Humanoids have begun to step into factories and warehouses in scenarios closer to real-world applications. Earlier humanoid robot deployments were largely confined to demonstrations, but the gap between lab and factory floor is closing. Recent high-profile deals such as the one between Toyota and Agility Robotics reflect the trend. Toyota Motor Manufacturing Canada has signed a Robots-as-a-Service agreement to deploy seven of Agility’s humanoid robots in manufacturing at its Woodstock, Ontario facility. Agility has also expanded its customer base in e-commerce, including a warehouse automation push by Mercado Libre in Latin America.

At BMW Group’s Spartanburg plant in the US, over ten months in 2025, Figure AI’s Figure 02 robot worked alongside human employees on the BMW X3 production line, completing approximately 1,250 operating hours and moving more than 90,000 components. The robot handled precise positioning of sheet metal parts for welding — a task requiring millimeter-level accuracy while being physically demanding for human workers. Within that period, it contributed to the production of 30,000 vehicles.

BMW has just announced that it is bringing robots and physical AI to some of its German plants.

These deployments also reveal how humanoids are being used currently. All involve structured, repetitive tasks in controlled environments. Material handling tasks dominate, including moving bins, positioning components, transporting totes between stations and so on. No manufacturer is deploying humanoids for assembly operations, welding cells, or machine tending — tasks that would require greater dexterity and adaptability.

Humanoid robot costs and ROI

The case for humanoid robots is gradually but increasingly supported by the math. A manufacturing worker in the US costs approximately $160,000 annually when wages, benefits, overhead, and payroll taxes are included. Current humanoid robot prices range from $13,500 for basic models like the Unitree G1 to $250,000 for more capable systems like Agility’s Digit.

At these prices, the payback period for replacing a single-shift worker ranges from two months for the least expensive models to nineteen months for premium units.

Multi-shift operations improve the economics further. A robot operating sixteen hours daily can effectively replace two workers at a combined cost of $320,000 per year. Maintenance, software updates, and integration costs add 50-100 percent to the purchase price, but the basic calculation remains attractive for high-wage environments.

In regions where labor costs are lower, or where existing automation solutions remain cost-effective, the business case weakens. The appeal of humanoid robots derives partly from their ability to work in facilities designed for human workers, avoiding the expense of redesigning production lines. But this advantage matters most in brownfield sites with space constraints or frequent reconfiguration needs.

The bill of materials for a humanoid could drop from the current $40,000 to $50,000 to $10,000 to $20,000 by 2035, the consultancy Bain & Co. projects.

Technical limitations of current humanoid robots

Current humanoids excel at mobility and basic manipulation in mapped environments but struggle with fine motor tasks requiring tactile feedback. Battery life remains limited — most industrial humanoids operate for 4-8 hours before requiring recharge. Reliability under continuous operation is improving but not yet proven over multi-year timeframes.

Safety certification for fenceless operation alongside human workers represents another hurdle. Most current deployments operate in semi-segregated zones or during off-hours when human presence is minimal. Agility Robotics has stated its next-generation Digit will be the first cooperatively safe humanoid capable of working directly alongside people at scale.

BMW notes that while motion sequences trained in the laboratory transferred to production faster than expected, early involvement of production IT infrastructure, occupational safety, process management, and shop floor logistics teams proved essential. Integration with existing manufacturing execution systems and warehouse management platforms requires substantial engineering effort.

Bain and Company’s research indicates that humanoid adoption will occur in three waves: industrial applications in automotive, mining, and construction first; then commercial uses in cleaning, healthcare, and hospitality; finally, consumer applications in domestic settings. The trajectory depends on technology maturation but also on measurable return on investment and building user risk tolerance.

From pilots to scale

The current landscape features many pilot programs, partnerships and testing agreements. It remains to be seen how many of those will transition to commercial deployments of hundreds and then thousands of humanoid robots.

Manufacturing executives require proof of reliability over extended periods — preferably years, not months. They need evidence that humanoids can match the uptime of existing automation solutions, which often exceed 95 percent. Questions about service and support networks, spare parts availability, and long-term vendor viability weigh heavily in capital equipment decisions of this magnitude. For context, a transition from pilot to meaningful scale would likely mean fleets of 50 or more units operating continuously across multiple shifts, a threshold no manufacturer has publicly reached with humanoid robots as of 2025.

The industry also faces a coordination challenge. Humanoid robots, unlike most other manufacturing technologies, lack established supply chains with tiered manufacturers building standardized modules. Each company develops its own actuators, control systems, and AI models. Some suppliers are beginning to address this by developing standardized humanoid components for use across multiple platforms. This vertical integration may accelerate innovation but creates vendor lock-in and complicates maintenance for end users operating multi-vendor fleets.

Humanoid robots are entering commercial use, but years away from scale deployment in most manufacturing environments. Even as they are being used for genuine industrial applications, deployments at scale are some years away.

Widespread adoption will require significant cost reductions, demonstrated multi-year reliability in production settings, expanded dexterity for complex manipulation tasks, and standardized safety certifications for human-robot collaboration. Progress on all fronts is happening, but unevenly.

Subscribe now

Autonomous mobile robots are gaining traction in warehouses and factories worldwide. Their long-term success depends not just on software wizardry but equally on hardware fundamentals.

In a recent conversation with Mobile Robotics Insider, Venkatesh Prasad BS, Head of Mechanical Engineering at Ati Motors, offered some insights into how mechanical design determines whether these machines perform reliably in demanding industrial environments.

His perspective, shaped by nearly three decades in product design across machine tools, forklifts, and mobile robotics, emphasises deliberate engineering over technological flair. Here are five takeaways from his conversation.

1. Design for reliability, not iteration

The philosophy at Ati Motors departs from the Silicon Valley mantra of rapid iteration and learning from failure. Venkatesh argues that in autonomous mobile robots, mistakes cannot be afforded. “It is fine initially, but now we cannot afford to do mistakes,” he notes, describing the shift from a learn-as-you-go approach to a “do it right first time” methodology. This stems from practical economics: fixing errors escalates exponentially as products move from design to prototype, pilot production, and field deployment. A design flaw discovered in a warehouse costs far more to remedy than one caught during engineering reviews.

This approach requires extensive upfront planning. The mechanical engineering team at Ati conducts thorough market studies, customer benchmarking, and competitive analysis before committing resources. They identify technical risks early and address components with long lead times to avoid delays in production readiness. The discipline extends to testing. When choosing between a linear ball bearing and a polymer bearing for suspension mechanisms, engineers build test setups to validate life cycles before finalising designs. Testing runs parallel to design rather than following it, compressing development timelines without sacrificing rigor.

2. Adaptability through materials and structure

Autonomous mobile robots operate in environments that vary considerably — smooth factory floors, rough outdoor concrete, wet surfaces, uneven terrains. Mechanical designs must accommodate these conditions without requiring multiple product variants. At Ati, this translates into selecting drive systems, protective enclosures, and structural materials that withstand diverse stresses.

Prasad describes the use of lightweight yet durable materials such as glass fibre and carbon fibre. These composites reduce the robot’s overall weight, lowering power consumption while maintaining structural strength. They resist corrosion, dampen vibration, and offer an additional benefit: transparency to Wi-Fi signals, which metal enclosures would otherwise block. Material selection extends beyond performance specifications. Load-carrying members undergo heat treatment for strength, and critical components are machined to tight tolerances to ensure accuracy in the tolerance stack—the cumulative effect of part variations on assembly performance.

The outer chassis provides protection against dust, impacts, and water splashes. Coatings prevent rust and wear. Moving parts are covered, and bumpers with sensors prevent collisions. These are not aesthetic choices but functional requirements for robots expected to operate continuously in harsh conditions.

3. Powertrain optimisation and energy efficiency

Transmission efficiency determines how long a robot operates between battery charges. Venkatesh explains that optimising the powertrain involves more than selecting a high-quality gearbox with low backlash. It requires attention to every component connecting the motor to the wheel—shafts, couplings, bearings, and fasteners. Engineers select appropriate grades of steel and aluminium and specify manufacturing processes to minimise friction and energy loss.

Small details matter. Bolts, for instance, must be tightened to specified torques. “You have made a fantastic vehicle, but you don’t tighten the bolts. That’s a bad failure point,” he observes. Battery management strategies also contribute to efficiency. Ati’s robots support both manual and automated battery swapping, depending on customer requirements and available downtime. Opportunistic charging allows robots to recharge during lunch breaks or shift changes, extending operational hours without dedicated charging cycles.

Failsafe mechanisms add redundancy without compromising efficiency. Ati’s Sherpa robots incorporate brakes that engage automatically if power is lost, preventing unintended movement and ensuring safety in warehouses where human workers are present.

4. Modularity and platform reuse

Modular design reduces development time and cost while increasing product versatility. Prasad cites the Sherpa 500 as an example — a modular platform that transforms into different configurations by changing top modules. It can lift payloads, move items on conveyors, or perform turning operations. The Sherpa Mecha, a humanoid robot built on the same chassis, uses interchangeable grippers for tasks ranging from pick-and-place to assembly and testing.

Platform reuse accelerates product launches. When Ati identified demand for a compact robot capable of navigating narrow factory aisles, engineers adapted the Sherpa Pivot — a robot designed for assembly line movement—by adding forks to create a miniature pallet mover. The modification enabled fast deployment in production lines where larger pallet movers could not operate. “This was brought because of the small change that we did to existing vehicle and it went into production like very fast,” he explains.

Modularity also addresses diverse customer requirements. Pallet sizes vary by width, length, and construction (open or closed stringers), and materials range from wood to steel. Fork-to-fork distances can be adjusted for stability across different pallet types. Tilt angles allow loads to rest against the chassis during high-speed transport, improving stability. The challenge lies in designing one product that meets multiple needs without becoming overly complex or sacrificing robustness.

5. Cross-functional integration from the start

Mechanical, electrical, and software teams at Ati collaborate from the outset of product development rather than working sequentially. Venkatesh describes this as concurrent engineering, where system engineers understand customer requirements, design constraints, and manufacturing capabilities simultaneously. If the control team requires high positional accuracy, the mechanical team optimizes steering, gearbox, wheels, and suspension to deliver precision, reducing the burden on software to compensate for mechanical imprecision.

This approach resembles a band fine-tuning instruments together rather than rehearsing separately. Problems are identified and resolved early, before they cascade into costly redesigns. Every engineer understands that delays in new product delivery mean losing first-mover advantage. The team focuses resources on select products to ensure timely completion rather than spreading efforts thinly across multiple projects.

Concurrent engineering shortens development cycles and matures products faster, meeting customer deadlines without compromising quality. Venkatesh emphasizes that all engineers at Ati grasp the value customers derive from their products and design with manufacturability in mind from day one.

Subscribe now

Venkatesh’s journey spans decades across industries, from early days sketching designs on drawing boards to deploying modular robots capable of thriving in India’s most challenging production environments.

At Ati Motors, he leads a team blending creativity with rigorous engineering principles, enabling the creation of robust, adaptable machines like the Sherpa series — robots praised for their resilience, modularity, and efficiency in both indoor and outdoor settings. His approach emphasises meticulous design, firsthand process understanding, and a relentless pursuit of reliability, reflected in over 130 patents to his name.

In this episode, Venkatesh shares insights on achieving “first-time-right” engineering, the value of cross-functional teamwork, and the evolving trends of material science and autonomy shaping the future of robotics. Watch this for a candid discussion on the challenges, breakthroughs, and aspirations that define one of India’s best-known robotics and automation ventures in the backdrop of the global robotics revolution.

Subscribe now

Leading humanoid robotics companies face a fundamental challenge that billions in venture capital may not solve, according to Rodney Brooks, the MIT roboticist who co-founded iRobot. His latest analysis targets the video-based training approaches championed by firms like Figure and Tesla, arguing they overlook a critical component of human dexterity: touch.

Brooks contends that companies pursuing humanoid robots are making a costly miscalculation. Both Figure and Tesla have shifted toward vision-only training methods, recording workers performing tasks with multi-camera rigs and using this footage to teach robots to mimic human movements. Tesla explicitly abandoned motion capture suits and teleoperation in favor of this approach, while Figure promotes learning “directly from everyday human video.”

The core of Brooks’ argument rests on human physiology. He points to research showing human hands contain approximately 17,000 mechanoreceptors and 15 distinct families of touch-sensing neurons. These sensors detect pressure, vibration, texture, and minute force variations that enable dexterous manipulation. Laboratory experiments demonstrate this dependency: when fingertips are numbed, simple tasks like lighting a match become four times slower and significantly more difficult.

Current humanoid training methods capture none of this tactile information. According to Brooks, video-based systems lack force feedback at the wrists, provide limited finger control precision, and operate without any sense of touch. This data gap may prove insurmountable for achieving human-level dexterity.

Brooks challenges the assumption that brute-force learning will succeed where traditional robotics has struggled. He argues that celebrated artificial intelligence breakthroughs in speech recognition, image processing, and large language models were not truly “end-to-end.” Instead, each relied on sophisticated, domain-specific preprocessing that mimicked human sensory systems.

Speech-to-text systems use filtering and frequency analysis originally developed for telephone networks. Image recognition employs convolutional neural networks that replicate visual cortex structures. Language models depend on tokenization and embedding techniques rooted in human linguistic understanding.

The implications extend beyond technical challenges. Brooks suggests the industry may be pursuing an expensive dead end, with “hundreds of millions, or perhaps many billions of dollars” at risk. His timeline estimates place practical humanoid deployment more than a decade away, potentially stranding current investment approaches.

The roboticist’s credibility in this debate is substantial. His previous ventures produced successful commercial robots including Roomba and industrial models deployed in factories worldwide. His criticism carries weight partly because he has navigated both academic research and commercial robot manufacturing.

Brooks predicts that viable humanoid robots will eventually emerge, but with forms bearing little resemblance to current designs or human anatomy. Companies may need to abandon anthropomorphic designs and video-training methods to achieve practical results, suggesting that today’s humanoid robot investments may prove as obsolete as their underlying assumptions.

The stakes involve more than venture capital returns. If Brooks proves correct, the timeline for humanoid robots entering workplaces and homes extends considerably, affecting labor market assumptions and automation investment strategies across industries.

Subscribe now

One of the areas in which advances will have an impact on autonomous mobile robots (AMRs) is event-based sensor technology. These neuromorphic vision systems, also known as dynamic vision sensors (DVS), could significantly change how robots perceive and navigate their environments.

Unlike conventional cameras that capture complete frames at fixed intervals, event-based sensors mimic the human retina by detecting only changes in brightness at the pixel level. This paradigm shift means robots equipped with these sensors generate 10 to 1000 times less data while operating at equivalent frame rates exceeding 10,000 frames per second.

For AMRs operating in dynamic industrial environments, this translates to dramatically reduced computational overhead and faster response times.

Recent research demonstrates that event-based sensors offer microsecond-level temporal resolution with over 120dB dynamic range, enabling robots to function seamlessly across diverse lighting conditions — from pitch darkness to blinding sunlight. This capability is particularly valuable for AMRs that must transition between indoor warehouses and outdoor loading docks without missing a beat.

Traditional AMRs rely heavily on LiDAR, cameras, and ultrasonic sensors for mapping and obstacle detection. While effective, these systems struggle with motion blur during high-speed operations and consume significant power processing redundant visual data. Event-based sensors address these limitations by focusing exclusively on relevant motion and changes in the environment.

Cutting-edge implementations show AMRs equipped with neuromorphic vision achieving 47.3 percent improved detection accuracy in challenging scenarios, such as navigating from dark tunnels into bright outdoor environments. This improvement stems from the sensors' ability to adapt quickly to lighting changes, eliminating the exposure adjustment delays that are a limitation with traditional cameras.

Perhaps most significantly, event-based sensors enable true edge computing for AMRs. As far back as 2017, Intel’s first-generation Loihi research chip enabled processing of 130 million events per second on single CPU cores. Since then, there have been significant advances.

This efficiency allows robots to make real-time decisions without relying on cloud connectivity or power-hungry GPU processing, crucial for battery-powered mobile platforms.

The integration of physics-guided neural networks with event-based perception creates what researchers call ‘neuromorphic navigation stacks’ — systems that combine energy-efficient sensing with intelligent planning algorithms. These advances enable AMRs to operate for extended periods while maintaining high-performance autonomous navigation capabilities.

The event-based sensor market is projected to grow from $500 million in 2025 to $2.5 billion by 2033, we can expect wider adoption across AMR applications. From warehouse automation to outdoor delivery robots, these sensors promise to unlock new levels of autonomy, efficiency, and reliability in mobile robotics, fundamentally changing how robots see and interact with the world around them.

Subscribe now

The robotics community stands at a crossroads in 2025, engaged in what UC Berkeley's Ken Goldberg calls a “paradigm shift” — a fundamental disagreement about whether the future of humanoid robots lies in data-driven artificial intelligence or traditional engineering approaches.

This debate, which has captured the attention of leading researchers and tech executives worldwide, centres on a critical question: Will robots achieve human-level capabilities through massive data collection and training, or do they require the structured foundation of physics-based modelling and engineering principles?

Reality or hype?

Tech luminaries have made extraordinary claims about the imminent arrival of general-purpose humanoid robots. Elon Musk predicts that by 2040, there will be 10 billion humanoid robots — more robots than people on Earth. NVIDIA CEO Jensen Huang has similarly bet big on robot-powered factories, while others suggest we'll see robots performing surgery and replacing factory workers within just a few years.[2][3]

However, robotics experts are pushing back against this timeline. As Goldberg emphasizes, while robots are advancing quickly, “it's not going to happen in the next two years, or five years or even 10 years,” Goldberg tells UC Berkeley News. He is co-founder of Ambi Robotics and Jacobi Robotics and William S. Floyd Distinguished Chair of Engineering at UC Berkeley.

The disconnect between public expectations and scientific reality has created what many researchers view as dangerous “humanoid hype” that could lead to a bubble and subsequent backlash.

100,000-year data gap

At the heart of the debate lies what Goldberg terms the “100,000-year data gap”. Large language models that power AI chatbots were trained on internet-scale text data equivalent to what would take a human approximately 100,000 years to read. This massive dataset enabled rapid advances in language fluency and reasoning capabilities.

Robots face a fundamentally different challenge. Unlike text data readily available on the internet, robots require training data that combines video inputs with precise robot motion commands — data that simply doesn't exist today at the required scale. The few approaches to bridge this gap each face significant limitations:

• YouTube videos can't provide the detailed 3D motion data needed for dexterous manipulation

• Simulation works well for basic locomotion like robot backflips, but fails for complex dexterity tasks like those performed by construction workers or electricians

• Teleoperation, where humans control robots remotely, generates data at an impractically slow rate — every eight hours of work yields only eight hours of data

Two schools

One camp, led by researchers like MIT's Daniela Rus and Russ Tedrake, argues that data-driven approaches are essential for robots to function in unpredictable, human-centred environments. Rus emphasizes that “physics gives us clean models for controlled environments, but the moment we step outside, those assumptions collapse”.

This approach focuses on the following:

• Multimodal datasets capturing human activities from cooking to object manipulation

• Learning from demonstration with robots adapting to variations in real-world scenarios

• Scaling effects where increased data leads to emergent robustness and common-sense behaviours

Tedrake demonstrated this with robots learning to slice apples, where each apple's unique characteristics required adaptive responses that emerged naturally from diverse training data.

The opposing camp, including MIT's Leslie Kaelbling, argues for the continuing importance of mathematical models and first principles. Kaelbling states that “data can show us patterns, but models give us understanding. Without models, we risk systems that work, until they suddenly don't”.

Key arguments include:

• Safety-critical applications demand deeper understanding than trial-and-error learning

• Physics-based models provide essential insights about motion, force, and control that data alone cannot capture

• Structured approaches can guide data collection and interpretation more effectively

Despite the debate, significant investment continues flowing into humanoid robotics. Goldman Sachs estimates the market could reach $38 billion by 2035, while other projections suggest $66 billion by 2032. Chinese companies are leading development, with China projected to control nearly one-third of the global market by 2029. Meanwhile, recent industrial pilots at companies like BMW and Toyota are beginning to generate the real-world data both camps say they need.

However, fundamental challenges remain. The most basic human capabilities — like dexterity in object manipulation — continue to elude robots. As Goldberg notes, “no robot can pick up a wine glass or change a light bulb”. This reflects Moravec's paradox: tasks that seem simple to humans often prove extraordinarily difficult for machines.

Hybrid solutions

Rather than choosing sides, many experts now advocate for hybrid approaches that combine both paradigms. Goldberg suggests using traditional engineering to create robots functional enough for real-world deployment, which then generates the data needed for improvement — creating a “data flywheel” effect.

This approach is already showing success:

• Waymo's self-driving cars collect real-world data daily while operating commercially

• Ambi Robotics improves package-sorting robots through continuous data collection in active warehouses

• Industrial automation continues advancing through combined model-based control and data-driven adaptation

It takes a village

Contrary to popular fears, Goldberg argues that blue-collar trades are relatively safe from robot displacement. Jobs requiring dexterity and adaptability — construction workers, plumbers, electricians — remain far beyond current robot capabilities. Instead, routine cognitive tasks like form-filling and certain aspects of customer service face more immediate automation risk.

The debate over humanoid robots reflects a healthy scientific process where different approaches compete and ultimately combine to drive progress. While the timeline for general-purpose humanoid robots remains longer than tech leaders suggest, the field is advancing through both data-driven learning and engineering innovation.

The key insight from this debate is that neither approach alone is sufficient. The future likely belongs to hybrid systems that leverage the pattern recognition capabilities of large-scale data while respecting the constraints and insights provided by physics-based models.

As the robotics community continues this paradigm shift, managing public expectations while pursuing rigorous research will be crucial to avoiding the boom-and-bust cycles that have historically plagued the field.

The humanoid robot revolution may not arrive in the next few years, but the foundations being laid today — through both camps in this great debate — are building toward a future where robots can safely and effectively operate alongside humans in the complex, unpredictable real world.

The Interview

Daily AMR

Subscribe now

The world of industrial robots is experiencing its biggest transformation in decades, and it's not just about faster machines or stronger arms. The real breakthrough is happening in how robots learn and think, thanks to what are called Large Behavioural Models (LBMs) - a technology that's making robots smarter and more adaptable than ever before.

Traditional industrial robots have always been like very precise, very expensive puppets. Engineers had to write thousands of lines of code to teach them even simple tasks, and any change in the environment often meant starting from scratch. If one wanted a robot to stack boxes differently or handle a new product shape, one would need weeks of reprogramming by specialists.

Large Behavioural Models flip this script entirely. Think of LBMs as the robot equivalent of ChatGPT - but instead of understanding human language, they understand actions and movements. These AI systems learn by watching thousands of hours of demonstrations, then figure out how to apply that knowledge to new situations they've never seen before.

Just this past week, Boston Dynamics and Toyota Research Institute showcased exactly how game-changing this technology can be. Their Atlas humanoid robot, powered by an LBM, performed a complex sequence of packing and sorting tasks that would have stumped previous generations of robots.

What made this demonstration remarkable wasn't just what Atlas could do, but how it adapted when researchers deliberately interfered - moving boxes, closing lids, and creating obstacles mid-task. The robot simply adjusted and continued working, treating its hands and feet as equally capable tools for solving problems.

This represents a fundamental shift from robots that separate walking from manipulation to machines that use their entire body as one integrated system - much more like how humans approach physical tasks.

The numbers tell the story of this transformation. Toyota Research Institute's research shows that LBMs can learn new skills with 80% less training data than traditional methods. Tasks that once took expert programmers weeks to code can now be taught to robots in an afternoon through simple demonstrations. One LBM can master hundreds of different manipulation tasks and then apply that knowledge to completely new challenges.

This isn't just about making robots more capable - it's about making them practical for businesses that couldn't afford custom automation before. Small manufacturers, warehouses with frequently changing products, and facilities dealing with seasonal demand fluctuations can now deploy robots that adapt rather than require expensive reprogramming.

The warehouse industry is likely to be the first major testing ground, where LBM-powered humanoid robots could work alongside human employees during busy periods without requiring massive infrastructure changes. Unlike traditional automation that demands rebuilding entire workflows, these intelligent machines can drop into existing operations and start contributing immediately.

The robotics industry is moving from an era of rigid automation to one of intelligent collaboration, where teaching a robot a new skill becomes as natural as showing a human colleague how to do a job.

The Interview

Daily AMR

Subscribe now

Investment in robotics startups has nearly doubled in four years, jumping from $7 billion in 2020 to more than $12 billion in 2024, according to JPMorgan's latest Applied Technology report. The surge reflects growing demand for AI-powered automation as companies confront persistent labour shortages and seek productivity gains amid economic uncertainty.

The shift toward later-stage funding rounds signals industry maturation. Many of 2024's largest robotics deals targeted companies with proven business models and established customer bases rather than early-stage concepts. Median deal sizes have grown substantially across all funding stages, with late-stage valuations topping $160 million — a sharp increase from previous years.

“This surge likely reflects growing investor interest in robotics, driven by advancements in AI, the expansion of industrial applications and an increased focus by organizations on boosting

productivity through automation,” the authors of the report note, led by Justin Krauss, managing director and head of applied technology at JPMorgan Commercial Banking.

The Bureau of Labour Statistics reports nearly half a million open manufacturing jobs, making labour shortages a primary catalyst for automation adoption.

Federal policy is amplifying private investment. Government agencies allocated $338 billion toward applied technologies in fiscal year 2024, with the Department of Defense continuing to award contracts to robotics startups. The CHIPS and Science Act has reinforced domestic manufacturing capabilities, particularly in semiconductors that power modern robotic systems.

Artificial intelligence is transforming robotics capabilities beyond traditional programmable machines. AI-driven predictive maintenance, intelligent navigation, and human-robot collaboration are reshaping manufacturing floors and supply chains. Physical AI—allowing robots to train in virtual environments before real-world deployment—represents the next frontier in autonomous system development.

Despite the funding surge, challenges remain. Higher interest rates have pressured public market valuations, making initial public offerings less attractive. Instead, 97 percent of exits now occur through strategic acquisitions as established companies race to integrate robotics capabilities. This consolidation trend reflects both market maturation and corporate urgency to automate operations.

The report identifies three robotics categories driving investment growth. Autonomous robotic systems attract the largest share of funding, followed by AI in the physical world applications. Humanoid robotics, while historically receiving minimal investment, captured increased attention in 2024 as several high-profile startups raised substantial capital rounds.

Industrial manufacturers remain the primary adopters, but new sectors are emerging. Construction companies, laboratory operators, and warehouse facilities increasingly deploy robotic solutions. Robotics-as-a-service business models are lowering barriers for small and mid-size enterprises that previously couldn't justify capital investments in automation equipment.

Geopolitical tensions and supply chain disruptions have heightened interest in domestic production capabilities. Robotics enables companies to reshore manufacturing without sacrificing cost efficiency, addressing both economic and national security concerns. Defense applications, particularly autonomous surveillance and drone technologies, represent a growing share of government contracts.

The sustainability angle is gaining importance. Robots improve manufacturing precision, reducing material waste and extending product lifecycles. In renewable energy production — solar panels, electric vehicle batteries, recycling equipment — robotics enables rapid scaling without compromising quality standards.

Market observers note that robotics requires more capital than pure software companies to reach commercial scale. Manufacturing, testing, and regulatory compliance create higher funding requirements. However, successful robotics companies also command premium valuations due to their hardware-software integration and defensible market positions.

Looking ahead, demographic changes in developed economies will likely accelerate robotics adoption. Aging populations in the United States, Japan, China, and Germany are straining labour markets across manufacturing sectors. Robotics offers a path to maintain production capacity and economic growth despite workforce constraints.

As the sector matures from experimental technology to commercial necessity, funding patterns suggest sustained growth ahead.

The Interview

Daily AMR

Previous

Subscribe now

In Part 1, Naveen gave us a quick intro to Ati’s ‘full-stack’ approach to how the company’s flagship family of robots, called Sherpas, were engineered from the beginning. In this episode, he gives us a sense of the capabilities Ati was able to build into the robots because of that approach.

A standout feature is that the Sherpas can operate in harsh conditions, including outdoors, in blazing summers, rain and even snow.

In this conversation, Naveen also talks about advances that are happening in the world of AMRs and how they will influence the evolution of these robots. He believes that, in general, AMRs will be much more ‘intelligent’ in the future. This will allow humans to put AMRs to work in many more environments and in more complex scenarios where AI-led software will act as overarching fleet managers, in both intra-fleet and inter-fleet applications.

Previous post

Subscribe now

Humanoid robots — once the stuff of science fiction — are closer to becoming a practical reality across industries. Recent advances in artificial intelligence, robotics, and sensor technology have propelled bipedal, dexterous robots out of the lab and into real-world pilot programs.

These machines are now being trialled in factories, warehouses, and even hospitals, signalling a new era where robots may soon work side by side with humans in dynamic, unstructured environments.

Several converging trends are driving the rapid progress and deployment of humanoid robots, according to a new report from the consultancy Bain & Company, titled Humanoid robots at work: what executives need to know.

Human-level Mobility and Dexterity: Today’s humanoid robots can walk, jump, and navigate complex terrain. Enhanced by AI, their fine motor skills now approach human precision, enabling them to manipulate objects and perform intricate tasks once thought impossible for machines.

Simplified training and management: Thanks to natural language AI, robots can be trained using spoken instructions rather than specialized coding. Their human-like form allows them to learn directly in the environments where they’ll operate, reducing the need for costly retrofits or custom infrastructure.

Cost parity with human labor: Between 2022 and 2024, the unit cost of humanoid robots dropped by at least 40 percent. Some models, such as Unitree’s $16,000 robot, now match or undercut the annual cost of a minimum-wage worker in the US, making them economically viable for a wide range of employers.

General purpose intelligence: Advances in generative AI and multimodal data analysis are enabling robots to respond autonomously to unpredictable changes in their surroundings, expanding their usefulness far beyond repetitive automation.

Major manufacturers and logistics companies are leading the charge in deploying humanoid robots. Automotive giants like BMW, Mercedes-Benz, and Toyota are piloting robots on assembly lines, where they handle monotonous and physically demanding tasks.

In warehouses, robots such as Agility Robotics’ Digit are being tested by Amazon and GXO Logistics to sort packages and transport goods, demonstrating their ability to operate in crowded, ever-changing environments.

NEURA Robotics has just struck a partnership with Hyundai Samho to develop quadruped and humanoid robots for use in shipbuilding.

These pilots are not just about novelty — early results show significant productivity gains. For instance, BMW’s deployment of the Figure 02 humanoid robot at its Spartanburg plant reportedly delivered a 400 percent efficiency boost in targeted tasks. As labor shortages intensify and the cost of human labor rises, more companies are expected to follow suit.

Hospitals and elder care facilities are also experimenting with humanoid robots. In China, a large-scale program is deploying 200 humanoid robots in nursing homes, where they assist with daily chores, monitor residents’ health, and provide social interaction. In US hospitals, robots like Moxi are already delivering medications and lab samples, reducing staff burnout and freeing up nurses for direct patient care.

While robots are not yet performing direct clinical tasks, their ability to handle logistical and support roles is proving invaluable. Early deployments have saved thousands of staff hours and improved operational efficiency, offering a glimpse of how robots could help address chronic labor shortages in healthcare.

Market forecasts anticipate steady growth in humanoid robot deployment through 2030, with a rapid uptick expected thereafter. Industry projections suggest the global market could reach $38 billion to over $200 billion by 2035. Funding for humanoid robotics start-ups has soared, rising from $308 million in 2020 to $1.1 billion in 2024, as both tech giants and specialized firms race to commercialize the technology.

For business leaders, the message is clear: while it may be too soon for mass adoption, the technology is maturing fast. Companies with physical operations should begin experimenting with prototypes and strategizing for a future where humanoid robots are a competitive necessity, according to Bain’s report. For a closer look at real manufacturing deployments already underway and the economics behind them, see our analysis of the current state of humanoid robots in factories.

Understanding the pace of technological change, regulatory developments, and workforce impacts will be critical for those aiming to stay ahead.

Humanoid robots are not poised to replace entire workforces overnight. But their growing dexterity, intelligence, and cost-effectiveness mean they will soon take on a meaningful share of physical jobs — reshaping industries, addressing labor shortages, and redefining what it means to work in the 21st century.

The interview

AMR Daily

Subscribe now

Our guest in this episode is Dr. Naveen Arulselvan, CTO at Ati Motors, a pioneer in self-contained robots with natural navigation built into them. This is Part 1 of our conversation with Arulselvan, in which he starts with an interesting piece of trivia about the location of Ati’s facility in Bengaluru, India’s unofficial tech capital.

He goes on to talk about their early engineering decision to stick with natural navigation and how that, as part of their ‘full-stack’ approach to building their Sherpa family of robots — Ati’s current flagship products — now allows them go in and out of their customers’ factories and warehouses very quickly, while deploying multiple robots.

Those decisions early on meant a lot of hard work for years, but it’s all paying off now. Ati is also now owns the mechanical design the hardware as well as the software stack of the Sherpas, making them amenable to rapid customisation based on different customer site scenarios.

Subscribe now

In 2025, Google DeepMind unveiled Gemini Robotics, a suite of AI models designed to bring artificial intelligence out of the digital realm and into the physical world. Building on the Gemini 2.0 platform, Gemini Robotics equips robots with the ability to see, understand, and interact with their environments — enabling them to perform tasks with a level of dexterity and adaptability previously out of reach for machines.

At the heart of Gemini Robotics is a “vision-language-action” (VLA) system. Unlike earlier models that focused on one skill at a time, Gemini Robotics can process and respond to text, images, audio, and video, while also reasoning about physical spaces and taking action. This means a robot can understand a spoken request, visually identify objects, and manipulate them with precision — folding origami, packing lunch boxes, or even playing games like Tic-Tac-Toe.

The models excel at generalization, allowing robots to handle unfamiliar objects and situations without retraining, a leap from traditional robotics that required repetitive, task-specific programming.

Gemini Robotics comes in two main variants: the flagship Gemini Robotics model and Gemini Robotics-ER. The former integrates advanced multimodal reasoning with physical action, while the latter is optimized for “embodied reasoning,” helping robots make sense of their surroundings, predict object trajectories, and generate code to execute complex tasks.

A new “On-Device” version allows these capabilities to run directly on robotic hardware, ensuring fast, reliable performance even without cloud connectivity.

DeepMind’s approach is holistic, emphasizing safety and responsible deployment. The team has introduced the ASIMOV dataset to benchmark safety in real-world robotic actions and is working with partners like Apptronik, Agile Robots, and Boston Dynamics to integrate Gemini Robotics into diverse platforms, from bi-arm research robots to humanoid machines.

The implications are profound: Gemini Robotics sets a new bar for dexterity, generality, and interactivity in robotics. By enabling robots to follow natural language instructions and adapt to dynamic environments, Google DeepMind is laying the groundwork for AI-powered machines that could one day assist in homes, factories, and beyond—turning science fiction’s vision of helpful, adaptable robots into a tangible reality.

The Interview

Daily AMR

Subscribe now

Conversations with founders, investors and other industry experts from the world of AMRs

In this recent interview, Ati Motors Founder and CEO Saurabh Chandra sat down at the company’s dev shop with Harichandan Arakali, editor and host at India Tech Report.

Chandra talks about how the company’s Sherpa family of autonomous mobile robots has been built ground up to work in harsh environments, including outdoors. He also explains the company’s full-stack approach to innovation, which has allowed Ati to build everything the Sherpas need into the robots themselves, eliminating the need for crutches like remote human operators, ‘happy paths’ and other props.

Going forward, Chandra believes that industrial environments will see an increasing number of AMRs, including humanoid versions.

Daily AMR

News from the world of autonomous mobile robots that caught our attention.

Subscribe now