, a 9-minute charge restores 90 minutes of endurance, achieving a work-to-charge ratio of 10:1. This is significantly higher than the 2:1 ratio of the previous generation, Digit 4. Under ideal cycling conditions, the cumulative effective working time within a 24-hour cycle can exceed 20 hours.
The underlying logic of this approach is to shift the energy capacity from the robot itself to external infrastructure.
Nyobolt batteries prioritize optimizing charging power, focusing on ultra-high-rate fast charging rather than maximizing single-cell energy density. In standardized industrial scenarios such as warehousing and manufacturing, there are natural gaps in the robot's task flow: waiting for materials, handovers, and station switching. These fragmented idle periods serve as its charging windows. The robot automatically enters the fast-charging dock, tops up its battery in a few minutes, and immediately returns to the production line. For factory customers, the ultimate concern is how many effective working hours the robot can deliver in a single day.
The shortcoming of this approach is equally clear: it is highly dependent on supporting fast-charging infrastructure. It is well-suited for fixed industrial environments that can be retrofitted in advance to allow batch deployment of charging docks. Once disconnected from the charging network and required to perform long-duration, uninterrupted independent operations, the advantages of this solution cannot be realized.
03. Two Approaches, Three Key Takeaways
The choices made by Unitree and Agility essentially answer the same question: How exactly should the endurance of humanoid robots be defined? They now provide two distinct answers.
First, breaking the "2-hour curse" is not solely about stacking batteries. Previously, the default assumption was to simply increase battery size if endurance fell short. However, the weight, spatial, and dynamic constraints of bipedal robots dictate that this path has an upper limit. Unitree's single-unit long-endurance approach pursues independent operational capability away from infrastructure; Agility adopts a systematic energy replenishment approach, using fast charging combined with scheduling to maximize the fleet's all-day utilization rate. With both solutions running in parallel, the industry no longer relies on a single problem-solving mindset.
Second, endurance evaluation criteria must distinguish between single-charge continuous endurance and all-day operational uptime. For single-unit scenarios such as special operations, inspections, and emergency response, uninterrupted single-charge endurance is a hard metric. For large-scale B2B deployment in factories and warehouses, customers place greater value on all-day effective working hours. Digit 5 demonstrates that energy solutions for commercial deployment must be designed in tandem with business processes.
Third, future B2B competition for humanoid robots will be about competing with complete solution packages, not just the robot hardware itself. What is sold to factories in the future will no longer be just a standalone robot, but a comprehensive package comprising the robot, charging docks, and a scheduling system. Batteries, energy replenishment infrastructure, and task scheduling algorithms will become inseparable components of commercial solutions, just like joints, controllers, and large models.
Of course, it must be objectively acknowledged that neither approach has bypassed underlying physical constraints. Unitree's 8 hours will shrink in heavy-load scenarios; Digit 5's 20 hours of effective working time is predicated on ideal scheduling and the stable availability of charging infrastructure. The curse of "looking good for no more than 2 hours" has merely been mitigated by different engineering solutions and has not yet been completely broken.
At this stage in the industry's development, endurance is no longer just a cold, hard parameter. It is intrinsically tied to application scenarios and business models. In the short term, both approaches will coexist for a long period, each capturing the markets best suited to them. The next true qualitative leap in the industry will require joint breakthroughs in cell energy density, overall power consumption, and novel energy replenishment solutions.
This article is compiled based on publicly available online information and is for reference only. It does not constitute investment advice.