The Humanoid Race: What's Actually Deployed in 2026

Overview

Humanoid robots crossed from demonstration into commercial deployment in 2026, and the shape of that transition is different from what the sector's pricing announcements implied.

Figure AI has a 40-unit fleet of Figure 03 robots working at BMW's largest assembly plant, billed at roughly $25 per robot-operating-hour. That is a services arrangement — BMW is buying labour hours, not machines.

Tesla runs over 1,000 Optimus units inside its Gigafactory Texas facility, sorting parts, carrying components and performing quality inspections. It also discontinued the Model S and Model X in early 2026 and converted those Fremont production lines to build robots. Despite all of that, Optimus was still unavailable for external purchase as of April 2026.

The units people can actually buy come from elsewhere. Unitree shipped over 5,500 robots in 2025 with the G1 listed at $17,990, and the OpenAI-backed 1X NEO has over 10,000 home pre-orders at $20,000 or $499 a month.

Figure's 40 Robots at BMW

The most significant deployment in the sector is also one of the smallest in unit terms. Figure AI has 40 Figure 03 robots commercially deployed at BMW's largest assembly plant in the world.

The word doing the work is commercially. This is not a pilot, a proof of concept or a demonstration arranged for press coverage — it is a paying arrangement with an automotive manufacturer that has no incentive to keep unproductive machines on a production line.

Forty units is a modest fleet, and that is appropriate for the stage. A manufacturer integrating a new class of machine into a working plant scales gradually because the cost of disruption to an assembly line vastly exceeds the cost of the robots.

$25 Per Robot-Hour

The billing model is the most consequential detail in the entire sector and receives almost none of the attention that price targets do. Figure charges roughly $25 for each hour a robot operates.

This inverts how humanoid robots have been discussed. Announcements have focused on unit prices — $16,000, $20,000, $25,000 — framing robots as capital equipment a business buys. Charging by the operating hour makes them a service, priced directly against the labour they replace.

It also relocates the risk. A customer buying a $25,000 robot bears the cost if it underperforms, breaks or becomes obsolete. A customer paying $25 an hour pays only for hours worked, and the manufacturer carries maintenance, upgrades and downtime. For an unproven technology, that is the arrangement most likely to get a purchase order signed — and it means adoption can happen without anyone committing capital.

Tesla Has 1,000+ Optimus, All Internal

Tesla operates over 1,000 Optimus robots inside Gigafactory Texas, performing parts sorting, component carrying and quality inspection alongside human workers.

A thousand units is a substantial fleet, considerably larger than any external commercial deployment. It gives Tesla something no competitor has at that scale: continuous operational data from robots doing repetitive industrial work in a real facility, generating training data and failure cases.

Using your own factory as the first customer is a sound strategy for exactly this reason. There is no customer to disappoint, no contractual performance standard, and problems can be fixed without commercial consequence. It is also why internal deployment numbers should not be compared directly with commercial ones — they measure different things.

You Still Can't Buy One

Despite years of announced pricing targets, Optimus remained unavailable for external purchase as of April 2026.

This matters because the widely quoted $25,000 figure is a stated target for a product that is not for sale. It has anchored public expectations about what humanoid robots should cost, and no external customer has been able to test it.

The pattern is familiar from other Tesla programmes, where ambitious timelines and prices are announced well before delivery. Treating an announced target as a market price distorts comparison with companies that are actually shipping — Unitree's $17,990 is a price you can pay today, and it is not really comparable to a number attached to an unavailable product.

Tesla Killed the Model S and X for This

The clearest signal of Tesla's commitment is not a price target but a production decision. The company permanently discontinued the Model S and Model X in early 2026 and converted those Fremont factory lines to humanoid robot manufacturing.

Those two vehicles were Tesla's flagship products and the cars that established it as a serious manufacturer. Ending them to free capacity for robots is a substantial and difficult-to-reverse allocation of industrial capital.

Manufacturing capacity is the scarcest resource in this sector. Designing a humanoid robot is achievable for a well-funded team; building thousands of them reliably is a different problem, and it is the one automotive manufacturers are unusually well equipped to solve. Converting existing automotive lines is the fastest route to volume anyone has.

Unitree Ships the Most Units

By volume, the leader is neither Tesla nor Figure. Unitree shipped over 5,500 units in 2025 and is targeting 20,000 in 2026 — close to a fourfold increase.

Its G1 is listed on Amazon at $17,990, which makes it the most accessible genuine humanoid available. Buying a walking, programmable humanoid robot through a mainstream retail channel would have seemed implausible a few years ago.

The units serve different purposes than industrial fleets. Much of this volume goes to universities, research labs, developers and demonstration use rather than production work. Shipping the most robots and doing the most economically useful work with them are separate achievements, and Unitree currently leads decisively on the first.

1X Has 10,000 Home Pre-Orders

The consumer market has its first real test in 1X's NEO, which is backed by OpenAI and has accumulated over 10,000 home pre-orders with deliveries expected through 2026.

The pricing offers both models: $20,000 to purchase or $499 a month. The subscription option matters for the same reason Figure's hourly billing does — it removes the requirement to commit substantial capital to an unproven product.

Household tasks are a considerably harder problem than factory work. A production line is structured, repetitive and controlled; a home has stairs, pets, children, clutter and no two layouts alike. Ten thousand pre-orders indicate genuine consumer appetite, and the deliveries will be the first real evidence of whether current humanoids can function in unstructured environments.

50,000 Shipments Forecast for 2026

Total humanoid shipments are forecast to exceed 50,000 units in 2026 across a field including Tesla, Figure, Boston Dynamics, Agility Robotics, Unitree, AgiBot, UBTECH, 1X, Neura Robotics and Kawasaki.

That is rapid growth and a small absolute number. For comparison, the global industrial robot market installs several hundred thousand units annually — conventional arms and automated systems that do not walk but reliably perform specific tasks.

Humanoids remain a niche within automation, and their advantage is specific rather than general. A robot shaped like a person can use environments and tools built for people, which avoids redesigning a facility around the machine. That flexibility is worth paying for in some settings and is straightforwardly inferior to a fixed arm in most others.

Why Robot-Hours Beat Robot Sales

The economics of the $25 hourly rate explain more about this sector's near-term trajectory than any specification sheet.

At $25 per robot-hour, a machine running two shifts a day generates roughly $400 daily. That comfortably exceeds the unit cost of a Unitree G1 within a couple of months of operation — which tells you that the price of the hardware is not the constraint. Reliability, capability and integration are.

It also sets the competitive benchmark clearly. Twenty-five dollars an hour is above minimum wage in most jurisdictions and below fully loaded costs for skilled industrial labour in high-wage economies. Humanoids do not need to be cheap to be adopted; they need to work reliably enough that an hour of robot time genuinely substitutes for an hour of human time.

That is why deployment is currently concentrated in structured industrial settings with a handful of customers, and why 40 robots at BMW is more meaningful evidence than a thousand demonstration videos.

Conclusion

Humanoid robots reached genuine commercial deployment in 2026, and the milestone was a business model rather than a machine. Figure has 40 robots working at BMW's largest plant billed at roughly $25 per robot-operating-hour, which prices them against labour rather than selling them as equipment.

Tesla has made the largest industrial commitment — over 1,000 Optimus units running inside Gigafactory Texas, and Model S and Model X production lines at Fremont converted to build robots. It also has nothing for sale, which means its widely quoted $25,000 target remains untested by any external buyer.

The units that actually ship come from elsewhere: over 5,500 Unitree robots in 2025 with a target of 20,000, and 10,000-plus pre-orders for the OpenAI-backed 1X NEO. Total shipments should exceed 50,000 in 2026 — real growth, and still a fraction of conventional industrial robot installations.

The number to watch is the hourly rate rather than the sticker price. At $25 an hour the hardware pays for itself in weeks, which means the constraint was never cost. It is whether a robot-hour can reliably replace a human one.

Data Source and Attribution

Figure AI1X TechnologiesUnitree

Deployment figures, fleet sizes, billing rates, shipment volumes, pre-order counts and pricing come from company disclosures and industry reporting current to mid-2026. Tesla's internal Optimus deployment, Fremont line conversion and external availability status reflect reporting as of April 2026. Shipment forecasts for 2026 are projections rather than recorded figures. Announced price targets for products not yet available for purchase are identified as targets in the text.

FactsFigs reviews, cleans, and cross-checks every source dataset before shaping it into a data story. Each visualization is created and designed in FactsFigs Design Studio — an internal tool developed and owned by FactsFigs — and is the original work of a FactsFigs author, not an AI-generated copy of any existing graphic. Individual assets within a visual may or may not be produced with AI tools, but the design of the visual itself is solely FactsFigs' own.

Figures are estimates at the time of publication, provided for information only — nothing here is financial advice or a guarantee of accuracy.

2026-07-20