Before the hype, there was Atlas
If you've watched a video of a humanoid robot do something impressive and felt a genuine sense of "oh, this is real" — there's a good chance that video was made by Boston Dynamics, and the robot was Atlas.
Atlas is the reason that humanoid robots went from being an academic research curiosity to something that serious investors, serious companies, and serious engineers began treating as an inevitable product category. Understanding Atlas means understanding the foundation that everything else is built on.
Where Boston Dynamics came from
Boston Dynamics was founded in 1992 as a spinout from MIT, by Marc Raibert and colleagues who had been doing groundbreaking work on dynamic locomotion — the science of how machines can move like animals, using momentum and active balance rather than slow, careful, statically stable movement.
The company's early work was largely funded by DARPA, the U.S. Defense Advanced Research Projects Agency, which has a long history of funding research that seems impractical until it suddenly isn't. The robots they built in the 2000s — BigDog, a four-legged robot that could carry heavy loads across rough terrain — were extraordinary engineering achievements that got little mainstream attention at the time.
Then in 2013, they unveiled Atlas.
The original Atlas: impressive and awkward
The first generation of Atlas was developed as part of DARPA's Robotics Challenge — a competition designed to push the state of the art in disaster-response robots. The scenarios involved tasks like driving a vehicle, opening doors, climbing stairs, and operating industrial equipment. The idea was that a robot capable of these tasks could enter environments too dangerous for humans.
The original Atlas was remarkable — it could stand upright, walk over uneven terrain, and use its hands with reasonable dexterity. It was also tethered to a power supply, moved slowly, and fell over fairly often. The DARPA challenge videos of various teams' robots tumbling in slow motion became an inadvertent demonstration of just how hard the problem was.
But Atlas stood and walked and manipulated objects with two hands. That was genuinely new. That was proof of concept.
The viral videos that changed the conversation
The real cultural turning point came in 2016 and 2017, when Boston Dynamics released videos showing Atlas navigating a snowy forest trail, being shoved and maintaining its balance, picking up boxes, and most dramatically — doing a standing backflip.
The backflip video, released in November 2017, was watched tens of millions of times. Comments ranged from amazed to genuinely unsettled. Something about a machine moving with that kind of fluid, confident physicality crossed a psychological line for many viewers. This didn't look like a clumsy robot. It looked athletic.
That video did something important: it made humanoid robots feel inevitable to a general audience in a way that years of technical papers and conference demonstrations hadn't managed. It seeded the cultural awareness that made the current wave of humanoid robotics investment legible to non-specialists.
A robot doing a backflip has no practical application. What it demonstrated was dynamic balance, full-body coordination, and the ability to execute a complex movement sequence without errors. Those capabilities have enormous practical applications. The backflip was proof of the underlying system, expressed in a way anyone could understand.
A decade of iteration
What's less appreciated in the viral moments is the sustained engineering work underneath them. Atlas went through continuous major revisions over its first decade — becoming lighter, more capable, more autonomous, and less reliant on external support.
Early Atlas versions were hydraulically actuated — meaning they used fluid pressure to move, like heavy construction equipment. Later versions moved toward electric actuation, which is quieter, more energy-efficient, and more controllable for precise manipulation. By the early 2020s, Atlas could run, jump, do parkour sequences, and handle objects with increasing dexterity.
In 2024, Boston Dynamics unveiled an entirely new Atlas — fully electric, with a redesigned body and a focus on industrial applications. The new Atlas can rotate its joints beyond human range, which sounds unsettling but is actually useful: a robot doesn't need to mirror human limitations. It just needs to accomplish human-world tasks.
Who owns Boston Dynamics — and why it's complicated
Boston Dynamics has had a turbulent ownership history that reflects the uncertainty of the robotics market. Google acquired the company in 2013, apparently believing that humanoid robots would be central to its long-term strategy. By 2017, Google had sold Boston Dynamics to SoftBank, the Japanese conglomerate, apparently deciding that consumer robotics wasn't where it wanted to compete.
SoftBank sold a majority stake to Hyundai, the South Korean automotive giant, in 2021. Hyundai has invested heavily in the company and brings obvious synergies: manufacturing expertise, supply chains, and a direct interest in automating its own production facilities.
What this ownership history tells you is that the question of "when does this become a real commercial product" has been live for over a decade — and the answers have consistently been further out than optimists projected.
What Atlas is actually being used for
Contrary to what the spectacular demos might suggest, Atlas isn't being sold to consumers. Boston Dynamics' commercial products are Spot (the dog-shaped inspection robot) and Stretch (a warehouse robot arm). Atlas has been a research platform — used to push the state of the art and demonstrate what's possible, not to generate revenue directly.
The new electric Atlas is intended for industrial deployment, specifically in Hyundai's manufacturing facilities. This is a significant step: moving from demonstration to actual operational use in a demanding environment. The results of that deployment will tell us a lot about whether Boston Dynamics' hardware can survive the transition from impressive demo to reliable tool.
Why Atlas matters even if you can't buy one
Here's the thing about Atlas: it has functioned as a kind of benchmark for the entire humanoid robotics industry. Every time Boston Dynamics releases a new video, every other robotics company measures themselves against it. The capabilities Atlas demonstrates set the expectation for what a serious humanoid robot should be able to do.
More concretely, Atlas has been the training ground for a generation of roboticists. The engineers who worked on Atlas, and the research that came out of those years, are now distributed across the industry — at Tesla, at Figure, at Agility, at startups you haven't heard of yet. The intellectual capital built at Boston Dynamics over three decades didn't stay at Boston Dynamics.
Atlas didn't just build a robot. It trained an industry. The people and ideas that came out of Boston Dynamics are woven into every serious humanoid robotics effort happening right now.
The honest legacy
Boston Dynamics proved that a bipedal machine could move with speed, stability, and physical versatility that earlier generations of roboticists thought was decades away. They proved it with hardware, not with theoretical papers. They proved it repeatedly, across a decade of improvements, in conditions that were increasingly demanding.
What they didn't do — at least not yet — is turn that capability into a commercial product that regular businesses or consumers can buy and use. That challenge remains. But the foundation they laid is the reason the current humanoid robot boom is happening at all. Atlas didn't win the race. It built the road.