From C-3PO’s polished diplomacy to R2-D2’s battlefield heroics, robots have long captured our imagination. Today, what was once confined to science fiction is inching toward industrial reality. General-purpose robots, powered by increasingly capable embodied AI, are being tested in warehouses, factories, hospitals, and fields.1 And unlike previous generations of robots, they’re not just performing a single preprogrammed task but adapting to dynamic environments, learning new motions, and even following verbal commands.
Much of the current buzz centers on humanoids—robots that resemble people—whose recent exploits include running marathons and performing backflips. General-purpose robots also come in many other forms, however, including those that rely on four legs or wheels for movement (Exhibit 1). But as executives weigh automation road maps and workforce evolution, their focus should not be on whether their robots look human but on whether these robots can flex across tasks in environments designed for humans. This issue is both urgent and intriguing because general-purpose robots, including those in the multipurpose subcategory, may become part of the workplace team: trained to pack, pick, lift, inspect, move, and collaborate with people in real time.2
But how quickly will the robot revolution take place? While embodied AI and robotics are clearly advancing, questions remain about whether they are evolving fast enough, cheaply enough, and reliably enough to deliver real value at scale. The potential is enormous, but progress will depend on several converging technology advances, regulatory factors, and organizational readiness. Rather than betting bullish or bearish, it’s more useful for executives to consider the conditions under which general-purpose robotics might deliver value. This article, which unpacks the promise and pinpoints the pain points, provides a pragmatic lens for senior leaders evaluating when, where, and how general-purpose robotics might matter to their business.
Why the excitement?
The growing momentum about general-purpose robots is real. What’s changed—and why now?
Surge in investment and innovation
The sector has seen an explosion in activity. General-purpose robotics funding grew fivefold from 2022 to 2024, surpassing $1 billion in annual investment, with leading start-ups such as Figure AI, Skild AI, and Agility Robotics raising hundreds of millions of dollars. Patent filings have also surged, with a 40 percent CAGR in volume since 2022.
Governments are taking notice, too. China has designated embodied AI a national priority, anchoring a $138 billion innovation fund.
McKinsey Global Institute’s recent research report, The next big arenas of competition, identifies embodied AI and robotics as one of five emerging frontiers that are shaping future global productivity and digital infrastructure.
AI foundation models as robotics brainpower
Just as large language models unlocked natural conversation for chatbots, vision-language-action (VLA) foundation models enable robots to interpret visual cues, follow spoken instructions, and execute complex sequences. These foundation models support key robotic functions, including perception, reasoning, and decision-making. When paired with multimodal sensors—those that can ingest and act on multiple inputs, such as touch and force—they create systems that can learn by observing humans, without being manually programmed step by step.
Hardware breakthroughs in mobility and dexterity
Robots are now more agile, more stable, and more dexterous than ever. Many models can handle unstructured tasks, such as lifting irregularly shaped items. Simultaneously, improvements in actuators (motors or other devices that convert stored energy into movement) and edge computing have accelerated decision-making and steadily increased energy efficiency.
Compatibility with human-centered spaces
Humanoid robots aren’t just photogenic—they’re practical. They can maneuver in spaces designed for people, including those that involve navigating narrow hallways, turning doorknobs, using tools, and picking items from shelving. While not all use cases require a bipedal form of robot, humanoid robots’ ability to navigate human environments—without needing a workspace redesign—is a real advantage.
Growing focus on safety and collaboration
Cobots—robots designed to safely work alongside people—are getting smarter and safer. Risks have fallen because of improved sensor fusion (for instance, combining vision, sound, and touch), better perception models, and programmable force limits that tell robots when they should exert less pressure than usual. Already, robots such as Agility Robotics’ Digit are operating safely in logistics centers (see sidebar, “A wealth of robotic functions”).
What are the challenges?
Despite recent progress, general-purpose robots are not yet plug and play, and their performance remains suboptimal in many areas. In a recent half-marathon in Beijing, which included 21 humanoid runners, the fastest robot recorded a time of two hours, 40 minutes, compared with just under an hour for the top human.3 Until further technological leaps are made, humanoids will not be knocking people off the podium or rivaling them in many other tasks. Although the current hype and viral robot demos have set expectations very high, the robot ecosystem must overcome substantial hurdles in software, hardware, economics, and operations.
Foundation models still need massive, task-specific data
Foundation models trained on vast internet data sets struggle in real-world physical settings. Directing a robot to manipulate sheet metal or load a dishwasher requires billions of physical interaction examples—from videos, simulation labs, or real life—not just labeled images. The field is racing to build and scale simulation environments to increase available data.
Power and battery limits reduce uptime
Humanoid robots performing dynamic tasks require batteries capable of sustaining high discharge rates. Maintaining such performance over time is challenging and can lead to overheating or reduced battery lifespan. Currently, the best humanoids can typically operate for two to four hours with conventional batteries—less than a full industrial shift. Tasks involving heavy lifting and torque-heavy motion drain batteries faster. What’s more, the recharging infrastructure remains immature. For industrial adoption, robots will need longer duty cycles, faster charging, and better battery-swapping capabilities.
Manipulation remains expensive and slow
The human hand relies on interactions among dozens of muscles, bones, ligaments, joints, tendons, and nerves to function. When working correctly, hands have up to 27 degrees of freedom—the number of ways they can move—involving fingers, thumbs, and wrists.4 While robotic “hands”—or end effectors—are also complex, they cannot fully mimic the precise, delicate action of human hands. For instance, a state-of-the-art humanoid hand from Sanctuary AI has only 21 degrees of freedom, giving it a more constrained range of motion than human hands.5 Until end effectors have better dexterity, speed, and sensitivity, tasks such as tying shoelaces or peeling a banana will remain moonshot-level challenges.
Tactile sensing and compliant actuators—those with a soft or elastic element that can yield under force—have made promising advances, but the devices are not yet industrial grade. In addition, many robots rely on proprietary actuators or tactile sensors that are manufactured only in limited volumes, often in China, posing a supply risk for Western OEMs.
Supply chain issues and integration problems are common
Most general-purpose robots are still assembled from bespoke parts, with no widely accepted standards, making it difficult to scale production. Supply chain bottlenecks, especially for high-precision actuators and sensors, continue to delay production, and integration with factory systems remains expensive and inconsistent.
Shortages of planetary roller screws, which are used in robots because they can handle heavier loads and provide smoother motion, are common, as are delays in torque-sensor shipments. Steadily increasing demand could create even greater constraints.
Furthermore, to make the robots function, developers must integrate hardware and software into a complex technology stack, which is hard to industrialize.
High costs and slow ROI can delay adoption
The development and deployment of humanoid robots is expensive. Manufacturing costs typically range from $30,000 to $150,000 per unit, depending on design complexity and materials.6 Many common and essential components are among the costliest, including planetary roller screws, which can cost between $1,350 and $2,700 each.7 These screws, along with reducers, make up about 33 percent of the typical humanoid bill of materials, according to JPMorgan Chase.8
Overall, general-purpose robots can cost anywhere from about $15,000 to $250,000 per unit, with payback periods often exceeding two years in early pilots.9 Humanoid maintenance costs are also high. Each repair, including shipping and technician labor, may require up to $1,000, and potential downtime could increase the expense.10
Widespread adoption of general-purpose robots will require lower manufacturing and maintenance costs. Returns must also accelerate, especially outside of high-margin sectors.
Organizational and workforce barriers slow progress
Finally, adoption is not just about technology—it’s about people. Companies face workforce resistance, a lack of technical talent for operating and maintaining robots, and no clear playbooks for scaling pilots. Safety, ethics, and regulatory frameworks also remain patchy, and there are numerous concerns. Humanoid robots, for instance, now weigh less than previously, but they are still large enough to hurt someone if they fall or otherwise malfunction. Boston Dynamics’ Atlas, for instance, originally weighed about 330 pounds (150 kilograms) and is now down to about 180 pounds (52 kilograms). Without aligned leadership, clear KPIs, safety guidelines, and accountability, even promising pilots can stall.
What could the market look like?
Despite these challenges, the market potential for general-purpose robotics is massive. Uncertainties persist, however, about timelines, adoption patterns, and technology readiness.
We analyzed general-purpose robotics adoption across sectors and geographies to estimate potential market size under multiple scenarios. Our base case assumes moderate progress in training data, hardware affordability, and integration capabilities, as well as steady cultural and organizational adoption. If this scenario materializes, the market could reach about $370 billion by 2040 (Exhibit 2). Around 50 percent of that value could come from China, with the rest split between Europe and North America. The top use cases could include warehouse logistics, light manufacturing, retail operations, agriculture, and healthcare.
While the base estimates are encouraging, they depend on progress in many areas. The key assumptions—or, to put it differently, what you have to believe—include the following:
- Foundation models continue to improve and are trained on rich physical-interaction data sets.
- Battery technology and power management improve enough to double autonomous uptime.
- Hand-like manipulation gets significantly cheaper and more robust.
- Component supply chains consolidate and standardize.
- Business leaders embrace long-term ROI and prepare their organizations for new workflows and skill sets.
What executives should do now
The general-purpose robotics industry may be at an inflection point. With humanoids, for instance, global manufacturing capacity is expanding rapidly, and many companies have announced ambitious plans for further production increases. This includes Agility Robotics, which intends to scale from 1,200 units of Digit robots in 2025 to 7,500 by 2027 at a new facility in Oregon. Eventually, Agility Robotics hopes the facility will have an annual capacity of 10,000 units. Similarly, Tianlian Robotics in China is building a production facility that aims to have an annual output of 1,000 to 3,000 humanoid robots upon completion.
Although full scaling for humanoids and other general-purpose robotics may take many years, the most forward-looking companies are already preparing for a future in which robotics transforms both the workplace and daily life. Here’s how executives can approach the opportunity:
- Set a long-term automation vision. General-purpose robotics is not the first wave of automation, but it could be the most transformative. Define a long-term vision for where these technologies could play a role in operations—logistics, quality control, inspections, assembly, material handling—and begin identifying pilot-friendly environments. Many tasks are not far from being automated at scale (Exhibit 3).
- Invest in data and data infrastructure. Companies must invest in data for foundation models and data infrastructure to optimize training for general-purpose robots. They must also clearly document what information is used so that researchers can easily replicate their methodology when creating new robot platforms.
- Watch the right indicators. Track progress not just in robot demos but in enabling technologies: battery density, foundation model size and latency, and haptics (the ability to sense touch). Greater regulatory clarity could also be an encouraging sign. These indicators are better barometers of progress than flashy videos that show robots operating in controlled environments.
- Prepare your people. Don’t wait for robots to arrive before upskilling the workforce. Start building the talent and culture needed to work alongside machines, especially in maintenance, operations, and programming. Consider building internal capabilities in simulation, robotics integration, or AI tool development.
- Build partnerships and ecosystems. Stay close to the innovation frontier. This may mean partnering with robotics start-ups, joining standard-setting groups, or investing in modular infrastructure in factories or other workplaces that can easily be shifted to accommodate different robot types.
- Move quickly to create upstream value. Investors and component suppliers, such as those in semiconductors, actuators, power systems, and edge computing, should track where demand concentration is forming across robot platforms. Early partnerships and strategic bets on common standards, open platforms, or shared intellectual property could define long-term winners. As with the early days of smartphones or electric vehicles, supplier alignment often determines scaling success.
Robots could soon do the labor-intensive work of picking and placing items, pointing and directing instruments ...
… and transporting objects, as well as other tasks such as operating hand-held tools and manipulating basic devices.
Robots with the general intelligence, dexterity, and autonomy of a C-3PO or WALL-E aren’t here yet, but the building blocks are emerging fast. The companies that win won’t be those that buy in late but those that prepare early, experiment wisely, and scale responsibly. The age of the electronic coworker is coming—just not overnight.
