A Field That Used to Belong to the Universities
For most of the humanoid robot’s existence, it was a creature of the laboratory, not the marketplace. Honda’s ASIMO, unveiled in the year 2000 and quietly retired in 2018, spent nearly two decades as the most recognizable humanoid on Earth without ever being sold to a single customer. Boston Dynamics itself began life in 1992 as an MIT spinoff, its early funding coming almost entirely from DARPA rather than from paying clients, its most famous feats of parkour and backflips built to win robotics competitions rather than to earn a return on investment. The 2015 DARPA Robotics Challenge, in which university and government-funded teams sent humanoid robots stumbling through simulated disaster response tasks, was in many ways the last great showcase of research-driven humanoid development as the field’s center of gravity. For the better part of thirty years, if you wanted to see the most advanced humanoid robot in the world, you looked toward a university lab or a defense research budget, not a factory or a store shelf.
That era is ending, and it is ending quickly.
The Companies Took Over
What has changed is not that research stopped mattering, but that commercial companies, backed by manufacturing budgets research labs could never match, began outpacing academic institutions on the one thing that ultimately decides whether a humanoid robot becomes real: the ability to build many of them, reliably, at a cost someone will actually pay. Tesla’s entry into the field in 2021 marked an unmistakable turning point, treating the humanoid not as a research question to be solved once and displayed at a conference, but as a manufacturing problem to be solved at automotive scale. Figure, Agility Robotics, Apptronik, 1X, and a fast-growing list of Chinese manufacturers led by Unitree followed with the same underlying premise: whoever can turn a humanoid into a product, not a prototype, wins the field. Boston Dynamics itself, the last great research-lineage holdout, was acquired by Hyundai and has since leaned into automotive-style production supply chains for its electric Atlas platform, a fairly clear signal about where even the most storied research pedigree in the industry now believes its future lies.
The practical result is a field where the newest, most capable machines increasingly come from company balance sheets and investor rounds rather than government grants, and where the pace of iteration has sped up dramatically as a result. A university lab might spend years refining a single research platform. A company racing quarterly production targets against several well-funded rivals cannot afford to move at that pace, and it shows in how quickly generation after generation of industry humanoids has been arriving.
The Price Collapse Nobody Expected This Fast
Perhaps the most striking evidence of this shift is what has happened to price, and how quickly it has happened. As recently as 2024, the average selling price of a humanoid robot sat above one hundred and fourteen thousand dollars, a figure squarely in the territory of a specialized research instrument or a low-volume industrial tool. By 2026, that average has already fallen dramatically, with industry analysts at IDTechEx projecting it will continue sliding toward roughly thirty seven thousand dollars by the end of the decade.
The lower end of the market has moved even faster than the averages suggest. Unitree’s R1 model, launched in the middle of 2025, arrived at a price point under six thousand dollars, a figure that reportedly forced nearly every competitor in the industry to reconsider its own roadmap. Its sibling, the G1, offers considerably more sophisticated manipulation for roughly thirteen to sixteen thousand dollars. Tesla’s Optimus is targeting a similar entry price point of twenty to thirty thousand dollars once mass production genuinely ramps up, while enterprise-grade platforms from Boston Dynamics, Agility, and Apptronik still command anywhere from fifty thousand to well over three hundred thousand dollars for their most capable configurations. Analysts at Goldman Sachs now believe that continued scaling could push per-unit prices as low as fifteen to twenty thousand dollars, and forecasts from Bank of America and other research houses broadly agree that a forty to fifty percent decline in average prices between now and 2030 is a realistic, not optimistic, expectation.
The underlying driver of this collapse is refreshingly unglamorous: actuators, the precision motors that move every joint in a humanoid’s body, typically account for thirty to forty percent of the total hardware cost of building one. As manufacturers invest in custom actuator designs and, crucially, produce them in far greater volume, the cost of the single most expensive component in a humanoid robot is falling in exactly the way automotive component costs fell a century ago once true mass production took hold.
Renting a Robot Instead of Buying One
The falling sticker price of the hardware itself is only part of the affordability story. A second, arguably more consumer-friendly shift is unfolding alongside it: the emergence of humanoid robots available as a service rather than a purchase. Figure AI now offers its robots on a subscription-style model priced at roughly one thousand dollars a month, covering hardware, software updates, maintenance, and support in a single package rather than a large upfront capital outlay. 1X’s Neo, a robot explicitly designed for the home rather than the warehouse, is available for twenty thousand dollars outright or, notably, for as little as four hundred and ninety nine dollars a month, a price point that begins to resemble a premium home appliance subscription more than an industrial capital purchase. Industry analysts expect this rental model to fall even further, with some projecting entry-level humanoid access at around three hundred dollars a month within the next year, a figure that would put a working humanoid robot within reach of small businesses and individual households that could never have justified the cost of outright ownership.
Becoming Ordinary
None of this means humanoid robots have become simple appliances overnight. Integration, training, maintenance, and compliance costs still add a meaningful premium on top of the sticker price for any serious deployment, and the most capable platforms on the market remain firmly enterprise-priced tools rather than impulse purchases. But the trajectory is unmistakable and, by the standards of any prior wave of robotics, unusually fast. A category of machine that spent three decades as an expensive research curiosity, built one at a time inside university labs and defense-funded programs, has in the space of only a few years become something companies are racing to sell by the tens of thousands, at prices that keep undercutting the assumptions of the year before. Research humanoids opened the door and proved the concept could walk. Industry humanoids are the ones now racing each other through it, and the price of admission is falling faster with every generation that arrives.