Engineering Analysis · Humanoid Robotics

The Humanoid Actuator Battle: Calculating Force for Robot Joints (Linear vs. Rotary Face‑Off)

Everyone is obsessing over the AI brain. I have spent 20+ years building the muscles. Here is the math that Tesla, Boston Dynamics, and LG do not want you to see.

The robotics industry is at war over a component most people have never heard of. While the media breathlessly covers every new AI benchmark and every ChatGPT-powered demo video, a far more consequential battle is being fought inside the joints of every humanoid robot on the planet. The battle is over actuators. And the companies that get this wrong will not survive the decade.

At CES 2026, LG unveiled its dedicated actuator brand LG Actuator AXIUM, betting that its 41 million motors manufactured annually for washing machines can translate to robot muscles. Hyundai Mobis confirmed its actuators are now installed in the next-generation Boston Dynamics Atlas. Samsung Electro-Mechanics publicly announced it is evaluating entry into the actuator market. And on January 21, 2026, Tesla began mass production of the Optimus Gen 3 at its Fremont factory, featuring hands with 50 actuators and 22 degrees of freedom.

This is not a software race. This is a hardware war. And I have been building the weapons for over two decades.

Industry analysts now project the global humanoid robot market will exceed $15 billion by 2030. Within that figure, actuators dominate the bill of materials, accounting for over 51% of hardware cost according to recent research from Roots Analysis. A professor at Hanyang University put it bluntly at CES: if actuator technology is not secured, commercialization of humanoids is impossible.

I founded Firgelli Automations in 2002. Before that, I was designing systems at BMW, Ford, and Isuzu. I have spent my entire career converting rotational energy into linear motion and back again. If you are new to actuator technology, I wrote a comprehensive primer that covers the fundamentals: The Complete Guide to Humanoid Robot Actuators. Read that first if you need to get up to speed.

This article goes deeper. We are going to do the actual physics. I am going to put rotary and linear actuators head-to-head, show you exactly how much force it takes to make a 200 lb robot walk, and give you an interactive calculator so you can run the numbers for your own builds.

The Architecture: Where Each Actuator Type Belongs

You cannot simply bolt off-the-shelf motors onto a metal skeleton and call it a humanoid. The engineering decisions about which actuator goes where are what separate a robot that walks from one that falls on its face. Every joint on a humanoid body has different demands for range of motion, torque density, speed, and shock resilience.

Rotary Actuators: Hips, Shoulders, and Torso

For joints that need a wide or continuous range of motion, the industry relies on highly integrated rotary actuators. These are typically a frameless Brushless DC (BLDC) motor paired with a harmonic drive or planetary gear reducer. The hip joint on the Boston Dynamics Atlas, for example, needs to swing the entire leg forward and backward through a large arc while simultaneously handling abduction and rotation. A rotary actuator sitting directly on the pivot is the only architecture that makes sense here.

Rotary actuators deliver predictable torque curves and allow limbs to swing freely. But they have a critical weakness: they are terrible at absorbing shock. A harmonic drive has thin, flexible steel teeth. When a 90 kg robot lands from a jump, the instantaneous spike load through a harmonic drive at the knee can exceed its rated torque by 5x to 10x. That is how you strip gears and destroy a $2,000 component in a fraction of a second.

Linear Actuators: Knees and Ankles

Now look closely at the legs of the Tesla Optimus or the electric Atlas. Below the hip, the architecture changes dramatically. The knees and ankles use high-power electric linear actuators, most commonly driven by planetary roller screws or ball screws.

This is not an arbitrary design choice. Linear actuators are used in the lower extremities for two reasons that I have been preaching to engineers for years. First, slender profiles: a linear actuator can be mounted parallel to the structural members of the leg, like a human calf muscle running alongside the tibia. This keeps the leg slim enough to fit through doorways and navigate human environments. Second, and far more importantly, shock absorption: a planetary roller screw mechanism is vastly superior at absorbing the massive kinetic energy generated by a 90 kg robot running, jumping, or falling. The screw distributes load across dozens of contact points simultaneously. A harmonic drive concentrates it on a few thin teeth.

The human body figured this out through 300 million years of evolution. Your knee is not a rotary motor. It is a linear lever driven by a tendon. The quadriceps muscle pulls linearly, and the patellar tendon converts that pull into rotation around the knee joint. The most successful humanoid robots mimic this biomechanical architecture exactly.

The Math: The Actuator Face-Off

Let us reverse-engineer a knee joint. I am going to use realistic parameters: a total robot mass of 90 kg (the approximate weight of the new Boston Dynamics Atlas, and very close to the 57 kg Tesla Optimus when carrying a 33 kg payload). The lower leg (knee to ankle) is approximately 0.4 meters long. When this robot bends its knee to 90 degrees to pick up a box, gravity creates a massive torque demand at the joint.

The critical distinction between the two actuator types starts here. A direct-drive rotary motor is the joint. It sits at the pivot and must directly resist the gravitational torque created by the mass of the limb and everything above it, acting through the limb length at the given joint angle. A linear actuator, by contrast, is mounted offset from the pivot and pushes or pulls against a short lever arm to generate that same torque. This mechanical disadvantage means the linear actuator must produce far greater raw force, but it does so from a protected, compact position.

That is the fundamental trade-off. The rotary motor must produce enormous torque and it sits directly at the point of maximum mechanical stress, fully exposed to every impact and shock load. The linear actuator needs even greater raw force due to its short lever arm, but it delivers that force from a compact, protected, shock-resistant package that mimics actual human biomechanics. This is not theoretical. This is the engineering reality that Tesla, Boston Dynamics, and every serious humanoid company has converged on.

Interactive Tool: Rotary vs. Linear Joint Calculator

I built this calculator so that engineers, researchers, and students can run their own numbers. Whether you are designing a humanoid robot knee, an exoskeleton joint, or an industrial lifting mechanism, plug in your specifications and see the exact trade-off between rotary torque and linear push force.

Both actuator types are calculated for the same scenario: the same mass, the same limb, the same joint angle. The rotary motor sits directly at the pivot and must resist the full gravitational torque. The linear actuator is mounted offset from the pivot and must produce enough push force through its lever arm to generate that same torque.

Why the Industry Is Getting It Wrong

I have watched this industry evolve from hydraulic nightmares to the current electric gold rush. And I am going to say something that will probably annoy a few companies: the current trend of trying to cram rotary harmonic drives into every single robot joint is a fundamental engineering mistake.

I understand why it is happening. Harmonic drives are elegant, well-understood, and available off the shelf from companies like Harmonic Drive AG and Schaeffler. When a startup gets $100 million in venture capital and needs to ship a demo robot in 18 months, the path of least resistance is to put harmonic drives everywhere and worry about durability later. That is fine for a demo. It is a disaster for a production robot that needs to work an 8-hour shift in a warehouse.

Look at the robots that actually work. The electric Boston Dynamics Atlas uses linear actuators in its lower extremities. The Tesla Optimus uses roller screws. These are companies with thousands of engineers who have converged on the same conclusion I reached 20 years ago: you need linear motion in the legs. A rotary drive will fail under the dynamic shock loads of a 200 lb robot doing real-world work. A linear actuator built around a planetary roller screw absorbs that kinetic energy across dozens of threaded contact points simultaneously. It is not even close.

The new entrants flooding into this space, including LG, Samsung, and dozens of Chinese manufacturers, need to understand this before they waste billions on the wrong architecture. If you are building actuators for humanoid robot legs and you are not building linear actuators with roller screw or ball screw mechanisms, you are building components that will fail in the field. Full stop.

As this industry scales from lab prototypes to 100,000-unit production runs, and Tesla is publicly targeting 1 million units per year, the winners will be the companies that engineer for durability and total cost of ownership, not just for the demo video. The actuator is not a commodity. It is the most critical, most expensive, and most failure-prone component on the entire robot. The companies that treat it as an afterthought will be acquired or liquidated by 2030.

Where the Market Is Heading

The next 36 months will define the humanoid actuator supply chain for the next three decades. Here is what I see coming.

First, vertical integration will win. Tesla is already manufacturing its own actuators in-house. This is the same playbook they used with battery cells: control the most critical component and you control your own destiny. Companies that rely on third-party actuator suppliers will be at the mercy of allocation decisions and price increases.

Second, the Korean supply chain will become dominant. LG, Hyundai Mobis, and Samsung Electro-Mechanics are bringing decades of precision manufacturing experience from consumer electronics into the actuator space. When LG says it manufactures 41 million motors per year, that is not marketing. That is a manufacturing capability that no robotics startup on earth can match.

Third, and this is the prediction I will stake my reputation on: linear actuator technology will become the primary differentiator in humanoid performance. The rotary actuator problem is largely solved. Harmonic drives, cycloidal drives, planetary gears — these are mature technologies. But high-performance, compact, affordable linear actuators for dynamic bipedal locomotion are still an unsolved problem at scale. The company that cracks high-volume manufacturing of reliable roller screw linear actuators under $100 per unit will own the market.

Run Your Own Numbers

I built the calculator above because I am tired of watching robotics engineers make actuator selection decisions based on gut instinct and supplier marketing decks. The physics does not care about your brand preferences. If you are designing a joint, whether it is for a humanoid robot, an exoskeleton, a prosthetic limb, or an industrial automation system, run the numbers. See what the math actually demands. Then make your decision.

If you need actuators that deliver real-world force, browse Firgelli Automations' line of Micro Actuators for tight-tolerance robotics, or our Super Duty Actuators capable of pushing thousands of pounds. We have been building these for two decades while the rest of the industry was still drawing humanoid robots on whiteboards.