Humanoid Deployment Status

The humanoid robotics field in mid-2026 looks substantially different from how vendor marketing and analyst forecasts describe it. Two gaps separate the announced from the actual.

The first is the maturity gap. The field has many pilots and few operational deployments. By operational deployment we mean a humanoid doing economically productive work for a paying customer, not equipment sales to research customers. By that definition, the Protocol team’s read of available data is that one humanoid company is operating at commercial scale as of mid-2026: Agility Robotics with GXO Logistics. Every other Western program sits in pilot or proof-of-concept. Unit sales to research labs and universities, including Unitree’s substantial volume, are a separate market measuring research community interest rather than commercial viability.

The second gap follows from the first. Vendors are producing units faster than customers are putting them to productive work, because most customer relationships have not matured past pilot. Headline production numbers are not operational deployment numbers.

A pattern sits underneath both gaps: the field is splitting into capability-first and volume-first players. Western vendors compete on capability through deep customer partnerships and slow, high-value deployments. Chinese vendors compete on unit volume and price, often an order of magnitude cheaper. The two may partially converge as Western costs fall and Chinese capability rises, but for now they validate different things, and conflating their numbers is the most common error in reading the field.

Most humanoid coverage conflates two markets that validate entirely different things. Separating them is important for reading the field honestly.

Market 1: Operational deployment. The robot performs economically productive work for the customer’s business. Success is measured in labor offset, units moved, error rate, and uptime. This is the market that tests whether humanoids are commercially viable.

Market 2: Equipment and research sales. The robot is the product being purchased. The customer studies it or develops on it for R&D or academic purposes. Success is measured by the platform’s usefulness as a research tool. This market measures research community interest, not commercial viability.

The Western deployment story is almost entirely Market 1, operational deployment. The high-volume deployment numbers from Chinese companies are largely Market 2, research sales. When an investor asks whether the technology is commercially viable, they are asking about Market 1. A vendor answering with Market 2 unit counts is often not an accurate depiction of overall commercial viability.

The rest of this report concerns Market 1 unless stated otherwise.

Western

• GXO Logistics + Agility Robotics. Multi-year Robots-as-a-Service agreement, Digit deployed at GXO for tote movement. Crossed 100,000 totes moved across the live GXO implementation by November 2025. The most mature Market 1 deployment in the Western market.
• BMW + Figure AI. Figure 02 at BMW Spartanburg for sheet metal loading. 90,000+ parts were loaded, contributing to the production of 30,000+ X3 vehicles.
• BMW + Hexagon Robotics. Hexagon AEON at BMW Leipzig. The pilot focused on collecting real-life data to train future manufacturing applications. The only wheeled-bipedal form factor in pilot.
• Mercedes-Benz + Apptronik. Apollo at Mercedes plants, announced March 2024, scaled through 2025. Mercedes-Benz is exploring use cases for Apollo in logistics, bringing parts to the production line for workers to assemble, and inspecting components.
• Toyota Motor Manufacturing Canada + Agility. Digit at Toyota’s Woodstock, Ontario facility, logistics work, starting early 2026. This is a commercial agreement following a successful pilot deploying Digit at Toyota Motor Manufacturing Canada to support employees with manufacturing, supply chain, and logistics operations.
• Hyundai + Boston Dynamics. Hyundai now owns Boston Dynamics, as of June 2026. Atlas deployment is internal but represents a real OEM-vendor pairing.
• Jabil + Apptronik. Announced February 2025. Jabil is both Apptronik’s worldwide manufacturing partner and a deployment customer, piloting Apollo inside its own operations (“Apollo building Apollo”). Jabil is NYSE-listed (JBL) with 100-plus global sites.
• Mercado Libre + Agility. Announced October 2025 to deploy commercially to increase ergonomic safety for Mercado Libre team members, reduce labor gaps, and increase productivity.

Chinese

The Chinese market is largely Market 2, with some Market 1 inside China. Public information is thinner than for Western pairings.

• UBTECH. Walker S2 is the clearest Chinese Market 1 activity, with hot-swappable batteries supporting continuous operation at Chinese automakers including BYD and Geely. At BYD’s Shenzhen factory, Walker S1 raised sorting efficiency by 120 percent. At Lynk & Co’s Chengdu factory, it cut component storage cycle time by 40 percent and labor costs by 65 percent alongside automated guided vehicles. The Walker S series is now in the manufacturing process of 12 leading enterprises including BYD, Geely, and Foxconn, with orders for over 500 robots.
• Unitree. High-volume sales to research institutions and universities. Predominantly Market 2; unit volumes exceed any Western vendor.
• AgiBot. Multiple manufacturer pairings within China, and a reported 39% share of robotics shipments.

Not all deployments are the same. The gap between a demo and commercial-scale operation is several orders of magnitude in business value. The ladder has six rungs.

1. 6Commercial scale
2. 5Commercial deployment
3. 4Proof-of-concept
4. 3Pilot
5. 2Demo
6. 1Research

Where platforms sit, as the Protocol team reads the field:

• Commercial scale: No one.
• Commercial deployment: Agility with GXO. Likely the only platform at this rung.
• Pilot to proof-of-concept: Figure at BMW Spartanburg, Apptronik at Jabil, Boston Dynamics at Hyundai. Apptronik at Mercedes, Hexagon at BMW Leipzig (scale needs verification).
• Demo to pilot: Tesla Optimus, Sanctuary Phoenix, 1X NEO (consumer-targeted, not customer-paid industrial), most Chinese platforms in Western contexts.

The industry has many pilots and few operational deployments. For capital allocation, this means production capacity and pilot count are weaker signals than the pilot-to-commercial conversion rate.

The maturity gap shows up directly in the production numbers. Per Interact Analysis, only around 10% of humanoid units produced in 2025 were deployed in real-world applications: global production exceeded 20,000 units in 2025, a tenfold increase from fewer than 2,000 units in 2024, with the vast majority used for research, data collection, and entertainment rather than productive work. The rest sit in inventory, evaluation pilots, research environments, or demos.

This changes the meaning of every headline production figure. Figure’s 12,000-unit BotQ capacity, Tesla’s million-unit target, and aggregate Chinese volumes describe potential, not Market 1 deployment. Interact Analysis attributes 2025 growth to an expanding customer base and diversified pilots rather than scaled commercial deployments, with most real-world projects remaining small-scale proof-of-concept work driven by government subsidies, strategic investments, and supply-chain partnerships. Watching this gap close is the most reliable maturity signal, independent of production numbers.

Adoption is uneven across industries. The pattern reveals where Market 1 is real and where it is still narrative.

• Logistics (most mature). The leading vertical. Repetitive tasks, standardized objects (totes, boxes), controlled environments, and a chronic labor shortage create genuine economic pull. Agility at GXO, Toyota, and Amazon. Chinese platforms in domestic logistics.
• Automotive manufacturing (active pilots, not scaled). The highest density of pilots. Parts insertion and body assembly need more dexterity than logistics but remain structured. OEMs have capital and patience. The 2026 to 2027 window will show whether automotive becomes the second mature Market 1 vertical.
• Electronics manufacturing (early evaluation). Higher dexterity requirements than current platforms reliably deliver. Apptronik at Jabil. Chinese platforms at Chinese electronics makers.
• Healthcare and hospitality (aspirational). Discussed by vendors, not meaningfully deployed. Safe human interaction and dynamic-environment adaptation exceed current capability.
• Consumer (mostly cold). Despite marketing focus, no meaningful volume to households. The Protocol team’s read is that consumer deployment lags industrial by 3 to 5 years: higher capability requirements, a higher safety bar, harder unit economics.

Western: capability-first, Market 1. Vendors compete on capability. Deep customer partnerships, multi-year pilots, slow scale-up, high value per deployment. Heavily capitalized for long pilot cycles (Apptronik, Figure, and Agility have each raised hundreds of millions, and cumulative totals vary across sources).

Chinese: volume-first, largely Market 2. Vendors compete on unit volume and price, often an order of magnitude cheaper. Broad distribution, lower value per deployment, faster production scale-up but slower scale-up of operational work. Capital includes state coordination and supply-chain integration Western vendors lack.

The Protocol team’s read is that the two patterns partially converge over the next 18 to 24 months, as Western vendors push costs down (Figure 03 and Tesla Optimus targeting ~$20K) and Chinese vendors push capability up. Whether this yields one global market or two parallel markets is a key open question.

It is hard to predict whether the capability-first or volume-first approach will win.