Walk the floor at a robotics show in 2026 and the humanoids get the crowd and the keynote stage too. The 2026 Robotics Summit and Expo, Boston, held May 27 and 28 at the Menino Convention and Exhibition Center, gathered more than 5,000 robotics developers, over 200 exhibitors, and 70-plus speakers across tracks on AI, design and development, enabling technologies, healthcare, and logistics. Humanoids ran through all of it. The “State of Humanoids” keynote panel alone brought together leaders from Agility Robotics, Boston Dynamics, Schaeffler, RealSense, and ASTM International, and the question in the room was no longer whether bipedal robots can fold laundry or sort totes on camera. They can. The harder question, the one that decides whether any of this reaches a factory floor at scale, is how long the robot keeps working before it has to stop.
That answer still measures in single-digit hours. And for engineering and procurement teams evaluating these platforms, the humanoid robot battery is the constraint that quietly governs everything else.
It is not only a humanoid problem. Consider Boston Dynamics’ Spot, the most widely fielded legged robot on the market and a useful reality check. Even this mature, commercially deployed quadruped runs about 90 minutes on a 605 Wh pack and solves continuous operation through a hot-swappable battery. If the most refined legged platform in service still lives inside a 90-minute window and leans on swapping, the runtime math facing newer humanoids is not a temporary growing pain. It is the shape of the problem.
The demo runtime and the deployment runtime are not the same number
Most humanoid platforms on the market today run between two and four hours on a charge, with pack capacities generally under 2 kWh. The reasons are physical, not a matter of picking a bigger cell. A humanoid has to carry its energy inside a human-shaped body and balance that mass over two feet, which limits both the volume available for a pack and where that pack can sit without wrecking the robot’s center of gravity.
Two hours of dynamic operation is not a spec sheet footnote. It is the difference between a robot that covers a shift and a robot that needs a babysitter and a charging queue. Industrial buyers expect uptime in the 95 to 99 percent range. A platform that works for ninety minutes and then sits is a long way from that number.
The platforms on the Boston floor told the same story. AgiBot, one of the humanoid OEMs exhibiting, lists its A2 humanoid at roughly two hours of runtime with support for battery swapping. Its newer wheeled G2 platform pushes to around four hours and leans on dual hot-swappable packs specifically so it can keep running through shift changes. The pattern is consistent across the category: raw runtime sits in the two-to-four-hour band, and the credible answers to “how do we run all day” are swapping and smarter pack design, not a single heroic battery.
There is also a subtler point that the better operators in this space already understand. Pras Velagapudi, Agility Robotics’ chief technology officer, has framed it well in industry commentary: the goal is not maximizing continuous runtime in the abstract. It is balancing the charge a robot takes on against the work it does, so the unit stays productive inside the real cadence of the task. That reframing matters, because it moves the conversation away from “biggest battery wins” and toward power-system design that fits a duty cycle. That is a battery-engineering problem, not a marketing one.
Why bigger cells alone won’t fix it
The obvious instinct is to wait for chemistry to bail everyone out. Higher energy density is coming, and silicon-anode and eventually solid-state cells will help. But two things temper that optimism for anyone planning a deployment in the next few years.
First, the timeline. Solid-state options are widely expected to reach commercial viability later this decade, and even a doubling of energy density would still leave many humanoids short of a full eight-hour shift under heavy load. Chemistry is a tailwind, not a finish line.
Second, and more important, the bottleneck is not only how much energy the pack holds. It is how the pack behaves under stress. Humanoid motion produces frequent, sharp transient loads as the robot walks, lifts, and corrects its balance, with peak draw that can spike into the kilowatts. The traits that decide whether a robot can survive a real duty cycle are peak current headroom, the guardrails the battery management system enforces, and thermal margin. A pack with impressive nameplate capacity that trips its protections under a lifting transient, or that heats up and throttles, is not a usable pack. It is a press release.
Where the real engineering happens
If the headline is “battery problem,” the substance is four engineering decisions that rarely make the keynote but determine whether a humanoid program reaches production.
Pack architecture for a hostile form factor.
A humanoid pack has to deliver high burst power from an awkward, weight-sensitive envelope. That favors purpose-built modular designs over off-the-shelf low-voltage packs borrowed from other applications. Modularity also simplifies maintenance and shortens downtime, which is the metric buyers actually pay for.
BMS as the safety and longevity layer.
A capable battery management system does more than count electrons. It holds the line on peak current, manages cell balancing, and protects the pack thermally so the robot can sustain repeated transients without faulting or degrading prematurely. As onboard compute and sensing grow, the BMS becomes more central, not less.
Thermal management under sustained load.
Heat is the quiet killer of both runtime and pack life. Dense packs in a sealed humanoid torso, running high transient loads, generate heat with nowhere easy to send it. This is exactly where advanced thermal approaches, including immersion-style cooling, move from nice-to-have to enabling technology. Keeping cells inside their optimal temperature window protects capacity, extends cycle life, and holds performance steady deep into a shift.
Operational uptime by design.
The most credible path to multi-shift operation today pairs incremental density gains with operational workarounds, chiefly hot-swappable packs that let a unit keep working while a depleted module charges off the robot. AgiBot’s dual hot-swap approach on the G2 is one example of this thinking in production. Designing for swap is a pack-level and connector-level decision that has to be made early, not retrofitted.
What this means for teams evaluating humanoids
For engineering and procurement teams, the practical takeaway is to interrogate the power system as hard as the AI: runtime under real load, peak current behavior, thermal performance across a full shift, pack serviceability, and whether the architecture supports swapping. Those answers live in how the battery and BMS were engineered, not in how smoothly the robot walks on stage.
This is the layer TWS Technology Mobile Power Solutions works in, and it maps directly onto the four decisions above. TWS is a lithium-ion and LiFePO4 battery pack OEM whose custom battery solutions are engineered to a program’s specific voltage, capacity, and form factor, which is the whole battle inside a weight-constrained humanoid torso. TWS is chemistry-agnostic, so a high-density cell can be specified where a humanoid needs runtime and a more stable chemistry where a service robot needs cycle life, instead of forcing one compromise. Its in-house BMS platforms, backed by proprietary IP and patents, are where peak-current guardrails, cell balancing, and protection get tuned to the actual duty cycle. For the heat problem, TWS develops immersion cooling to hold cells inside their optimal temperature window under sustained load. And because a humanoid pack has to clear safety and regulatory review before it ships, TWS runs in-house test labs for performance, environmental, and safety validation and manages global certification directly.
That work is delivered through a global production network across China, Indonesia, and Thailand, coordinated by a single unified operating system so a custom program stays consistent at volume, with North American customers supported through the TWS customer care and logistics center in Las Vegas, Nevada.
Humanoids will get smarter. The harder, less glamorous work is making sure they can stay on long enough to use that intelligence. That work starts with the battery.
Engineering or sourcing power for a humanoid or advanced mobile platform? TWS works directly with OEM engineering teams on custom pack design, BMS, and thermal strategy. Reach out to start a conversation about your duty cycle and runtime targets.
Powering critical applications across six core markets, TWS Technology Mobile Power Solutions delivers advanced lithium-ion battery solutions—from standard packs to fully customized systems built to your exact needs.
Interested in the full technical dossier? Get in touch with our team: info.mps@tws.com