Humanoid robots are impressive when they walk, run, or manipulate objects. But beyond these demonstrations, one question remains: where can they provide enough practical value to justify their deployment?

Assisting police officers represents one potential application. During dangerous operations, particularly when violence breaks out on the sidelines of demonstrations, certain tasks expose officers directly to danger. A machine capable of handling some of these responsibilities could become a valuable support tool. This also raises a broader question: what first task should be assigned to a robot to achieve concrete, measurable results?

During the recent demonstrations in France on October 17, these robots could potentially have been useful in assisting police officers and helping protect town halls and businesses.

Assisting Police Officers in Their Duties

The first potential role would be logistical support. Shields, medical supplies, and communication equipment all need to be transported during police operations, sometimes under difficult conditions.

A humanoid robot could deliver this equipment to officers, provided it has sufficient battery life, mobility, and carrying capacity. It could also transport first-aid kits or make repeated trips between a vehicle and a secure operational location.

The benefit would be to reduce certain physical demands and allow police officers to concentrate on their primary responsibilities.

These applications may seem less spectacular than a robot making an arrest. Yet they could provide a relevant starting point for experimentation, with clearly defined tasks and measurable outcomes.

Additional Protection in the Field?

The possibility of a humanoid robot carrying a shield also deserves consideration. It could help temporarily protect an exposed officer or accompany personnel moving through an area where projectiles pose a risk.

The objective would be to reduce human exposure to danger. However, such a function requires far more than simply being able to hold a piece of equipment.

The robot would need to remain stable despite obstacles, coordinate its movements with those of police officers, and avoid dangerous contact. A fall could injure someone or block a passage. In this type of environment, an artificial intelligence error can quickly become a physical safety risk.

This application therefore remains a possibility that requires testing. A demonstration in a controlled environment does not prove that a machine can provide reliable protection on a crowded street.

Security Robotics Already Exists

The concept of sending a machine into a dangerous area has already demonstrated its value in certain specialized operations.

During the Notre-Dame Cathedral fire in Paris in April 2019, a firefighting robot equipped with cameras was deployed inside the nave, where conditions prevented firefighters from operating safely. Its value came from a clearly defined function: operating remotely in an environment too dangerous for humans.

In New York, authorities introduced Digidog in 2023, a quadruped robot remotely operated by trained police officers to assess certain high-risk situations, particularly those involving hazardous substances or barricaded suspects.

These examples demonstrate the value of robotic assistance. However, they do not establish that humanoid robots are ready to operate in close proximity to crowds.

Why Choose a Humanoid Rather Than Another Type of Robot?

Two arms can enable a robot to handle various types of equipment. A human-like body structure could also make it easier to access certain spaces designed for people.

This versatility offers potential advantages, but it also introduces additional complexity. Walking over debris, maintaining balance, and carrying loads must remain possible under real operational conditions.

For certain tasks, a wheeled or tracked platform, a drone, or a quadruped robot may be more suitable than a humanoid. The choice of robot design should depend on the mission, terrain, desired level of autonomy, and safety requirements.

The total cost of ownership will also be a determining factor, including purchase, maintenance, operator training, batteries, and recovery procedures in the event of a breakdown.

Technical and Human Limitations

Crowds represent particularly demanding environments. Unpredictable movements, reduced visibility, and close physical contact make navigation more complicated.

Understanding a situation is even more challenging. A person running may be escaping danger, while a sudden movement could be an attempt at self-protection. Recognizing a movement does not mean understanding the intention behind it.

The collection of images also raises separate concerns. France’s data protection authority, the CNIL, has highlighted the implications of augmented cameras for individual rights and freedoms. Equipping a robot with observation or analytical capabilities therefore requires a specific assessment of how those technologies will be used.

Image protection, remote access, and data flows must also be examined, as with any connected mobile robot exposed to cybersecurity risks.

Finally, a clear distinction must be made between assistance and the use of force. Several manufacturers, including Boston Dynamics, have publicly opposed the weaponization of their general-purpose robots.

Finally, a Practical Application in the Near Future?

The video accompanying this article presents examples of tasks that humanoid robots could perform to support police officers. These scenarios illustrate potential applications; they do not, on their own, constitute operational validation.

At Robot Magazine, we believe humanoid robots could provide genuine value for policing and public order operations, particularly by transporting equipment, conducting observations, and helping protect teams.

In the short term, the most credible opportunities involve limited tasks performed under controlled conditions and human supervision. Close-range protection will require more substantial evidence of reliability.

Could this finally be a useful humanoid robot in the field? The possibility deserves rigorous testing. Success will ultimately be measured by the reliability of the service provided and the actual reduction of risks for both police officers and the public.

FAQ – Humanoid Robots in Policing: How Could They Assist Law Enforcement?

Cet article Humanoid Robots: future teammates to assist police officers in the field? est apparu en premier sur Robot Magazine.

The Robot Magazine 2026 survey explores these questions among business leaders and decision-makers who read the magazine. Its findings point to practical expectations: a clearly defined application, a manageable investment and a measurable financial return.

Robotics demonstrations attract attention. Turning that interest into an investment decision requires answers to more specific questions: which task can be automated, what will deployment cost, and how will the business benefit?

Conducted online from September 13 to 26, 2026, among 300 respondents, the barometer provides insight into how this audience approaches its first robotics project within the French robotics market.

The survey used a self-selected sample of Robot Magazine readers and newsletter subscribers. It reflects the intentions of decision-makers already interested in robotics and is not representative of all French businesses.

Production, handling and logistics account for 59% of first-choice applications

Production and material handling are the preferred first applications for 38% of respondents. Logistics follows at 21%, bringing these two categories to a combined 59%.

Quality inspection accounts for 15% of responses, followed by cleaning at 11%. Other applications represent 9%, while maintenance receives 6%.

These choices suggest that respondents primarily see their first robot as a way to address physical, repetitive operations.

For robot manufacturers and systems integrators, the implication is straightforward: start with an identifiable operational need. Moving goods, feeding a workstation or automating a recurring handling task gives a prospective customer a concrete basis for assessing a solution.

The starting point is the task to be performed. Robot selection, tooling and integration requirements can then be evaluated against that task.

For businesses considering their first robot, the starting
point is not the technology. It is the task that needs to be
automated.

58% would budget less than €50,000

The Robot Magazine 2026 Barometer finds that 58% of respondents would allocate less than €50,000 to their first robot.

More specifically, 31% envisage a budget below €25,000, while 27% would spend between €25,000 and €50,000. At the upper end, 8% anticipate investing more than €100,000.

These figures describe intended budgets. They do not establish whether a complete installation would be achievable within those amounts.

The robot’s purchase price is only one component of the total cost of a robotics deployment. Depending on the application, businesses may also need to account for tooling, safety measures, integration, training, maintenance and operating expenses.

Suppliers can help buyers assess feasibility by clearly explaining what a quotation includes and how the total project cost relates to the proposed application.

Meanwhile, 18% of respondents do not yet know how much to budget. For this group, understanding the economics of a robotics project comes before choosing equipment.

Robot as a Service, or RaaS, may also be worth evaluating where available. Subscription or usage-based arrangements can change the timing of expenditure, although their overall cost and contractual terms still require careful assessment.

48% expect payback within two years

Financial expectations are equally clear: 48% of respondents want to recover their investment in less than 24 months.

This includes 14% expecting payback within one year and 34% targeting between one and two years.

For these decision-makers, a robotics proposal needs to connect expenditure with measurable operational improvements. Relevant measures may include working time saved, increased throughput, reduced rework or fewer physically demanding manual operations.

A credible robotics ROI calculation should make its assumptions explicit. Expected utilisation, production volumes, maintenance costs and deployment time can all influence the outcome.

However, 20% of respondents do not yet know what payback period to target. Documented customer examples and accessible calculation methods could help them assess whether a proposed project meets their needs.

A successful robotics demonstration proves technical
capability. A credible ROI calculation helps justify the
investment.

Price and uncertain returns account for 46% of the main barriers

Asked to identify their principal obstacle, 27% of respondents cite price and 19% uncertainty about return on investment. Together, these economic concerns account for 46% of responses.

Integration complexity follows at 17%, while a lack of internal expertise accounts for 12%.

These findings suggest that selling a first robotics project requires more than demonstrating the machine’s capabilities. Buyers also need to understand how deployment would work, what support would be available and who would be responsible for each stage.

A clearly scoped proposal, including the choice between a traditional industrial robot and a collaborative robot, can make those questions easier to answer. It should explain the application, integration requirements, training needs and criteria used to assess performance.

What the findings mean for a first robotics project

Among the Robot Magazine readers surveyed, expectations converge around three priorities: a specific task, an affordable project and a relatively short payback period.

For robotics suppliers, this creates an opportunity to make proposals easier to evaluate. Clear costs, realistic deployment assumptions and documented operational results can help decision-makers move from curiosity to a structured investment assessment.

For businesses considering automation, the findings also suggest a useful starting point: identify a recurring operational problem, define the result required and assess the complete cost of solving it.

The full Robot Magazine 2026 Barometer presents the four survey questions, detailed response breakdowns, methodology and editorial analysis.

Source: Baromètre Robot Magazine 2026 – Robot-Magazine.fr.

FAQ – Robot Magazine 2026 Barometer: What Do Businesses Expect from Their First Robot?

Cet article Robotics Survey 2026: What Businesses Want ? est apparu en premier sur Robot Magazine.

Billions for Robots, but What Kind of Capital?

Eleven billion dollars. According to Dealroom, startups classified in its “humanoids” category attracted this amount of venture capital between the beginning of 2026 and October 6. That is more than double the $4.8 billion recorded for the whole of 2025. This record concerns a specific segment and does not represent total investment across the global robotics industry.

The rise of Physical AI, artificial intelligence capable of perceiving and acting in the physical world, is bringing together two fields that were long treated separately: AI software and machines. Humanoids are attracting much of the attention, but financing is equally important for industrial robots, logistics, construction and healthcare.

The paradox becomes clear when examining the investors themselves: financing robotics companies does not necessarily mean managing a fund dedicated exclusively to robotics. A generalist fund may participate in a major funding round. A deeptech specialist may understand autonomy without investing solely in physical machines. A corporate venture capital arm may support a startup for both financial and industrial reasons.

This overview therefore distinguishes between robotics-focused investors, specialists with broader investment mandates and corporate venture capital funds. It does not claim to provide an exhaustive list or label investors as “100% robotics” when they do not describe themselves that way.

Independent Venture Capital Funds: Examining Their Specialization

Cybernetix Ventures: A Specialist Open to European Startups

Based in Boston, Cybernetix Ventures, founded by Fady Saad and Mark Martin, focuses its investment thesis on robotics, automation and industrial AI. It invests in North America and Europe, from pre-seed to Series A.

Its specialization also extends to the technologies that enable robotics, including sensors, autonomy and software.

Its portfolio includes Cambrian, which specializes in robotic vision, and Rugged Robotics, which operates in construction robotics. The standard initial investment amount is not specified on the official pages reviewed.

Cybernetix also contributes to the robotics ecosystem through Robotics Invest and Robotics Tech Week, whose Boston edition took place from May 26 to 29, 2026.

For European founders, its sector expertise and transatlantic investment mandate provide two concrete reasons to consider the fund.

Pathbreaker Ventures: Investing Before Takeoff

Based in San Francisco, Pathbreaker invests from company formation, particularly at pre-seed and seed stages.

Its investment thesis combines robotics, AI, computer vision and emerging hardware technologies, with applications in manufacturing and other industries. It is therefore a relevant robotics investor, although its mandate extends beyond the sector.

Its portfolio includes Diligent Robotics and Reliable Robotics, which specializes in aviation autonomy.

Reliable Robotics announced a $160 million funding round in April 2026, with renewed backing from Pathbreaker. This amount represents the total funding round, not Pathbreaker’s individual investment.

The indicative investment range of $250,000 to $2 million should be confirmed directly with the team, as it is not specified on the official pages reviewed.

The Robotics Hub: Investing Close to Research

Based in Pittsburgh, The Robotics Hub operates within the Carnegie Mellon ecosystem.

The university documents its origins as an initiative designed to identify and support emerging advanced robotics companies. This connection does not mean that it is a university fund restricted to Carnegie Mellon graduates.

The team highlights its ability to support projects even before a company has been formally established.

Agility Robotics is a documented example of an investment, particularly during its 2018 Series A round.

Its stated focus includes robotics and AI, although investment ticket sizes and the precise conditions applicable in 2026 remain to be confirmed.

Calibrate Ventures: AI and Automation Serving Industry

Based in Pasadena, California, Calibrate Ventures invests across deeptech, AI, computer vision, autonomy and robotics.

One of its co-founders describes the firm as a seed-stage investor.

Its documented investments include Gray Matter Robotics and Collaborative Robotics, also known as Cobot.

Its emphasis on technologies capable of solving commercial problems makes it particularly relevant to startups that can clearly demonstrate their industrial value.

The standard investment amount is not disclosed in the sources reviewed.

Calibrate illustrates an essential distinction: strong robotics expertise can exist within a portfolio that also includes AI companies developing no physical robots.

Corporate Venture Capital: Funding Driven by Industrial Strategy

MISUMI Ventures: $50 Million for Hardware and Automation

Announced on September 17, 2026, MISUMI Ventures is a $50 million fund backed by the Japanese MISUMI Group, with a presence in the San Francisco Bay Area.

Despite a management structure described as autonomous, it belongs to the corporate venture capital category.

Its initial investments are generally expected to range from $500,000 to $1.5 million, targeting seed and early-stage companies.

Its mandate includes robotics, advanced hardware, automation and industrial AI, but also aerospace, medical technologies and clean energy. It is therefore not exclusively focused on robotics.

MISUMI plans to provide access to its component supply and manufacturing network. For a startup, this support may be as valuable as the capital itself.

However, the announcement of the fund must be distinguished from its investment portfolio. The Japanese press release indicated that the fund would be established in October, with its first investment expected before the end of 2026.

No portfolio investment is included in this overview as of October 12.

Toyota Ventures: Robotics Within a Global Deeptech Investment Strategy

Based in California, with offices in Los Altos and San Francisco, Toyota Ventures invests in early-stage companies across several sectors, including AI, robotics and mobility.

Its portfolio includes Intuition Robotics and SLAMcore.

It would be misleading to automatically apply the $500,000 to $2 million investment range published for certain earlier innovation calls to its 2026 activities.

Those programs demonstrate previous investment practices, not a current general investment policy.

For founders, the appeal lies in combining technological expertise with potential access to an industrial environment.

That access must be discussed in practical terms: an equity investment does not guarantee a purchase order or commercial contract.

ABB Ventures: Financing Connected to Automation Needs

ABB Ventures, part of the Swiss group’s ecosystem, covers several areas of electrification and automation.

ABB also identifies activities under ABB Robotics Ventures.

Documented investments include Scalable Robotics and Automata.

The development of Sevensense, a Swiss startup originating from ETH Zurich, illustrates this strategy.

ABB initially made a minority investment before acquiring the company in 2024.

This is a historical example of industrial integration, not a promise of a similar exit for other portfolio companies.

The sources reviewed do not provide a standard investment ticket or a single investment-stage framework covering all these activities.

Europe: A Shortage of Dedicated Funds, Not a Shortage of Funding

Claiming that no specialized robotics funds exist in Europe would be an overstatement.

However, this overview does not reveal an extensive European network of independent investors dedicated exclusively to robotics.

Without an exhaustive survey, this observation should not be presented as a statistical conclusion.

European robotics startups can secure capital from investors whose mandates extend beyond the industry.

RobCo announced a $100 million funding round in January 2026, co-led by Lightspeed Venture Partners and Lingotto Innovation, with participation from Sequoia Capital, among others.

In the United Kingdom, Humanoid announced a $152 million Series A round in July 2026, led by Prime Movers Lab, with participants including Schaeffler, Bosch and Aglaé Ventures.

These transactions demonstrate that investors do not need to specialize exclusively in robotics to finance substantial funding rounds.

In France, Bpifrance also supports robotics financing and development.

Its Large Venture fund participated in Exotec’s Series D funding round announced in 2022.

The National Industrial Venture Capital Fund (Fonds national de venture industriel) follows a different approach: it helps structure industrial investment funds rather than operating as a direct financing channel dedicated exclusively to robotics startups.

Deeptech investors represent another option.

Supernova Invest is one French example, with historically documented support for Robocath in medical robotics.

This investment should not be presented as a new transaction in 2026.

Why do exclusively robotics-focused funds appear relatively uncommon?

Our interpretation is that robotics combines several risks simultaneously: technical development, manufacturing, customer integration and commercialization.

A medical robotics startup, a logistics robot manufacturer and a robotic manipulation specialist do not operate under the same development cycles or regulatory requirements.

For an investment manager, a broader deeptech or industrial investment strategy makes it possible to diversify these risks.

This explanation is an analysis of the investment model, not a statistically demonstrated causal relationship.

Nevertheless, it helps explain why European robotics capital often operates under broader investment categories.

Advice for Founders: How to Target Robotics Investors

Start With the Fund’s Actual Investment Mandate

Verify the geographical scope, investment stage, types of products financed and the fund’s ability to lead a funding round.

An American robotics portfolio does not prove that the investor accepts European companies.

Examine existing investments as well.

They help determine whether the team understands your market, your product’s constraints and your customers’ purchasing cycles.

Adapt Your Evidence to the Investment Stage

At the pre-seed stage, present a credible team, a clearly defined technical challenge and a structured experimentation program.

When raising capital for deployment, demonstrate results under real operating conditions: uptime, human intervention requirements, installation time, maintenance costs and customer return on investment.

A successful demonstration must have the potential to become a repeatable commercial deployment.

Present the Complete Economics of the Robot

The selling price or subscription fee alone is not enough.

Include manufacturing, installation, after-sales service, working capital requirements and any financing needed for machines retained on the company’s balance sheet.

The investment presentation should make it possible to understand the real cost of a robot and its deployment, and how those economics evolve as the number of deployed machines increases.

Link the Investment Amount to a Measurable Milestone

Explain what the funding round will enable the company to achieve: a validated prototype, a paid pilot, certification, an initial production run or repeatable deployment.

Always distinguish between the total amount being raised and the expected investment from each individual investor.

Alignment between the capital requested and the next development milestone is more useful than financial ambitions disconnected from industrial timelines.

Clarify Corporate Venture Capital Conditions

Discuss customer access, industrial trials and promised resources, but also exclusivity clauses, data rights and intellectual property.

The value of the partnership must remain compatible with relationships involving future customers and investors.

An industrial partnership can accelerate development, provided its commitments are clearly defined.

These recommendations are editorial guidance, not investment criteria uniformly attributed to the funds mentioned.

What This Overview Reveals About European Robotics Maturity

Europe is capable of producing robotics companies that raise substantial amounts of capital.

The funding rounds announced by RobCo and Humanoid demonstrate this capacity.

However, these achievements alone do not establish the existence of a complete venture capital ecosystem dedicated exclusively to robotics.

Maturity must also be measured by the ability to finance intermediate stages: moving from prototype to pilot, and then from pilot to repeatable, profitable deployment.

This is where an investor’s industrial expertise becomes decisive.

For European founders, fundraising strategies should therefore combine transatlantic specialists, local deeptech funds, growth investors and relevant corporate venture capital partners.

The best robotics investor is the one who understands the risks being financed and can support the next stage of development.

An investor’s sector label provides an initial indication; its actual investment practices and portfolio offer more meaningful evidence.

Robotics Venture Capital Funds: Summary Table

Investment amounts are expressed in US dollars and refer to initial investments where a range is available. “Not disclosed” means that no standard investment amount was identified in the sources reviewed.

FAQ – Robotics Venture Capital Funds: Who Is Really Financing Robotics Startups in 2026?

Cet article Robotics Venture Capital Funds: Who Is Really Financing Robots? est apparu en premier sur Robot Magazine.

Beyond performance and price, it must examine data flows, remote access, software updates and maintenance conditions. A connected robot also becomes part of the industrial IT system.

Consider the case of a French company purchasing mobile robots directly from a Chinese manufacturer to equip its factory or warehouse. It compares payload capacity, battery life, navigation accuracy and price. But before connecting these machines to the site network, it must also understand their digital architecture.

Can the robots operate without a connection to the manufacturer’s cloud? Do site maps, images and diagnostic data remain within the company? Who can access the machines remotely, from which countries and with what permissions? How are updates validated?

Without a European integrator, the company must organize these checks itself or entrust them to a specialist. However, the presence of an integrator does not, by itself, guarantee cybersecurity: the scope of its intervention, its expertise and its responsibilities must be clearly defined.

To identify manufacturers, integrators and industry specialists, companies can consult the robotics professional directory. However, the provider’s specific expertise in industrial cybersecurity should still be verified before any intervention.

The Chinese nationality of the manufacturer is not, in itself, a vulnerability. However, a lack of documentation, opaque communications or the inability to revoke remote access are concrete criteria that should be examined before purchase.

For the industrial company, the question is simple: after installation, who controls the robots, their data and their connections?

An Attack Surface That Extends Beyond the Robot

An AMR, or autonomous mobile robot, rarely operates on its own. Depending on its architecture, it communicates with a fleet management server, a supervision interface, a warehouse management system and maintenance services. The perimeter to be assessed therefore also includes administration workstations, network equipment and software interfaces.

NIST points out that OT systems technologies that monitor or control physical processes have specific availability, reliability and safety constraints. A cyber compromise must therefore be assessed in terms of its potential operational consequences.

Consider a hypothetical scenario: a compromised maintenance account provides access to the fleet manager. Depending on the permissions granted, an attacker could view information, modify missions or interrupt operations.

This does not mean that the attacker could automatically disable the robots’ physical safety protections. However, immobilizing a fleet may already be enough to disrupt the supply of a production line.

The assessment should therefore cover data confidentiality, configuration integrity, service availability and the possibility of propagation to other systems.

Understand Which Data Leaves the Site

Depending on the sensors and functions enabled, a robot may generate factory maps, images, mission logs and diagnostic information. Not all of this data necessarily leaves the site. The supplier should specify which information is processed locally and which is transmitted externally.

For each external communication flow, the buyer should know the destination, purpose, transmitted content and activation conditions. The company should also identify the service providers involved, retention periods and the consequences of disabling the service.

Encryption protects data while it is being transmitted. However, it does not, by itself, answer questions about how the data will be used or who may access it once it reaches its destination.

One key document to request is a data-flow matrix showing sources, destinations, ports, protocols and purposes. This documentation should be compared with the communications actually observed during a pilot project, particularly during a software update or a support intervention.

A connection to an external server is not automatically suspicious. However, an unexplained or uncontrollable connection requires further investigation.

Control Remote Maintenance Access

Remote maintenance can reduce diagnostic time. However, it becomes a sensitive issue when access remains permanently open, relies on a shared account or is not visible to the operator.

For a robotics project, Robot Magazine recommends named user accounts, multifactor authentication at the remote maintenance access point and permissions limited to the operations that are actually required.

Sessions should be authorized for a defined period and generate usable logs. The company must be able to identify the person carrying out the intervention, determine which systems are accessible and revoke access when necessary.

A VPN secures a communication channel. It does not guarantee that the user’s permissions are appropriate or that access is sufficiently restricted.

When robots are purchased directly, these arrangements must be negotiated with the manufacturer before deployment. In particular, the agreement should specify who authorizes interventions and who retains the access logs.

Isolate Robots Without Disrupting Their Operation

Placing robots on a dedicated VLAN provides an initial level of network separation. However, communications between this network, the fleet server and the company’s other applications must also be filtered.

The recommended approach is to identify the communications required for normal operation and authorize only those flows. Administration interfaces should be accessible only from controlled workstations or controlled access points.

Network segmentation must be verified in practice: an overly permissive rule between networks can eliminate the apparent benefit of separation.

The pilot project should also test system dependencies. What happens when the cloud becomes unavailable, when the Wi-Fi connection is lost or when the fleet server restarts?

The answers should specify which functions remain available, the state adopted by the robots and the conditions required to resume operations. These behaviors must be compatible with the site’s safety and business continuity requirements.

ROS 2 Does Not Eliminate the Need to Check Configuration

For equipment using ROS 2, the name of the middleware itself is not a guarantee of cybersecurity. Research into SROS2 presents tools and methods for securing ROS 2 communications and graphs while also highlighting the challenges involved in implementation.

The buyer should ask which protections are actually deployed: participant authentication, communication permissions, encryption and certificate management.

The system’s behavior should also be examined when a certificate expires or when a security configuration becomes invalid.

This analysis does not replace the examination of the operating system, web interfaces, APIs and third-party components. Security must be assessed on the configuration that is actually delivered.

Plan for Updates Throughout the Entire Operating Life

The company must understand how the robot verifies the origin and integrity of an update, who is authorized to initiate it and how the installation is logged.

Signing an update package helps protect against unauthorized modification. However, it does not guarantee that the new version is free from vulnerabilities or compatible with every configuration used on the site.

The process should include prior validation, a defined maintenance window and a recovery procedure. Any rollback mechanism must remain controlled to avoid reinstalling a vulnerable version.

The duration of software support deserves particular attention. For a machine expected to operate for ten years, the contract should specify the end date of maintenance, the conditions for correcting vulnerabilities and the solutions offered when support ends.

Vendor dependency should also be anticipated: can the company back up its configurations, retrieve its data and continue certain operations if a remote service disappears?

Distinguish Cybersecurity From Physical Safety

The cybersecurity analysis must examine the interfaces between control, supervision and safety functions. Which configurations can be modified remotely?

Which protections remain in place if communications are lost or compromised?

A penetration test does not replace the validation of robotic safety functions. Conversely, the presence of physical safety devices does not demonstrate that IT access and data are properly controlled.

Tests that may cause movement, shutdown or degradation should be prepared with the manufacturer and carried out in a controlled environment.

The objective is to verify the possible consequences of an incident without putting people or the installation at risk.

The European Regulatory Timeline Must Be Anticipated

The Cyber Resilience Act introduces cybersecurity requirements for products with digital elements that fall within its scope.

Its main obligations will apply from 11 December 2027. The reporting obligations for actively exploited vulnerabilities and serious incidents have applied since 11 September 2026.

For a robotics purchase intended for a French site, the supplier should be asked how it is preparing the product concerned for compliance. A general statement about CRA readiness does not replace technical documentation or an assessment of the actual installation.

Evidence to Request Before Signing

For a direct purchase without a European integrator, the industrial company should obtain at least:

• An architecture diagram and a communication-flow matrix.
• A description of the data collected, its destination and retention period.
• Documented rules for authorizing and revoking remote access.
• Procedures for updates, backups and restoration.
• Precise commitments regarding the duration and conditions of software support.
• A vulnerability-reporting contact and a remediation process.
• Documented tests covering connection loss and service recovery.
• When an audit is provided, a report specifying the versions tested, scope, limitations and verification of corrective actions.

The purchasing decision should be based on these verifiable elements. An attractive price is not enough: the total cost of a robotics project must also include network connectivity, cybersecurity controls, technical support, software updates and maintenance. These costs can significantly reduce the initial financial advantage of a direct purchase.

FAQ – Cybersecurity for Mobile Robots Purchased Directly from a Chinese Manufacturer

Cet article Mobile Robot Cybersecurity: What to check before buying est apparu en premier sur Robot Magazine.

After nearly ten years at Meta, why did you decide to join PRESAGE and return to Paris to build a company?

There are two dimensions to this decision: scientific continuity, first, and a more personal choice, second.

From a scientific perspective, joining PRESAGE is a natural progression from the work I have been doing for several years. Much of my research has focused on how models can perceive the world, build representations of it, understand its dynamics and anticipate how it will evolve. At Meta, I explored these questions in relation to the physical world and the interactions between humans, machines and their environment.

With PRESAGE, I am continuing on that path in a different domain: the digital world. Cloud infrastructure or a data center is, in its own way, an extremely complex environment, with millions of entities and interactions constantly evolving. With Digital World Models, or DWM, our ambition is to enable models to understand this complexity, anticipate its evolution and, above all, simulate different possible futures.

That is precisely what makes the subject so exciting: in the digital world, we can simulate a failure, an attack, an intervention or an extreme scenario, observe the consequences and learn from those simulations without taking the risk of causing them in the real system. There is genuine scientific continuity, applied to a new domain.

On a personal level, this decision also reflects a desire that gradually took shape. I built my career across France, Japan and the United States, in highly international research and technology environments. After nearly ten years at Meta, I wanted to return to entrepreneurship: to build something, assemble a team and pursue a scientific vision over the long term.

Returning to Paris is part of that ambition. The idea is not to come back to build a purely French or European project. It is to demonstrate that we can build a technology and science company from Paris with truly global ambitions.

World Models are also central to current advances in robotics. Do you see connections between the technologies developed by PRESAGE and the future of autonomous robots?

Yes, absolutely. From a scientific perspective, the principles are very similar. A truly autonomous robot must be able to understand the state of its environment, anticipate how it will evolve and, above all, simulate the consequences of different actions before deciding how to intervene. That is precisely one of the challenges World Models address.

At PRESAGE, we apply this approach to the digital world through Digital World Models: understanding complex systems, predicting their evolution and simulating different scenarios. But a data center is not just a software system. It is also a physical environment, with machines, networks, power, cooling and equipment that needs maintenance or repair.

As these infrastructures become more autonomous, those two dimensions will converge. Digital World Models will be able to understand the system as a whole, anticipate incidents and determine which actions to take, while robotics will be able to intervene in the physical world. Ultimately, digital and physical autonomy will become part of the same continuum.

A truly autonomous robot must be able to understand the state of its environment, anticipate how it will evolve and, above all, simulate the consequences of different actions before deciding how to intervene.

AI infrastructure and data centers are becoming increasingly complex. How autonomous could they become with your technologies?

In the short to medium term, the goal is to enable infrastructure to do more than react to events: it should anticipate its own operation. If it can predict changes in its workloads, computing requirements or energy consumption, it can allocate resources more effectively, adjust cooling or intervene before an incident occurs.

At the scale of the infrastructure required for AI, these optimizations could deliver substantial gains in reliability, performance, energy efficiency and environmental impact.

Over the longer term, our vision goes much further. We can imagine data centers operating with minimal human intervention, or even without a permanent human presence on site. Combined with autonomous robots capable of carrying out certain maintenance tasks, they could be deployed in environments that are highly challenging today: deserts, Arctic regions, at sea or perhaps one day in space.

For us, autonomy is therefore more than a way to improve the automation of existing data centers. It could allow us to rethink their architecture, their location and, more broadly, what a data center can be.

What are your ambitions for PRESAGE over the coming years, in France and internationally?

Our ambition has several dimensions.

First, from a scientific and technological perspective, we are currently developing Digital World Models for cloud and AI infrastructure. The next step is to move beyond specialized models and build true foundation models of digital worlds: models that can learn principles general enough to work across very different environments, adapt to new situations and transfer what they have learned from one system to another.

This is a major scientific challenge. Digital infrastructure is constantly evolving and combines hardware, software, networks, energy and, increasingly, AI agents at an enormous scale. Learning the dynamics of such complex systems opens up a field of research that remains largely unexplored.

Then there is our international ambition. We want to build this fundamental technology from France and Europe, but with global reach from the outset. The infrastructures we aim to model are global, so our technologies must be able to operate on the most complex systems, regardless of their architecture or location.

Ultimately, our ambition is clear: to make Digital World Models a fundamental intelligence layer for autonomous computing systems.

FAQ – Tony Tung, PRESAGE and Digital World Models

Cet article Tony Tung: “Digital and physical autonomy will become part of the same continuum” est apparu en premier sur Robot Magazine.

Robotic automation is only as good as the motion behind it – and nowhere is that more visible than in automated storage and retrieval systems, where robots are pushed to cover long distances, at speed, without losing accuracy. As French system integrators and manufacturers scale up robotic cells to handle heavier loads and longer strokes, the actuator choice behind the robot arm is becoming as critical as the robot itself.

“We see this challenge repeatedly across the French market,” said Sébastien Richard, Sales Manager France, at Rollon. “Integrators are being asked to build automated storage and retrieval systems that scale in load and travel distance without sacrificing accuracy, and too often the actuator is treated as an afterthought rather than the component that determines whether the whole cell performs reliably in the longer.”

The challenge: precision at scale

For robotic system integrators, the demands of automated storage and retrieval systems rarely stay modest for long. What starts as a straightforward pick-and-place brief often evolves into something far more demanding: longer travel distances, heavier loads, tighter cycle times, and, critically, no tolerance for drift in accuracy as the system scales.

One recent project, where global linear motion specialist, Rollon, supported Turn-Key Solutions (TKS), a robotic systems integrator, illustrates the point well. The brief called for two robots to move parts vertically up to 30 feet and horizontally over 60 feet, in continuous, high-frequency operation, as part of an automated storage and retrieval system. Loads reached 136 kg, with positional accuracy specified to 1.5 mm – a demand profile that will be familiar to any integrator handling heavy-duty, high-throughput automation.

TKS needed reliable equipment that could endure a heavy-duty schedule, while still holding precise accuracy and speed. The actuator is what determines whether the cell delivers on throughput and reliability and is the deciding factor between a system that runs continuously under heavy load and one that cannot sustain the accuracy the application demands.

The solution: Rollon ROBOT 220

To meet the brief, Rollon’s ROBOT 220 belt-driven actuator was recommended as it is ideally suited to long-stroke, high-dynamic applications. Its load capacity, suitable for the 136 kg pallets involved, combined with proven performance on the vertical lift axis under continuous, high-frequency use, made it the right choice for this automated storage and retrieval system.

Sébastien Richard continued: “Long-stroke, high-load actuation isn’t a catalogue decision – it has to be engineered into the cell from the start, alongside duty cycle, contamination risk and maintenance access. That’s where we work closely with integrators, well before a project reaches the build stage, to make sure the actuator is specified to the application rather than adapted to it afterwards.”

“What stood out in this project wasn’t just the actuator’s technical spec, but its role as an enabling component within a fully integrated robotic cell – one supporting material handling, assembly and machine tending, not just a single isolated task. That’s a pattern Rollon sees echoed across robotics and automation projects in France: as cells are asked to do more – cover longer lines, handle heavier loads, add new stations – the motion system underneath has to be capable of scaling with them, without requiring a redesign every time requirements shift.

“This is particularly relevant for French integrators working across sectors where floor space is at a premium and vertical storage density is increasingly the answer – warehousing and logistics, automotive parts handling, and heavy manufacturing among them. In each case, the actuator isn’t a background component; it’s what determines whether a robotic cell can be trusted to run unattended, at full load, for years rather than months.”

A practical takeaway for integrators

For system integrators navigating similar demands, the message from this project is a practical one: specifying the right actuator requires forward thinking. A well-chosen linear motion system gives a robotic cell room to expand, without the compromises that come from retrofitting an underspecified system later.

As robotic automation continues to mature across French manufacturing and logistics, that distinction is likely to become one of the defining factors in how successfully automated storage and retrieval projects perform over their working life.

Cet article Scaling automated storage and retrieval systems est apparu en premier sur Robot Magazine.

The maintenance technician of 2030 will still need skills in mechanics, electricity, automation and electrical engineering. However, their working environment could be radically different.

When faced with a breakdown, technicians will be able to instantly access an asset’s maintenance history, search for similar incidents, consult manufacturers’ documentation, analyse sensor data and obtain a diagnostic procedure tailored to the situation.

This development builds on a transformation that has been underway for several years with the Industrial Internet of Things, Computerised Maintenance Management Systems (CMMS), predictive maintenance and augmented reality. Robot Magazine previously examined the emergence of the augmented technician and the new skills associated with this profession.

The arrival of generative AI and, more recently, agentic AI is opening a new chapter: the system no longer simply displays information. It can gradually assist with diagnostics and handle some of the tasks required to carry out an intervention.

The augmented technician is becoming an industrial reality

This development addresses a very practical problem: industrial equipment is becoming increasingly sophisticated, while the skills required to maintain it are difficult to find.

In its Smart Manufacturing and Operations 2025 study, Deloitte surveyed 600 executives from large industrial companies headquartered or operating in the United States. Of those surveyed, 35% consider adapting employees to the “Factory of the Future” to be a major concern. In addition, 48% report moderate to significant difficulties in recruiting for production and operations management roles.

The issue is therefore not purely technological. It directly concerns the organisation of work and the transfer of skills.

The principle behind the augmented technician is precisely to use technology to make more knowledge available to professionals at the exact moment they need it.

The technician of 2030 may no longer simply be
the person who repairs the machine. They will become
the person who orchestrates the data, AI and robots around them.

When technical documentation becomes directly actionable

Consider a common situation: a production line comes to a halt.

The technician may have to consult several software applications, search for the manufacturer’s documentation, review previous interventions, analyse machine alarms and ask more experienced colleagues for advice.

A significant share of the intervention time is therefore spent searching for information before the equipment can even be repaired.

New interfaces are changing this process.

The technician can ask the system:

“Has this fault occurred on this machine before?”

“What caused the last three similar shutdowns?”

“Which manufacturer procedure corresponds to this error code?”

“Is this part available in stock?”

Access to technical knowledge therefore becomes much more direct.

In the semiconductor industry, McKinsey estimates that applying large language models to existing capabilities could improve labour productivity by up to 35%, particularly by providing technicians more quickly with the information they need for maintenance operations.

Real-world example: up to 90% less unplanned downtime

Some projects are already producing measurable results.

McKinsey describes the case of a consumer goods company that developed a copilot for production operators.

The system brings together various sources of technical knowledge, including breakdown histories, equipment documentation, failure mode analyses and root cause investigation methods.

When a piece of equipment stops, the operator is guided through a procedure that progressively eliminates the different possible causes.

According to McKinsey, the system reduced certain types of unplanned downtime by up to 90%.

Maintenance labour costs fell by approximately one-third, while technicians gained 40% additional capacity, partly because they were called on less frequently to resolve the simplest incidents.

This result illustrates one of the potential effects of the augmented technician: it is not simply about helping specialists work faster.

Operators themselves can become capable of resolving more first-level incidents, allowing experienced technicians to focus on complex problems.

AI does not eliminate the technician’s expertise:
it allows them to access and apply the knowledge
accumulated across the entire factory more quickly.

AI is addressing the final mile of industrial digitalisation

Factories already use a wide range of digital systems: ERP, MES, CMMS, supervisory control systems, quality management platforms, Industrial IoT, inventory management solutions and document databases.

The problem is that this information often remains scattered across multiple applications.

For technicians, finding a piece of information may require navigating between different screens and databases.

This could be described as the final application mile of industry.

The data already exists. The challenge now is to make it immediately actionable on the factory floor.

An interface capable of querying several systems could allow technicians to express their needs directly, without necessarily knowing which application contains the relevant information.

This simplification could become one of the main benefits offered by the new generation of industrial software.

From an assistant that advises to an agent that acts

A second development is beginning to emerge alongside this transformation: agentic AI.

The distinction is significant.

An assistant primarily provides an answer or recommendation. An agent can be authorised to perform a sequence of operations in order to achieve a defined objective.

After detecting an anomaly, for example, an agent could search the machine’s history, identify similar incidents, retrieve a technical procedure, check whether a part is available, prepare a maintenance work order in the CMMS and then generate the intervention report.

Critical operations would, of course, remain subject to the company’s safety rules and approval procedures.

However, some of the administrative and computer-based tasks currently performed by technicians could gradually be automated.

In its 2026 outlook for the industrial sector, Deloitte also notes that agentic AI, with its ability to reason, plan and autonomously perform certain actions, is expected to become a new component of smart manufacturing.

Productivity gains are already visible in smart factories

This transformation extends far beyond maintenance.

The industrial companies surveyed in Deloitte’s 2025 study reported that, following the implementation of their smart manufacturing initiatives, they achieved an average:

10% to 20% improvement in production; 7% to 20% improvement in workforce productivity; and 10% to 15% increase in released production capacity.

In addition, 46% of respondents rank process automation among their top two investment priorities for the next two years, compared with 37% for physical automation.

These figures reveal an important shift.

Factory transformation is no longer based solely on installing additional robots. It also involves automating the processes surrounding human work.

What if the technician’s next tool were a humanoid robot?

Another stage could gradually emerge: the integration of these digital systems with physical robotics.

Mobile robots already perform certain transport, monitoring and inspection operations in industrial environments.

Humanoid robots aim to go further by operating in environments originally designed for humans.

In the future, some inspection, handling, measurement and intervention tasks in hazardous areas could be entrusted to these machines.

Technicians could then supervise several tools: diagnostic software, sensors, mobile robots and potentially humanoid robots.

The division of roles would become relatively clear: sensors observe, software analyses, robots perform certain physical actions, and technicians supervise and make decisions.

Today, this organisation appears more credible than the vision of a fully autonomous factory from which humans have disappeared.

Will the technician of 2030 become an orchestra conductor?

The central question is therefore probably not whether artificial intelligence will replace the technician.

It is how many machines, robots and systems a technician will be able to supervise in the future with the help of these new tools.

Traditional technical skills will remain essential. However, they will be complemented by the ability to use data, work with diagnostic support systems, verify their recommendations and supervise increasingly autonomous equipment.

For industrial companies, the challenge will therefore involve training just as much as technological investment.

In Deloitte’s study, 92% of the executives surveyed consider smart manufacturing to be one of the main drivers of industrial competitiveness over the next three years. However, the study also shows that human capital and maintenance are among the areas where the gap between current maturity levels and desired maturity levels remains significant.

The factory of 2030 will probably be neither one of humans versus robots nor one of AI versus technicians.

It will be defined by a new distribution of work between humans, software and machines.

Within this organisation, the augmented technician could become less of a simple maintenance operator and more of an orchestra conductor for industrial maintenance.

FAQ – Factory of 2030: How Is AI Transforming the Role of the Technician?

Cet article Factory of 2030: How AI Is Transforming the Role of Maintenance Technicians est apparu en premier sur Robot Magazine.

Humanoids can walk, sort components, manipulate objects and execute increasingly complex industrial tasks. Videos of robots dancing, running or folding laundry continue to attract millions of views.

But industrial buyers are beginning to ask a much less glamorous question: Where is the ROI?

As the sector moves from prototypes to production environments, the benchmark is changing. A successful demonstration is no longer enough. Humanoid manufacturers now need to prove uptime, throughput, safety, integration costs and, ultimately, that somebody is willing to pay for the work their machines perform.

The evidence available in September 2026 suggests that this transition has started. But only a small number of companies can currently point to meaningful commercial deployments.

Agility Robotics: perhaps the clearest commercial case

If the criterion is simply “Is a customer paying to put humanoids to work?”, Agility Robotics is one of the strongest cases today.

Its Digit robot has been operating commercially with GXO Logistics since 2024 under a multi-year Robots-as-a-Service agreement. Digit performs a deliberately unglamorous logistics workflow: moving totes from autonomous robots and placing them onto conveyors.

That is precisely what makes the deployment interesting. It is repetitive work occurring inside an existing warehouse process rather than a carefully staged robotics demonstration.

Agility subsequently reported that Digit had moved more than 100,000 totes at GXO’s Flowery Branch facility.

The company has since expanded its commercial footprint. Toyota Motor Manufacturing Canada signed a commercial agreement in February 2026 following a pilot, while Agility says its robots are operating or being introduced with industrial groups including GXO, Toyota and Schaeffler.

More importantly, Agility disclosed in 2026 that it had secured more than $300 million in multi-year contracted orders for Digit v5.

That does not mean $300 million has already been recognized as revenue. Orders, deployments and recognized revenue are three very different metrics. But it is one of the clearest signs yet that industrial customers are prepared to commit significant budgets to humanoid automation.

In financial disclosures surrounding its planned public listing, Agility presented an illustrative model in which a Digit v5 could cost approximately $200,000, plus around $20,000 for deployment and approximately $36,000 per year for software and maintenance. Using a five-year operating life, that implies a total customer cost around $400,000.

The fundamental industrial question therefore becomes measurable:

Can one humanoid create more than roughly $80,000 per year of economic value after integration, supervision and operational constraints?

If it can, humanoid robotics starts becoming an automation investment rather than a research experiment.

The winner of humanoid robotics may not be
the company with the most impressive robot.
It may be the company that can prove its robot
creates more economic value than it costs to deploy.

Figure: impressive production data, but different economics

Figure has produced perhaps the most important independently confirmed industrial dataset in the sector.

During its Figure 02 deployment at BMW’s Spartanburg plant, the robot worked on an active automotive production line handling sheet metal components. BMW says the deployment accumulated approximately 1,250 operating hours, more than 90,000 components handled, around 1.2 million robot steps and more than 30,000 BMW X3 vehicles supported.

This matters because industrial robotics ultimately lives or dies on repetition. Performing a manipulation once proves capability. Performing it tens of thousands of times begins to demonstrate reliability.

BMW has since moved to Figure 03 and a more complex logistics sequencing application, where robots must select components from containers and organize them for just-in-sequence production.

Figure therefore appears to have crossed an important technical threshold: its humanoids are no longer confined to the laboratory.

But an important distinction remains: operational success is not the same thing as demonstrated profitability.

Figure is privately held and does not disclose detailed humanoid revenue, gross margins, deployment costs or customer economics. We know considerably more about what its robots can accomplish than about how much money those deployments generate. That distinction will become increasingly important.

China is selling humanoids, but who is buying them?

China presents a completely different picture. Chinese manufacturers are rapidly industrializing humanoid production, while competition is pushing hardware prices downward.

UBTECH is particularly interesting because, unlike many private Western humanoid startups, it publishes financial results. For the first half of 2026, UBTECH reported approximately RMB 590 million in revenue from full-size embodied intelligence humanoid robot products and services, with 921 units sold.

That makes UBTECH one of the few companies where significant humanoid-related revenue can actually be observed.

But revenue needs context. A significant portion of Chinese humanoid demand has been driven by government procurement, state-backed training centres and subsidies rather than conventional private-sector automation demand.

The distinction matters. Selling a robot is not necessarily evidence that the robot is economically replacing or augmenting human labour in a factory. A machine purchased by a research laboratory or government-backed robotics training centre generates revenue for the manufacturer, but it does not demonstrate industrial ROI.

China may therefore currently be the world’s fastest-growing humanoid hardware market without yet being the world’s most mature humanoid labour market.

Tesla: enormous ambition, limited external evidence

Tesla remains impossible to ignore. Optimus potentially benefits from something many robotics startups lack: factories, manufacturing expertise, computing infrastructure and an enormous internal environment in which robots can be developed and deployed. In theory, Tesla can become both manufacturer and first customer.

But from an industrial procurement perspective, the evidence remains limited. Tesla has demonstrated increasingly sophisticated Optimus capabilities and has discussed internal factory deployment at scale. Yet detailed, independently verified figures covering uptime, task throughput, cost per operation or external customer ROI remain scarce.

The interesting question is therefore no longer whether Optimus can perform useful tasks. It is:

When will Tesla publish the equivalent of BMW’s 90,000-part Figure deployment or GXO’s 100,000-tote Digit deployment?

Until then, Optimus remains one of the industry’s most important technology programmes, but not yet its clearest commercial benchmark.

Moving 100,000 totes may be less
spectacular than a humanoid dancing.
Economically, it is far more interesting.

Apptronik, Boston Dynamics and the rest of the field

The same distinction applies elsewhere. Apptronik has been working with Mercedes-Benz on Apollo humanoid applications. Boston Dynamics and Hyundai are preparing Atlas for industrial environments. Numerous Chinese manufacturers, including Unitree, AgiBot and Fourier Intelligence, are rapidly improving hardware and reducing costs.

These programmes matter. But pilots, partnerships and purchase announcements should not be confused with scaled commercial operations.

For technical buyers, five metrics are becoming much more useful than demo videos:

• Autonomous operating hours
• Intervention rate
• Task throughput
• Total deployment cost
• Cost per completed operation

Add safety certification, maintenance requirements and integration with factory and warehouse management systems, and the real difficulty of humanoid deployment becomes apparent.

The hidden metric: human intervention

Perhaps the most important number in humanoid robotics is one companies rarely advertise: how often does a human need to help the robot?

A humanoid could theoretically work 20 hours per day. But if remote operators frequently intervene, reset failed tasks or manually control difficult manipulations, the economics change dramatically.

This is why autonomy percentage alone can be misleading. A commercially meaningful humanoid needs not only high autonomy but also a sufficiently low intervention frequency and short recovery time.

A robot completing 99% of actions autonomously sounds impressive. But in a workflow containing thousands of actions per shift, the remaining 1% can still generate substantial human supervision.

The industry’s next transparency battle may therefore concern intervention rates rather than walking speed or dexterity.

Why warehouses and factories are winning first

The earliest viable humanoid applications share several characteristics. They occur in relatively structured environments. The workflows repeat frequently. Objects and infrastructure are reasonably predictable. The economic value of automation can be measured.

Labour shortages, ergonomics and undesirable repetitive work also create an existing business problem. This explains why logistics and automotive manufacturing are emerging ahead of household robotics.

The humanoid’s theoretical advantage is not necessarily that it can outperform a specialized industrial robot. Often it cannot. Its advantage is that it may eventually operate inside infrastructure designed for humans without requiring the facility to be rebuilt around automation.

Doors, shelves, carts, workstations and production lines were designed around human geometry. A sufficiently capable general-purpose robot could exploit that installed infrastructure. That is the economic hypothesis behind the humanoid.

The real competitor isn’t another humanoid

There is another uncomfortable question the industry needs to answer. A humanoid robot is not competing only against Figure, Digit, Optimus or Apollo. It competes against an industrial robot arm, an autonomous mobile robot, a conveyor, machine vision, a purpose-built automation cell and, of course, a human worker.

If a $50,000 automation solution can perform the same workflow more reliably than a $200,000 humanoid, there is little economic reason to use a humanoid.

The humanoid only wins when its generality creates value. That could come from performing several tasks with the same hardware, being redeployed without major infrastructure changes, handling environments designed for humans, or learning new workflows through software rather than mechanical reengineering.

This is why the ultimate commercial metric may not be cost per robot. It could be cost per task across the robot’s lifetime.

From cost per robot to cost per productive hour

Industrial buyers may soon stop asking: How much does the humanoid cost?

The better question is: How much does one productive autonomous hour cost?

Imagine a $200,000 robot operating reliably for five years. If it works only a few hours per day, requires frequent supervision and performs one narrow workflow, the economics may be poor. If the same hardware works across multiple shifts, performs several tasks and learns additional workflows through software updates, its economics change dramatically.

The real value is therefore not simply making robots more capable. It is allowing the same physical asset to become more productive over time through software. That is fundamentally different from traditional fixed automation.

So, who is actually making money?

As of September 2026, the answer requires nuance.

Agility Robotics has one of the strongest cases for genuine paid industrial deployment and has disclosed a substantial contracted order book.

Figure has some of the strongest publicly verified production evidence, particularly through BMW, but its underlying revenue and deployment economics remain private.

UBTECH is generating substantial identifiable humanoid-related revenue, although the Chinese market’s dependence on public procurement and training infrastructure complicates comparisons with Western industrial deployments.

Tesla, Apptronik, Boston Dynamics and many emerging competitors remain strategically important but have yet to publish enough standardized commercial data to make direct ROI comparisons possible.

Nobody has yet demonstrated, publicly and at massive scale, that general-purpose humanoids can economically replace conventional automation or human labour across multiple industries.

But something important has changed. Two years ago, the central question was: Can humanoid robots work? In 2026, there is enough evidence to answer yes, under specific conditions.

The next question is much harder: Can they work cheaply enough, reliably enough and autonomously enough to generate sustainable profit?

That is the race that matters now.

The winner of humanoid robotics may not be the company with the robot that runs fastest, looks most human or produces the most impressive demonstration. It may simply be the company that can prove one equation:

Value of work performed > total cost of deploying the robot.

Everything else is still a demo.

Cet article Humanoid Robots: Enough Demos. Who Is Actually Making Money? est apparu en premier sur Robot Magazine.

Robot Magazine is pleased to welcome a special contribution from Dr. Gleb Tsipursky, a leading behavioral scientist and internationally recognized expert on AI adoption. Named the “Office Whisperer” by The New York Times, he examines a critical issue for the next generation of Physical AI: how organizations can learn from near-misses before autonomous machines scale across factories, laboratories and workplaces.

PHYSICAL AI NEEDS A NEAR-MISS LEDGER BEFORE AUTONOMOUS MACHINES SCALE

Anthropic’s August 27 research preview of the Model Hardware Standard moves AI agents another step out of software and into the physical world. The proposed standard lets agents operate programmable equipment including microscopes, liquid handlers, lasers, and robotic arms, coordinate multiple devices, adjust parameters during a workflow, and in some cases recover from hardware errors without human intervention.

That progress changes the safety problem. When an AI assistant drafts a bad paragraph, the failure is usually visible and reversible. When an agent controls physical equipment, the consequences can include damaged materials, lost production time, spoiled experiments, unsafe motion, or a chain of small deviations that eventually becomes a serious incident.

Traditional safety systems already record accidents, breakdowns, and maintenance events. Physical AI needs one additional layer: a near-miss ledger that records moments when an autonomous system was heading toward a bad outcome but a person, safeguard, or other system caught the problem first.

Near-misses are especially valuable for adaptive systems because they reveal where nominal success hides fragile judgment. If the machine completed the task only because a technician intervened, the final “success” metric can erase the most important evidence about the system.

Consider a robotic arm that reaches toward the wrong tray and stops because an operator notices the path. Production continues, so the event may never enter a failure log. Yet the episode contains information about perception, task context, human vigilance, and whether the same error could recur when nobody is watching closely.

A useful near-miss ledger would capture five things.

First, record the intended action. What was the agent trying to do, and what authority had it been given? A vague entry such as “robot error” tells the next operator almost nothing. The record should identify the task, the decision the system made, and the physical action it attempted.

Second, record the operating context. Physical systems behave differently across lighting, temperature, load, equipment state, workspace layout, and other conditions. Context matters because the same model decision can be harmless in one setting and dangerous in another.

Third, record the intervention. Did a person stop the robot? Did a software limit block the command? Did another sensor contradict the agent’s interpretation? The intervention shows which control actually prevented the bad outcome.

Fourth, record the prevented consequence. This does not require pretending to know exactly what would have happened. A short description such as “possible collision with adjacent fixture” or “wrong reagent likely to enter process” is enough to classify the risk without inflating certainty.

Fifth, record the corrective change. The point is not merely to accumulate warnings. Teams should note whether they changed a threshold, revised an instruction, altered the physical setup, added a sensor check, narrowed the agent’s permissions, or determined that no change was necessary after review.

This is a different mindset from treating human intervention as an embarrassing exception. In early physical-AI deployments, intervention is part of the learning system. A worker who notices that the machine is about to make a mistake is producing operational data that developers cannot get from successful runs alone.

That matters as integration gets easier. Anthropic says its Model Hardware Standard can reduce hardware integration that normally takes weeks or months to hours or minutes. Lower integration costs should expand experimentation. They can also increase the number of settings in which an agent encounters unfamiliar combinations of equipment, workflows, and environmental conditions.

Robot Magazine’s recent overview of Spain’s robotics market points in the same direction. Robotics is spreading across automotive production, food and beverage, logistics, healthcare, and service applications, while cobots and AI integration are broadening the range of tasks machines can perform. As robots move beyond tightly controlled, repetitive cells, the number of meaningful edge cases grows.

The near-miss ledger should therefore become a leading indicator, not an archive nobody reads. Managers can review patterns weekly or monthly: Which tasks generate the most interventions? Which safeguards catch problems? Are near-misses falling after corrective changes? Are operators repeatedly compensating for the same weakness? Do some locations or shifts encounter more edge cases than others?

These questions create a better scaling signal than raw uptime. A system can show excellent uptime while depending heavily on invisible human rescue. If those rescues are not counted, leaders may automate more aggressively precisely because people are quietly preventing failures.

The ledger also improves the relationship between frontline workers and robotics teams. Operators often know where a system is brittle before formal metrics show it. Giving them a simple way to record near-misses turns that tacit knowledge into evidence. It also reduces the incentive to hide interventions in order to make a pilot look cleaner than it really is.

The mechanism should stay lightweight. If documenting a near-miss requires a long investigation, workers will skip it. A short structured entry, ideally generated partly from machine logs and confirmed by the person involved, is enough for most cases. Serious patterns can then trigger deeper review.

Physical AI will not become trustworthy because every possible failure is predicted in advance. Dynamic environments make that impossible. Trust will come from building systems that notice weak signals, preserve evidence from interventions, and improve before those signals become injuries, shutdowns, or costly mistakes.

As agents gain the ability to operate more kinds of hardware, organizations will need to learn faster than the machines spread. A near-miss ledger gives them a practical way to do that.

Cet article Physical AI Needs a Near-Miss Ledger Before Autonomous Machines Scale est apparu en premier sur Robot Magazine.

A study published in 2025 examining 303 accidents involving human–robot interactions notably highlights unexpected robot activations and sensor errors. With the emergence of Physical AI, artificial intelligence errors are gradually moving beyond screens and entering the physical world.

For several years, we have grown accustomed to artificial intelligence making mistakes.

A chatbot invents information. An image generator produces an anomaly. A model misinterprets a document.

In most cases, the error remains digital.

With Physical AI, this situation changes radically

When artificial intelligence controls a robotic arm, a mobile robot or, in the future, a humanoid robot, a misinterpretation of its surroundings can trigger a real physical action.

An error of just a few centimetres can have consequences far more serious than a ChatGPT hallucination.

303 accidents analysed

A study published in 2025 in the scientific journal Safety Science analysed 303 accident reports involving interactions between humans and robots.

The researchers identified seven major categories of incidents.

Two phenomena stand out in particular: unexpected robot activations and sensor-related errors.

This is precisely one of the challenges that Physical AI will need to address.

A robot must perceive its environment before it can act.

Cameras, proximity sensors, vision systems, contextual data and artificial intelligence models become its eyes and, to some extent, its decision-making system.

But what happens when that perception is incorrect?

A wrong decision becomes a physical movement

Consider an extremely simple scenario.

A robot equipped with a camera must identify a crate and move it.

Its vision system misinterprets the scene.

The software nevertheless decides that it has identified the correct object.

In a conventional computer system, this error might produce incorrect data.

In a robot, it can potentially produce an incorrect physical movement.

If an operator is standing in its path, the algorithmic error becomes a safety issue.

With Physical AI, an artificial intelligence
error no longer produces only an incorrect
answer. It can produce a physical
movement.

Accidents are already a reality

Data from the United States provide a tangible measure of this risk.

In its report on workplace accidents in the manufacturing industry, the US Occupational Safety and Health Administration identified 550 incidents involving robots.

In 33% of cases, the victims were treated in emergency departments, while 2% of incidents required hospitalisation.

Reported injuries included cuts, bruises and fractures.

Robot maintenance and cleaning also appear to be particularly sensitive situations.

Of course, not all these accidents can be attributed to artificial intelligence.

That distinction is essential.

However, they demonstrate what can happen when humans and machines capable of powerful movements share the same environment.

Physical AI is now adding a new variable: the machine’s decision-making autonomy.

From software bug to physical risk

This is probably one of the major changes that the robotics industry will need to address in the coming years.

Conventional software follows relatively deterministic logic.

New generations of robots will be increasingly capable of interpreting their surroundings, adapting their actions and making certain decisions in real time.

This capability is precisely what makes them valuable.

But it also creates a new problem.

How can we guarantee the safety of a machine when not all of its decisions can be predicted in advance?

The question becomes even more important with humanoid robots.

Unlike industrial robots, which have traditionally operated behind safety barriers, humanoid robots are specifically designed to work in environments created for humans.

• Factories
• Warehouses
• Hospitals
• Shops
• And potentially, in the future, our homes

Robot errors are becoming a new KPI

Until now, manufacturers have mainly measured availability, productivity, breakdowns and accidents.

Physical AI could introduce a new indicator:
the number of incorrect decisions made by a robot before they result in an incident.

A robot operating for 1,000 hours without an accident may appear extremely reliable.

But if an operator had to intervene 30 times to prevent an incorrect action, the risk assessment becomes very different.

These “near misses” could become an essential metric for assessing the true reliability of autonomous robots.

Accidents tell us what went wrong.
Near misses could show us what may
go wrong tomorrow.

Data remains insufficient to anticipate the risk

The problem is that organisations do not yet have a common language for measuring these situations.

An emergency stop, a trajectory corrected by an operator, an incorrect detection or a collision avoided by only a few centimetres are not always recorded in the same way.

Yet these events could provide the most useful warning signs before an accident occurs.

For example, a robot recording zero accidents over 1,000 operating hours does not present the same level of risk if it required zero, 10 or 30 human interventions to prevent a dangerous action.

The next generation of indicators will therefore probably need to distinguish between:

• Accidents resulting in injury
• Property damage
• Emergency stops
• Perception errors
• Unexpected activations
• Preventive human interventions
• Near misses with no consequences

Without this level of detail, companies risk underestimating the true danger posed by systems that appear to be functioning correctly.

The next battle for Physical AI may be about trust

Robot performance is advancing rapidly.

But if millions of autonomous machines are to work alongside humans every day, their intelligence will need to be more than just efficient.

It will need to be predictable, controllable and capable of failing without putting people at risk.

Artificial intelligence introduced us to a new category of error: hallucinations.

Physical AI now confronts us with a much more tangible question: what happens when a hallucination becomes a physical movement?

Anticipating the risk before the first major accident

The question is no longer whether autonomous robots will make mistakes.

They will.

The real question is whether we will be able to detect those errors early enough to prevent them from becoming accidents.

How should near misses be recorded? What thresholds should trigger a shutdown or a system reassessment? How can an AI model be tested in unpredictable physical environments? Who should be held responsible when an autonomous decision causes harm?

These issues already concern manufacturers, robotics integrators, safety managers, insurers, researchers and regulators.

At Robot Magazine, we want to open this discussion with the organisations and experts developing, deploying and regulating Physical AI.

If you work on autonomous robot safety, near-miss detection, embedded AI system certification or the prevention of human–robot risks, please contact us.

Together, we can begin defining the methods, indicators and best practices needed to anticipate this future challenge before it becomes a crisis.

Cet article Robotics: When an AI Error becomes a Physical Risk est apparu en premier sur Robot Magazine.