The 2020s, when machines that had left the laboratory first squared off against one another on factory floors and on stages. A map of the giants.
If Part 1 covered half a century of laboratories and contest grounds, the stage for this part is the factory floor, the exhibition hall, and the social-media timeline. As the 2020s began, humanoid robots multiplied so quickly that a new machine seemed to be unveiled almost every month, and behind that surge stood the two changes noted at the close of Part 1: cheap, powerful electric drivetrains, and an artificial intelligence that had finally become usable.
Yet the fact that so many names appeared at once does not mean they are all chasing the same goal. One bets on a mass market of millions of units, another tries to seize first the artificial-intelligence model that will serve as the robot's "brain," and still another aims to shake the market with overwhelming price and volume, so in this part we will divide the major giants competing right now into camps and lay out, like a single map, where each one is placing its bet.
The trigger for this explosion was, surprisingly, not a robotics company but a carmaker. When Tesla first hinted at its humanoid plan at "AI Day" in 2021, with a video of a person dancing in a white suit, few took it seriously, but the mood shifted quickly as the company showed an actually-walking prototype the following year and unveiled a far more refined second-generation Optimus at the end of 2023.
Tesla's rationale was at once simple and provocative. The plan was to transplant directly into a human-shaped machine the motors and batteries it had built up while mass-producing electric cars, together with the artificial intelligence honed in self-driving, and eventually to stamp them out by the millions like automobiles; by first deploying them in its own factories for simple repetitive work and driving the unit cost down to somewhere around twenty thousand dollars before widening to the general market, this was the first serious declaration of "the robot as a consumer product" rather than a single research unit.
The distance between a staged demonstration and actual mass production is of course long, and it is true that announced timelines have slipped again and again. Even so, the symbolic weight Optimus carries is far from small, because the moment one of the world's most automation-capable manufacturers declared a humanoid to be "a core product of the future," other companies and investors could no longer dismiss the field as mere fantasy.
If Tesla pushed forward on the strength of manufacturing, another camp sought to win on the robot's "brain." At its head is Figure, a startup founded in 2022 by Brett Adcock, which drew quick attention by unveiling its second machine, Figure 02, in 2024 and deploying it on a trial basis at a BMW plant in the United States.
Its moves around artificial intelligence were especially interesting. Figure once joined hands with OpenAI to experiment with giving robots conversational and perceptual ability, but it later wound down that collaboration and turned toward developing its own artificial-intelligence model in-house, a choice that grew out of the judgment that even if the bodies of robots eventually converge, the real moat lies in who controls the intelligence that moves those bodies.
Challengers sharing a similar "software-first" philosophy are not few. Texas-based Apptronik worked with Mercedes-Benz and NASA through its Apollo machine; 1X, which began in Norway, professes a humanoid for use in the home and has taken investment from OpenAI; and Oregon's Agility Robotics has actually deployed its two-legged robot Digit in warehouses including Amazon's, and precisely because they focus on "carrying boxes in a warehouse today" rather than on dazzling stage demonstrations, this camp may fairly be called the most pragmatic of all.
Boston Dynamics, one of the protagonists of Part 1, had to redesign itself in this new phase. The hydraulic Atlas that had long been the company's symbol was unrivalled in power and dynamism, but it was heavy, expensive, and prone to leaking fluid, which left it far from mass production.
And so in 2024 Boston Dynamics announced the retirement of the symbolic hydraulic Atlas while at the same time unveiling a fully redesigned, electrically driven Atlas. The new machine drew attention for rotations more flexible than the human range of motion, an extension of the same pragmatism seen in Part 1's HUBO, which chose motions more advantageous to the machine over an exact imitation of a human. Now that Hyundai Motor Group has acquired the company and become its parent, the task that remains is to carry its world-class motion-control technology out of the research demo and onto the actual manufacturing floor.
While American firms across the Pacific competed on technology and capital, China came forward with an entirely different weapon: price and volume. Hangzhou's Unitree carried the mass-production experience it had built up with quadruped robots over to bipedal walking and, in 2024, released the compact humanoid G1 at the startling price of around sixteen thousand dollars, a figure on a wholly different order of magnitude from the existing research humanoids that ran into the hundreds of thousands.
Videos of Unitree's robots dancing on stage or performing martial-arts moves became a repeated sensation on social media, and in early 2025 a troupe of them even took the stage of China's largest holiday broadcast to perform a group dance. Yet flashy demonstrations are not the whole story. Chinese firms including UBTech are deploying humanoids on a trial basis in their own automobile and electronics factories and rapidly accumulating real-world data, and this war of volume, interlocking a formidable component supply chain, wholehearted government support, and an enormous domestic market, becomes an ever more threatening weapon the more the technology gap narrows.
Optimus and mass production, Figure and artificial intelligence, Boston Dynamics and motion control, and Unitree and price. Today's giants are each betting on a different place, yet the outward appearance of their machines is, if anything, growing steadily more alike, and the broad outline of a human-sized machine that walks on two legs, grips with five fingers, and carries cameras on its head has effectively converged into one.
If so, the real contest will in the end be decided inside that shell. What matters is which motors and reducers move the joints, what holds the balance and perceives objects, and what battery keeps all of it running for hours on end. In the next part we will dissect the humanoid one layer at a time, in the language of technology, to see what actually happens beneath that smooth exterior.