On this page
The humanoid network operator
The previous chapter mapped the emerging structure of humanoid robotics: an hourglass architecture with diverse applications at the top, diverse hardware at...
The previous chapter mapped the emerging structure of humanoid robotics: an hourglass architecture with diverse applications at the top, diverse hardware at the bottom, and a standardized core in the middle. It asked who would own, deploy, and maintain the millions of robots entering the world. The answer may come from an unexpected direction.
It is 3 AM at a Marriott in downtown Chicago. A humanoid robot glides through the lobby, adjusting thermostats, restocking towels, and checking that fire exits are clear. When it encounters a spill near the elevators, it cleans it without waiting for a supervisor's approval. The robot does not belong to Marriott. It is leased from a company called RoboFleet, which manages 4,000 humanoid robots across 200 hotels in North America. RoboFleet handles maintenance, software updates, and operator training. When a robot malfunctions, a replacement arrives within hours. Marriott pays a monthly fee per robot and never worries about repairs, obsolescence, or resale value.
This is one possible future: a world where humanoid robots are not sold like appliances but managed like fleets. A world where specialized operators, not manufacturers, own and deploy the machines that will reshape work.
The Starting Point: 2025#
The humanoid robot market is taking shape faster than many predicted. In October 2025, 1X Technologies opened pre-orders for NEO Gamma, what they called "the world's first consumer-ready humanoid robot." The pricing revealed the industry's emerging business models: $20,000 for outright purchase or $499 per month on a subscription with a 6-month minimum commitment. A $200 refundable deposit holds your place in line, with U.S. deliveries beginning in 2026.
But the consumer story quickly evolved. In December 2025, 1X announced a strategic partnership with EQT, a large Swedish multi-asset investor whose venture fund had backed 1X. The deal involves shipping up to 10,000 NEO robots between 2026 and 2030 to EQT's more than 300 portfolio companies, with concentration on manufacturing, warehousing, logistics, and other industrial use cases. A robot marketed for home use was pivoting to factories before a single consumer unit had shipped.
Agility Robotics offers similar flexibility. Their Digit robot is available through Robot-as-a-Service contracts that bundle hardware, software, accessories, and support into a single subscription, or through conventional purchase with ongoing SaaS (Software as a Service) for updates and improvements. Digit is already operating at Amazon, GXO Logistics, and Spanx facilities, handling tote movement and material transport. The GXO deployment, announced in mid-2024, marked the industry's first formal commercial deployment of humanoids generating real revenue. In June 2025, Agility closed a $400 million Series C at approximately $2.1 billion valuation, with its RoboFab facility in Salem, Oregon scaling toward 10,000 units per year.
In China, small businesses are buying Unitree G1 robots and renting them out for $1,400 to $20,000 per day, mostly for exhibitions and corporate events. Tesla plans to sell Optimus robots outright, with Elon Musk targeting prices under $20,000 once production hits scale.
These models work at small scale. But as production ramps from thousands of units to millions, a new question emerges: who will own, operate, and maintain these fleets when they move beyond factories into hospitals, hotels, warehouses, and homes? Will manufacturers handle everything themselves, or will a new industry arise to manage the robots?
The Precedent: Mobile Networks and Robotaxis#
There is a pattern in technology. When products require ongoing service, complex logistics, and regional expertise, a separation often occurs. Manufacturers build the hardware. Operators manage deployment.
Consider mobile telecommunications. Verizon and AT&T do not manufacture smartphones. Apple and Samsung do. But telecom operators own the infrastructure, manage customer relationships, and handle billing, repairs, and network optimization. This division allows Apple to focus on product design while Verizon focuses on coverage and service.
A closer parallel is taking shape in autonomous vehicles. Tesla manufactures robotaxis, but the company may not operate every fleet. Specialized robotaxi network operators could own thousands of vehicles, manage charging infrastructure, coordinate maintenance, and optimize dispatch algorithms. Tesla builds the cars and provides software. Operators handle the messy details of keeping vehicles on the road. This separation is not yet settled, but the logic is clear: building a car is different from running a fleet.
Humanoid robots may follow the same path.
Why Humanoid Network Operators Might Emerge#
The case for humanoid network operators rests on several forces.
First, manufacturers excel at different things than operators. Tesla, Figure, and Boston Dynamics are engineering companies. Their strength is in hardware design, AI development, and manufacturing at scale. Running a service business requires different expertise: contract negotiations, regional maintenance networks, workforce training, and customer support. A company optimizing for production efficiency may not want the burden of managing 10,000 service contracts across 50 cities.
Second, customers may prefer dealing with operators rather than manufacturers. A hospital purchasing 20 humanoid robots does not want to manage software updates, troubleshoot hardware failures, or negotiate with multiple vendors. An operator can bundle robots, maintenance, insurance, and training into a single contract, simplifying procurement. For industries like hospitality or retail, where margins are thin and technical expertise is limited, the operator model reduces risk.
Third, economies of scale favor specialization. An operator managing 50,000 robots can build centralized repair facilities, negotiate volume discounts on parts, and employ specialized technicians. A single hotel owning five robots cannot. The operator spreads fixed costs across a large customer base, lowering the effective cost per robot. This is the same logic that made equipment leasing a multi-billion-dollar industry for construction machinery and medical devices.
Fourth, local operators may understand regional needs better than global manufacturers. A robotics operator in Japan, where regulations around elder care are strict and cultural expectations for robot behavior differ from the West, can customize deployments in ways a Silicon Valley manufacturer cannot. Operators become intermediaries, translating customer needs into technical requirements and feeding insights back to manufacturers.
The Teleoperation Imperative#
But there is a fifth reason, more immediate and more practical, that could make humanoid network operators essential: teleoperation.
The promise of humanoid robots is full autonomy. A robot that learns a task once and executes it flawlessly forever, requiring no human supervision. This vision remains distant. Current humanoid robots struggle with unexpected situations. A robot trained to fold laundry may freeze when it encounters a shirt with an unusual button. A robot stocking shelves may fail when a box is dented or a label is torn. Edge cases are everywhere, and training AI to handle every variation takes years of data collection and iteration.
This means that for the next decade, possibly longer, humanoid robots will require teleoperation. When a robot encounters a situation it cannot handle, a human operator takes control remotely, guiding the robot through the task. The robot learns from this intervention, gradually reducing the need for human help. But in the early years of deployment, teleoperation will not be occasional. It will be constant.
The 1X NEO launch in October 2025 validated this prediction in striking detail. For complex tasks NEO does not know, owners can schedule a "1X Expert" to remotely guide the robot, helping it learn while getting the job done. The robot's "Emotive Ear Rings" change color to indicate when a teleoperator is active. Owners control when sessions occur, can designate no-go zones in their homes, and can blur faces for privacy. 1X cannot take control without the owner's explicit approval.
This is not a workaround or a temporary measure. It is the core of the business model. The robot arrives with basic autonomy and grows in capability through teleoperated training. Every intervention teaches the AI something new, and those learnings flow back to improve all NEO robots. Teleoperation is not a bug. It is the training mechanism.
Now consider the logistics at scale. A manufacturer in Shenzhen builds 100,000 humanoid robots and sells them to customers in Europe, Africa, and South America. When those robots encounter problems, who operates them remotely? If the manufacturer tries to handle teleoperation from China, several issues arise.
Time zones create the first problem. A robot working a night shift in a German factory needs support at 2 AM local time, which is 9 AM in Beijing. But a robot in a Kenyan hospital needs help at 3 PM local time, which is 10 PM in Beijing. A robot in a Brazilian warehouse operates on yet another schedule. Managing teleoperation across global time zones requires either massive staffing or robots that sit idle waiting for operators to become available.
Latency creates the second problem. Teleoperation requires low-latency communication. An operator in Beijing controlling a robot in Lagos faces delays measured in hundreds of milliseconds, enough to make precise manipulation tasks frustrating or impossible. A surgeon operating a robotic system demands near-instantaneous response. A factory line cannot tolerate delays when a robot needs real-time guidance to avoid a collision. When Serve Robotics acquired Voysys in September 2025 for $5.75 million, they gained technology achieving glass-to-glass latency as low as 50 milliseconds. That acquisition signals how critical low-latency teleoperation has become.
Language and local knowledge create the third problem. A robot operating in a French hospital may need to read handwritten labels in French, navigate cultural norms around patient interaction, or understand regional medical terminology. An operator in China may lack the context to make good decisions. A robot in a South African mine may encounter safety protocols specific to that region. Remote operators without local expertise will make mistakes, and those mistakes will be costly.
The solution is decentralization. Instead of manufacturers handling teleoperation from a single headquarters, regional operators take responsibility. An operator in Berlin manages robots across Germany and neighboring countries, staffing teleoperators who speak the language, understand local regulations, and work in time zones aligned with the robots. An operator in Nairobi handles robots across East Africa. An operator in São Paulo covers Brazil. Each operator employs local teams, reducing latency, improving response times, and ensuring cultural fluency.
This model transforms teleoperation from a bottleneck into a distributed service. Manufacturers focus on building better robots and improving autonomy. Operators focus on keeping robots functional in the real world, managing the messy transition from supervised machines to fully autonomous ones. As robots become more capable, the need for teleoperation declines, but operators can shift their workforce toward maintenance, training, and deployment rather than real-time control.
Teleoperation, then, is not just a temporary necessity. It is a forcing function that makes the operator model practical. A centralized manufacturer trying to teleoperate millions of robots across every continent faces an impossible task. A decentralized network of regional operators makes it manageable.
The Infrastructure Is Already Emerging#
The building blocks for humanoid network operators are appearing across the industry.
Boston Dynamics launched Orbit, fleet management software that provides centralized dashboards aggregating data from all sites, giving operators a unified view of robot activity, site performance, and fleet health. In May 2025, Orbit 5.0 added AI-powered visual inspections using vision-language models that automatically detect anomalies like debris, spills, or corrosion without manual image review. The software can respond to queries for yes/no answers, numeric readings, or descriptive text, flagging safety issues like missing fire extinguishers or monitoring equipment wear. Orbit includes Site View, a visual history of facilities powered by 360-degree images captured by Spot robots. The company plans to integrate Atlas and Stretch into the same platform, creating fleet management infrastructure that works across multiple sites and robot types.
Bain & Company's 2025 Technology Report identified an emerging category: "humanoid robot integrators" who combine robots, AI, fleet management, and workflow redesign into turnkey offerings. These integrators develop industry-specific playbooks, service contracts, change management approaches, and safety certification pathways. They establish operating ecosystems covering spares, services, battery swap infrastructure, and ongoing maintenance. This is precisely the operator model, described by management consultants as the next business opportunity.
Agility Robotics built Arc, a cloud-based automation platform that gives customers complete control of robots and equipment by deploying and integrating automated workflows into logistics and manufacturing operations. Arc provides operational visibility into critical key performance indicators (KPIs) like uptime, throughput, and robot status, with industry-standard APIs for integration with existing warehouse management systems. Arc is not just robot software. It is the management layer that operators need.
The teleoperation industry is consolidating around the same logic. The market for teleoperated robotic systems reached approximately $890 million in 2025 and is projected to exceed $4 billion by 2032. DriveU.auto in Israel, Ottopia, Designated Driver, and others are building platforms that can work across vehicle types and use cases. Shadow Robot sells dexterous hand teleoperation systems for nuclear, pharmaceutical, and research sectors. These companies are creating the operational infrastructure that humanoid network operators will need.
The Counter-Arguments#
But there are strong reasons why humanoid network operators may never materialize, or may remain niche players.
The most obvious is vertical integration. Tesla has shown little interest in separating vehicle manufacturing from service operations. The company operates its own Supercharger network, handles repairs through Tesla-owned service centers, and maintains tight control over software updates. Elon Musk has stated that Optimus could generate over $10 trillion in long-term revenue, a figure that assumes Tesla captures not just hardware sales but ongoing service fees. Why would Tesla hand that revenue to a third party?
Hyundai offers an even more compelling example. In April 2025, Hyundai Motor Group announced it would purchase "tens of thousands" of robots from its subsidiary Boston Dynamics in the coming years, becoming the company's biggest customer. Atlas will be deployed at Hyundai's Georgia Metaplant starting in late 2025, with plans to expand across the automaker's global manufacturing facilities. Hyundai's $21 billion U.S. investment, including $6 billion for innovation and strategic partnerships, signals that the automaker sees robotics not as something to outsource but as a core capability to own. At CES 2026, Boston Dynamics will bring Atlas out of the lab and onto the stage for its first public debut, marking what Hyundai calls "a tangible step toward the commercialization of AI Robotics." When you own the robot manufacturer, the operator model becomes redundant.
Apple offers another model. The company does not manufacture iPhones but controls every aspect of the user experience, from retail stores to software ecosystems. Applying this to robotics, manufacturers might lease robots directly to customers, maintaining control over data, software, and the customer relationship. In this future, operators exist only at the margins, handling specialized deployments that manufacturers find unprofitable.
Control raises another concern. A humanoid robot is not a smartphone. It operates in physical space, potentially in homes and around vulnerable people. Manufacturers may be unwilling to cede control over how their robots are used, fearing liability, brand damage, or misuse. If a robot injures someone, who is responsible? The manufacturer, the operator, or the customer? These legal uncertainties may push manufacturers toward direct relationships, where liability is clearer.
Finally, the infrastructure argument is weaker for humanoids than for telecom. Mobile networks require cell towers, spectrum licenses, and regulatory approval. Humanoid robots need charging stations, which are trivial to install, and maintenance facilities, which any competent repair shop can provide. The barriers to entry for operating humanoid fleets are lower than for telecom, which means manufacturers may find it easier to handle operations in-house.
Even the teleoperation argument has limits. Manufacturers could establish regional teleoperation centers themselves, hiring local teams without ceding control to third-party operators. Tesla could open a teleoperation facility in Berlin and another in Nairobi, maintaining vertical integration while solving the time zone and latency problems. 1X is doing exactly this with its Expert network, keeping teleoperation in-house rather than outsourcing to operators. The question is whether manufacturers want to manage that complexity at global scale or prefer to focus on hardware and software, leaving operations to specialists.
Two Futures#
The first future is decentralized. Manufacturers sell or lease robots directly. Customers handle operations themselves, perhaps with third-party maintenance contracts. Teleoperation, when needed, is managed by manufacturers through regional centers they control. Competition remains high, innovation moves quickly, and no single entity controls the infrastructure. This is the smartphone future: diverse manufacturers, direct customer relationships, minimal intermediaries.
The second future is consolidated. A handful of large operators emerge, purchasing robots by the tens of thousands and managing deployments across industries. These operators become essential infrastructure, much like cloud providers in the tech industry. They negotiate aggressively with manufacturers, shape industry standards, and accumulate vast datasets on robot performance. They employ thousands of teleoperators, making them indispensable during the transition to full autonomy. Customers gain simplicity but lose control. This is the telecom future: oligopoly, stability, and concentrated power.
Which future arrives depends on choices made in the next five years. If manufacturers prioritize control and vertical integration, operators will remain marginal. If manufacturers focus on hardware and delegate services, operators will fill the gap. If regulations mandate local operation of humanoid fleets, operators with regional expertise will thrive. If insurance companies refuse to cover customer-owned robots without professional management, operators will become indispensable. And if teleoperation remains critical longer than manufacturers expect, the need for regional operators will become unavoidable.
The Concentration of Power#
But if humanoid network operators do emerge, they will not be neutral. They will be gatekeepers.
An operator managing 100,000 robots controls labor on an extraordinary scale. That operator decides which tasks robots perform, which customers gain access, and which manufacturers succeed or fail. If an operator chooses to source exclusively from one manufacturer, it can make or break a robotics company overnight. Conversely, a dominant manufacturer could dictate terms to operators, ensuring that only its robots are deployed widely.
This concentration raises uncomfortable questions. If three or four operators dominate the humanoid market, as three or four telecom companies dominate mobile service in most countries, what prevents monopolistic behavior? Could operators raise prices arbitrarily once customers become dependent? Could they sell data collected by robots, tracking not just workplace efficiency but private behavior in homes?
The geopolitical dimension adds further complexity. If a Chinese operator manages the robots in American factories, or if an American operator controls the robots in European hospitals, who truly governs those machines? Humanoid robots are not passive tools. They see, hear, and interact with sensitive environments. The question of who operates them is a question of sovereignty.
Teleoperation intensifies these concerns. A teleoperator controlling a robot remotely sees everything the robot sees. In a hospital, that could mean patient records. In a factory, proprietary manufacturing processes. In a home, private conversations. If operators are responsible for teleoperation, they become custodians of vast amounts of sensitive data. The potential for abuse, surveillance, or data breaches is immense.
The Question That Remains#
Humanoid network operators are not inevitable. They are one path among many, and the forces pulling for and against them are evenly matched. What is certain is this: someone will manage the millions of humanoid robots entering the world in the 2030s. The question is whether that someone will be the companies that build the robots, the companies that deploy them, or an entirely new category of firm that has not yet appeared.
The answer will determine not just how we use humanoid robots, but who controls the labor, the data, and the infrastructure that shapes the next century of work.