Gospodarka robotow ma problem z właścicielami: do kogo należy praca maszyny?

In October 2025, 1X Technologies announced a pricing model for its NEO Gamma humanoid robot that seemed almost quaint: $20,000 to buy, or $499 per month to rent. The press covered the headline numbers. What nobody asked was the more interesting question.

If a NEO Gamma performs household tasks worth $15-$20 per hour for a family that would otherwise hire help, the robot generates roughly $25,000-$35,000 in labor-equivalent value per year. At $499 per month, 1X captures about $6,000 of that annually. The family captures the remaining $19,000-$29,000 in value. A clean split, more or less. The family pays a fraction of what a human helper would cost. 1X collects recurring revenue. Everybody wins.

Now scale that math to a warehouse. A $100,000 humanoid robot working two shifts in a logistics facility generates approximately $180,000-$240,000 in labor-equivalent value per year, based on the Bureau of Labor Statistics’ all-in compensation cost for warehouse workers at $22-$28 per hour. Under a purchase model, the deploying company captures nearly all of that value after hardware, maintenance, and energy costs. Under a Robot-as-a-Service subscription model, the manufacturer captures a recurring slice, potentially 30-50% of the value generated.

The difference between those two models is not a pricing decision. It is a question about who will own the economic output of the next industrial revolution.

Three models, three futures

The humanoid robot industry is converging on three distinct business models for getting robots into the world. Each model creates a fundamentally different economic relationship between the company that builds the robot, the company that deploys it, and the workers whose labor it replaces or augments. The industry talks about these models as pricing strategies. They are not. They are blueprints for how trillions of dollars in economic value will be distributed over the next two decades.

Model 1: The purchase model

The simplest structure. A company buys a robot outright, operates it, maintains it, and captures the full economic output. This is how traditional industrial robots have been sold for 50 years. It is how most humanoid robots are sold today.

Tesla’s Optimus, priced at $100,000-$150,000 for enterprise buyers, follows this model. So does the Unitree G1 at $16,000. Boston Dynamics sells Atlas units through enterprise purchase agreements with multi-year service contracts attached.

Under the purchase model, the manufacturer’s revenue is front-loaded. They collect the hardware price, plus annual maintenance fees that typically run 15-30% of the purchase price. The deploying company bears the capital risk but also captures the ongoing value.

The economics are straightforward. If a $100,000 robot generates $200,000 per year in value and costs $40,000 annually to operate (maintenance, energy, software, depreciation), the deploying company earns $160,000 per year in net value. The robot pays for itself in roughly 8 months. After that, every dollar of value it generates flows to the company that bought it.

This model works well for large companies with the capital to buy fleets of robots and the operational expertise to maintain them. It works poorly for small and mid-size businesses that cannot afford $100,000 per unit upfront, cannot hire robotics technicians, and cannot absorb the risk that the technology will be obsolete in two years.

Model 2: The lease model

Leasing shifts the capital burden from the deploying company to a financing entity, usually a bank, equipment leasing company, or the manufacturer itself. The deploying company pays a fixed monthly amount over a set term, typically 36-60 months, and either returns the robot or buys it at residual value at the end of the lease.

This is the model that scaled industrial robot arms across automotive manufacturing in the 1990s and 2000s. It is also the model that dominates commercial vehicle fleets, medical imaging equipment, and enterprise IT hardware.

For humanoid robots, the lease model solves the capital problem but introduces a new dynamic: the manufacturer or financing entity retains ownership of the physical asset. They own the robot. The deploying company owns nothing. They are renting productive capacity.

This distinction becomes important when you consider what happens at lease end. If the robot has improved through software updates over the lease term, is it worth more or less than when the lease started? If the manufacturer has collected operational data from the deployed robot and used it to train better AI models, who owns the value of that data? The lease contract says the deploying company is renting hardware. The reality is that data is flowing back to the manufacturer, making the manufacturer’s product better, and the deploying company has no claim on that value.

Model 3: Robot-as-a-Service

This is where it gets interesting. And where the landlord analogy starts to apply.

Robot-as-a-Service, or RaaS, is a subscription model where the manufacturer deploys, operates, maintains, and updates the robot, and the customer pays a recurring fee based on usage, output, or a flat monthly rate. The customer never owns the robot. They never touch the robot’s software. They may never even interact directly with the robot’s controls. They simply pay for the work the robot performs.

1X Technologies’ $499/month NEO rental is an early consumer version of this model. On the industrial side, several companies are exploring RaaS structures for enterprise deployments. The logic is compelling: instead of asking a warehouse operator to spend $100,000 on an unproven technology, offer them productive capacity at $3,000-$5,000 per month. Lower barrier to adoption. Predictable costs. No maintenance burden. The manufacturer handles everything.

But RaaS also means the manufacturer captures a much larger share of the value the robot generates, permanently. Under the purchase model, the manufacturer gets paid once (plus maintenance). Under RaaS, the manufacturer gets paid every month, forever, for as long as the customer needs the work done. And the customer never builds equity. They never own anything. They rent access to productive capacity that someone else controls.

The landlord analogy

In real estate, a landlord owns a productive asset (a building) and charges tenants for access to it. The tenant does the actual productive work inside the building, running a business, living a life, but the landlord captures a share of the value through rent. Over time, the landlord’s asset appreciates. The tenant’s payments build no equity.

The RaaS model creates an almost identical structure for physical labor. The robot manufacturer owns the productive asset (the robot). The deploying company rents access to it. The deploying company does the productive work around the robot, managing the warehouse, fulfilling orders, serving customers, but the manufacturer captures a recurring share of the robot’s output through subscription fees. Over time, the manufacturer’s asset improves (through software updates and data-driven AI improvements). The deploying company’s payments build no equity in the robot.

There is one critical difference that makes the robot version potentially more powerful than the real estate version: the manufacturer controls the means of improvement. A landlord does not make a building get better over time through software updates. A robot manufacturer does. Every improvement to the AI, every software update, every new capability added through the cloud makes the robot more valuable, and the manufacturer decides when, whether, and at what price to deliver those improvements.

The numbers behind the models

Let us run the economics over a 5-year deployment to see how the three models compare. Assume a single humanoid robot deployed in a logistics facility, generating $200,000 per year in labor-equivalent value.

Under the purchase model, the manufacturer captures $80,000-$120,000 over five years (hardware sale plus maintenance contracts). The deploying company captures the remaining $680,000-$720,000 in value after operating costs.

Under RaaS at $4,000 per month, the manufacturer captures $240,000 over five years. The deploying company captures $560,000. The manufacturer’s share roughly doubles compared to the purchase model.

But the real difference is in the terminal economics. After five years under the purchase model, the deploying company owns the robot and can continue capturing value at near-zero incremental cost (just maintenance and energy). After five years under RaaS, the deploying company owns nothing. They must keep paying the subscription or the productive capacity disappears. And the subscription price can increase.

McKinsey estimated in a 2026 report that the shift from purchase to RaaS in industrial robotics could increase manufacturer lifetime revenue per unit by 2-3x. For an industry projected to ship millions of robots by the early 2030s, that multiplier changes the economic landscape by trillions of dollars.

Who is building what model

The industry is not evenly split. Different companies are gravitating toward different models based on their strategic positions, financial needs, and views about long-term value capture.

Tesla is committed to the purchase model, at least initially. Elon Musk has repeatedly stated that Optimus will be sold, not rented, with a long-term consumer price target of $20,000-$30,000. Tesla’s business model for Optimus mirrors its approach to vehicles: sell the hardware, then monetize through software subscriptions (Full Self-Driving for cars, potentially premium AI capabilities for Optimus). This is a hybrid that captures hardware margin upfront and recurring software revenue over time.

1X Technologies is the most explicit about RaaS in the consumer market. The $499/month rental model for NEO Gamma is a pure subscription play. At that price, 1X captures roughly $6,000 per year per robot. If the robot costs $15,000-$18,000 to manufacture, it pays for itself in manufacturing cost within 2.5-3 years, and everything after that is margin.

Agility Robotics has signaled interest in RaaS for its Digit platform, particularly for logistics deployments where customers want to pay per unit of work done rather than per robot purchased. Amazon, Agility’s largest deployment partner, has deep experience with usage-based pricing models and may prefer a structure where it pays per tote moved rather than per robot deployed.

Boston Dynamics offers both purchase and lease options for Atlas, with premium service contracts that blur the line between the two models. The service contracts are so comprehensive (dedicated engineers, preventive maintenance, priority parts) that the ongoing costs approach RaaS levels even under a purchase agreement.

Figure AI has not publicly committed to a specific model but has the financial position (backed by $1.85 billion in funding at a $39 billion valuation) to experiment with RaaS without needing immediate hardware revenue.

The gig economy parallel

There is a useful parallel in recent history: the gig economy.

Before Uber, most taxi drivers owned or leased their vehicles. They bore the capital cost, but they also captured the full fare minus a dispatch fee. The vehicle was their asset. When Uber introduced its platform model, drivers still bore the capital cost (their own cars), but now Uber captured 25-30% of every fare through its platform fee. Drivers did the work. Uber owned the algorithm.

The Robot-as-a-Service model inverts this dynamic. Instead of workers owning the means of production (their bodies, their cars) while a platform extracts a fee, the robot manufacturer owns the means of production (the robot) while the deploying company extracts value from its operation. The worker is removed from the equation entirely.

This is not a subtle shift. In the gig economy, at least the worker retained the ability to work independently. If Uber’s take rate became too high, the driver could work for Lyft, or go back to a taxi company, or find another job. In the robot economy, the deploying company cannot switch to a competing robot on short notice (integration costs, retraining, workflow redesign). And the workers who were displaced have no robot of their own to fall back on.

The labor surplus question

The most uncomfortable dimension of the robot economy is what happens to the economic surplus previously captured by human workers.

When a warehouse employs 100 workers at $50,000 per year in total compensation, those workers collectively capture $5 million in economic value. That money flows into the local economy: rent, groceries, healthcare, education, entertainment. When 30 of those workers are replaced by humanoid robots, the $1.5 million in labor value they previously captured does not disappear. It is redistributed.

Under the purchase model, most of that value flows to the warehouse operator as increased profit margin. Under RaaS, it is split between the operator and the robot manufacturer. Under neither model does any meaningful share flow to the displaced workers or the communities they live in.

This is not a hypothetical concern. The Congressional Budget Office estimated in a 2025 report that automation-driven labor displacement could reduce aggregate consumer spending by 0.5-1.2% annually in affected regions if displaced workers do not transition to equivalent-paying jobs within 18 months. The historical evidence from manufacturing automation suggests that roughly 40% of displaced workers find equivalent or better employment, 35% find lower-paying work, and 25% exit the workforce or experience extended unemployment.

The robot economy will generate enormous aggregate value. Morgan Stanley projects a $5 trillion total ecosystem by 2050. Goldman Sachs projects $38 billion in annual revenue by 2035. The question that these projections do not answer, and that business model decisions will determine, is who gets to keep that value.

The data question nobody is asking

There is a dimension of the robot economy that is potentially more consequential than the hardware or the subscription fees: data.

Every humanoid robot operating in the real world generates gigabytes of sensor data per day. Cameras record workspaces. Force sensors record how objects feel. Joint encoders record how movements succeed or fail. This data is the raw material for training better AI models, which is the single most valuable input in the entire robotics value chain.

Under the purchase model, the question of who owns this operational data is ambiguous. Some contracts give the deploying company full ownership. Others include clauses allowing the manufacturer to collect anonymized telemetry for product improvement. The deploying company rarely negotiates these clauses aggressively because the data seems like a byproduct, not an asset.

Under RaaS, the data question is typically settled in the manufacturer’s favor. Since the manufacturer owns the robot, maintains the software, and manages the cloud infrastructure, operational data flows naturally to the manufacturer’s servers. The deploying company may not even realize what data is being collected, how it is being used, or what it is worth.

This creates a compounding dynamic. The manufacturer with the most deployed robots collects the most data. The most data trains the best AI. The best AI makes the robots more capable. More capable robots attract more customers. More customers generate more data. The flywheel is not about hardware margin or subscription revenue. It is about data accumulation. And the business model determines who controls the flywheel.

The policy vacuum

As of April 2026, no major economy has enacted legislation that specifically addresses the economic structure of robot labor deployment. Tax codes were written for a world where productive capacity was either human (and subject to payroll taxes, minimum wage laws, and labor protections) or capital equipment (and subject to depreciation schedules, investment tax credits, and property taxes). Humanoid robots under RaaS fit neatly into neither category.

Should a robot that replaces a human worker be subject to some form of payroll-equivalent tax? Several European countries have debated this. South Korea briefly had an “automation tax” that reduced tax incentives for companies investing in automation, though it was structured as a reduction in investment tax credits rather than a direct tax on robots.

Should workers displaced by robots have a legal claim on the economic surplus the robot generates? This sounds radical, but it has historical precedent. When containerized shipping displaced longshoremen in the 1960s, the International Longshoreman’s Association negotiated guaranteed annual income provisions that were funded by a share of the productivity gains from automation. The economic value of automation was split, by contract, between the shipping companies and the workers whose jobs the containers eliminated.

No equivalent mechanism exists for the robot economy. And the business model decisions being made right now, in boardrooms and pricing spreadsheets, are determining the default distribution in the absence of any policy framework.

Three scenarios for the robot economy

The business model decisions being made in 2026 will compound over the next two decades. Here is what the robot economy could look like depending on which model wins.

Scenario 1: Purchase dominates. Humanoid robots follow the trajectory of personal computers. Prices fall to $10,000-$20,000. Businesses buy them like they buy forklifts. Value flows primarily to deployers. Manufacturers compete on hardware and software quality, margins compress over time, and the industry looks like traditional equipment manufacturing. This is the most decentralized outcome, with value distributed broadly across millions of deploying companies.

Scenario 2: RaaS dominates. Humanoid robots follow the trajectory of cloud computing. Manufacturers capture 30-50% of the value through perpetual subscriptions. A small number of manufacturers with the best AI models and the most operational data become the dominant players, collecting rent from every warehouse, factory, and household that uses their robots. This is the most concentrated outcome, with value flowing primarily to 3-5 major manufacturers.

Scenario 3: A hybrid market. Large companies buy robots. Small companies rent them. The market splits along the same lines as the car market, where corporations buy fleets and consumers lease or finance. Data rights become a negotiated term in enterprise contracts. Regulation eventually imposes some form of displaced-worker compensation funded by a share of automation-driven productivity gains.

What to watch

The robot economy’s ownership structure will be determined in the next 3-5 years, before regulation catches up, before the market reaches scale, and before the compounding effects of data accumulation become irreversible. Here are the signals that will tell you which scenario is playing out.

Enterprise contract terms. Watch for the data rights clauses in large humanoid robot deployments. If deployers start demanding data ownership or data profit-sharing, the market is resisting manufacturer concentration. If they do not, the manufacturers win the data game by default.

RaaS adoption rates. If RaaS accounts for more than 40% of new humanoid robot deployments by 2028, the rental model is winning. If it stays below 20%, the purchase model is holding.

Subscription price trends. If RaaS prices stay flat or decline as the technology improves, the market is competitive. If prices increase while robot capabilities improve, manufacturers are exercising pricing power from lock-in. This is the clearest indicator of rent-seeking behavior.

Policy responses. Watch South Korea, the EU, and California. These jurisdictions have the most active automation policy discussions. Any legislation that taxes robot labor or mandates displaced-worker compensation will reshape the economics.

Manufacturer margin disclosures. As humanoid robot companies go public (Unitree IPO’d in early 2026, others will follow), their financial statements will reveal the margin structure of different business models. If RaaS margins are significantly higher than hardware margins, expect every manufacturer to shift toward subscriptions.

The question worth $5 trillion

Morgan Stanley projects the total humanoid robot ecosystem will be worth $5 trillion by 2050. That number includes hardware, software, services, supply chains, maintenance, and all the economic activity that robots enable.

What Morgan Stanley does not project, and what no analyst report addresses directly, is the distribution of that $5 trillion. Will it flow to thousands of companies that own their robots and capture the value of their work? Or will it concentrate in the hands of a small number of manufacturers who own the robots, collect the subscriptions, and accumulate the data?

The answer depends on decisions being made right now. Not in legislatures. Not in think tanks. In the pricing models and contract terms that humanoid robot companies are drafting this quarter. Every RaaS contract signed, every data rights clause accepted without negotiation, every subscription price set without competitive pressure shapes the economic architecture of an industry that does not yet exist at scale but soon will.

The robot economy does not have to have a landlord problem. But it will, unless the people building it, buying it, and regulating it start asking a question that nobody in the industry seems to want to answer: who owns the work a machine does?