Boston Dynamics’ New Atlas Humanoid Robot: ‘Order Of Magnitude’ Simpler

Boston Dynamics's new Atlas robot is kicking soccer balls like a World Cup athlete. But it's also much simpler — and cheaper — to make than before.

Hyundai Motor Company

Boston Dynamics’ new Atlas humanoid robot has been wowing the world for months now, even before the company unveiled it performing unprecedented soccer acrobatics in celebration of the current World Cup. The company unveiled the fifth generation of Atlas at CES early this year, and it might just be the strongest and most capable humanoid on the planet. Atlas has always been cutting edge, but the downside has been cost: historically upwards of $200,000. Maybe not anymore. In an exclusive interview with Forbes, Director of Robot Behavior for Atlas Alberto Rodriguez told me that the new Atlas is much simpler.

It is, in fact, "almost an order of magnitude reduction in complexity compared to the previous generation," Rodriguez told me. "It has way, way less parts, and way less unique parts. And the process of manufacturing it is much faster and simpler, which leads to higher reliability and lower cost."

That is a big statement, and he knows it. Rodriguez was careful to say "almost" an order of magnitude, not a full 10x simpler.

But even a fraction of that is a key signal of a maturing product: increased capability with decreased complexity. And the resulting lower cost should make Atlas even more interesting for potential customers. Hyundai, which recently bought the remaining 20% of Boston Dynamics that it did not already own, is planning to keep most or all of this year’s production run for its own uses. But soon the company plans to build at least 30,000 Atlas humanoids a year, and others will be able to purchase it.

Why simplicity is the whole game

Right now, except for a few robot makers who are already scaling like Agibot and Unitree, humanoid robots are expensive, hand-built, and finicky. Every unique part is a supply chain to manage, a failure point to engineer around, a line item that keeps the price high.

The companies that win the humanoid race will need to have a great product, for sure. But they’ll also need to be capable of building tens of thousands of reliable units at a price that makes customers reach for their checkbooks.

Boston Dynamics appears to be optimizing for exactly that.

By stripping out parts and, crucially, reducing the number of unique parts, the company shortens manufacturing time, raises reliability, and drives down cost, all at once. Fewer unique components means simpler assembly, fewer things to break, and a cleaner path to volume production.

"We're happy that we've been able to demonstrate the same level of performance, or higher, with a robot that is fundamentally way, way simpler," Rodriguez said. "Which we think puts us in a really good place for the next step of mass manufacturing."

That last phrase — mass manufacturing — is where this gets interesting. A humanoid that matches or beats the previous generation's capability while being far cheaper and more reliable to produce isn't just a better robot. It's a better business.

That’s one of the benefits of having Hyundai as a corporate parent: this company knows how to manufacture. Hyundai Motor Group, which also owns Kia and Genesis, sells around seven million vehicles a year combined, making it one of the world’s three largest automakers. If you’re making millions of anything, chances are you’ve gotten pretty good at it.

That’s a fact that should put a dash of fear into every humanoid robot competitor.

Capability is hardware plus software, and software is now the bottleneck

But capability isn’t just about hardware.

"It’s really the combination of hardware and software," Rodriguez told me. When engineers design a new robot, they build in potential, but that potential stays theoretical until the control software grows more capable of unlocking it. I’ve heard this from other humanoid robot makers like 1X in California and Humanoid.ai in the UK as well.

"The things we can show with existing hardware are still limited by our ability to know how to control it," he said. "Atlas — but I'm sure it's the same for many other robots — is capable of much more than we've been able to squeeze performance out of it today."

In other words, the AI and control algorithms remain the main bottleneck on getting value out of robotic hardware. And humanoid robots are no different than other modern products like software-heavy cars and consumer electronics: they’ll ship with a default set of capabilities, but over-the-air software updates will unlock more.

Rodriguez frames the Boston Dynamics AI stack as two systems working together.

The first is physical intelligence, the core control loop that handles balance, agility, and physical skill: jumping, grabbing, moving objects with speed and generality. The second is reasoning intelligence, the layer that looks at a task, breaks it into steps and figures out how to do it, including things like making a judgement call on whether an object will be heavy or light.

Physical intelligence is Boston Dynamics' home turf.

For years the company has pushed the frontier of what legged robots can do, taking inspiration from dancing, parkour, and gymnastics. "That's clearly a competitive strength we have," Rodriguez said.

But over the last couple of years, the team has invested heavily in the reasoning side because that’s what unlocks generality. If a robot is deployed on a factory line and the workflow changes two weeks later, you don’t want to spend another few months reprogramming and revalidating it. "You want the robot to learn through experience or through demonstration, in a way that is more natural," Rodriguez said. Here’s the human corollary: you get hired at the factory and a coworker or boss shows you what to do. You learn it, and as things change, you adapt on the fly rather than starting from scratch.

That’s kind of a holy grail for humanoid robots right now.

Speaking 'factory language'

But there’s a much less glamorous layer that’s just as important: getting robots to talk to the systems that run a factory. Integration is one of the biggest hidden costs of deploying new technology, and a humanoid that can’t communicate with a warehouse management system is dependent on being told what to do in person, over and over again.

Here, Boston Dynamics leans on experience most of its humanoid rivals don’t have. Stretch, its wheeled-base warehouse robot, already ships with a fleet management system deployed across hundreds of customers. That management system assigns inspection routes and coordinates which robot does what, rather than requiring each unit to be programmed individually.

"That's something we've been doing for a while now," Rodriguez said.

Bringing that orchestration muscle to Atlas is a meaningful head start.

The legs-versus-wheels question

I asked Rodriguez where he lands on the industry’s favorite debate: legs or wheels. Most robot buyers I talk to are more than fine with legs, which saves battery, allows for great stability and – crucially – can fit into existing safety regulations.

But Rodriguez shared some new insights that undercut the common assumption in the factory/warehouse/logistics spaces that legs are an expensive indulgence.

First, the complexity math is closer than I would have guessed. A wheeled omnidirectional base typically uses four wheels with two actuators each — one to drive, one to steer — for eight actuators total. That’s roughly what you need for two legs, Rodriguez says.

"The mechanical complexity that goes into building legs today is not that different," he said.

Second, legs go more places, and not just up stairs. Factories look flat until you need to cross the gap between a loading dock and a trailer, or climb to a mezzanine. Third, legs enable a slimmer form factor. A mobile base has to be wide in every direction to stay stable when an arm extends, but a legged robot can orient its stance to stay thin where space is tight.

In warehouses, where every square foot is at a premium, that matters.

As for balance and locomotion, the problem that stumped roboticists for decades?

"It's actually not that difficult anymore," Rodriguez said. "We've figured out the right recipes for how to do it reliably."

(Of course, plenty of much more recently founded humanoid startups are still working on exactly that.)

From parkour to the pitch: Atlas and the World Cup

As part of its "Next Starts Now" campaign for the FIFA World Cup 2026, Hyundai Motor created School of Football, a five-part social film series following Atlas as it learns to play the game.

The series builds to Atlas pulling off a “Ghost Rabona,” an advanced cross-leg kick that demands precise timing, balance, and controlled motion well beyond simple movement replication.

The kicker, so to speak: every movement was performed by Atlas itself, with no CGI. It’s a marketing campaign, sure, but it’s also a genuine demonstration of the agility Rodriguez says is the company’s core strength: the physical intelligence that will do far more prosaic work on dusty factory floors.

The bottom line

Boston Dynamics' pitch is coming into focus.

Match or beat the previous Atlas on capability, build it with a fraction of the complexity, at higher reliability and lower cost. Layer on reasoning AI so it can be retasked without months of reprogramming and hook into orchestration software borrowed from a fleet already running at hundreds of customers.

Do it all on legs, because the complexity penalty is smaller than the industry assumes and the payoff — reach, and a slim footprint, and just a much cooler robot — is real.

The demos are always cool. (Of course!)

But the order-of-magnitude simplification is the big story. That’s what makes the difference between a hobby and a scalable, manufacturable, profitable product.