Humanoid Robot Shipments Soar in 1H26

The global humanoid robotics industry has reached a pivotal juncture for large-scale commercial implementation, according to Counterpoint Research. Worldwide shipments of humanoid robots exceeded 22,000 units during the first half of 2026, representing a year-over-year increase of nearly 300% and sustaining robust growth momentum.

AGIBOT maintained its position as the leading global vendor by volume, delivering approximately 9,700 units in the first half to secure a market share exceeding 43%. Unitree followed with more than 7,000 units, capturing 31% of the market. Since the beginning of 2026, Galbot has experienced significant shipment expansion, with cumulative first-half deliveries surpassing 1,100 units for a 5% market share. UBTECH, a publicly traded entity, ranked fourth, reaching the thousand-unit milestone and accounting for 4.4% of the market, while Leju Robotics recorded 650 units for a 2.9% share.

Regarding application sectors, the aggregate share of the entertainment and performance category, along with data production and research, declined slightly in the first half but remained above 60%. Within this context, there is a discernible demand for small-form-factor humanoid units. Despite the entry of new competitors, the AGIBOT X and Unitree G1 series continue to dominate this niche, bolstered by established toolchains and content creation ecosystems.

Humanoid robots utilized in the service and guidance sector currently rank third by shipment volume, representing an estimated 19% market share. Offline retail environments, such as autonomous pharmacies, have become crucial arenas for data training. During the same timeframe, market shares for the intelligent manufacturing and warehousing sectors also increased, reaching 13% and 5%, respectively.

In the second half of 2026, several early-stage pilot programs in the service and industrial domains are expected to transition into mass deployment. Humanoid robots are being integrated into routine operations at major transportation hubs, while automotive manufacturers are increasingly deploying these units across the entire value chain, from production to sales and after-sales services.

For standardized operations such as material handling, sorting, and inspection within 3C manufacturing and logistics settings, practical deployment solutions tailored to real-world conditions are surfacing. As world-model technologies are integrated into existing Vision-Language-Action (VLA) architectures, the industry is entering a critical window for breakthroughs in intelligence and task generalization.

Recently, the convergence of world models and VLA has evolved into a prominent technical paradigm. Utilizing world-model components, the new end-to-end Transformer architecture can generate actions while simultaneously predicting continuous states for long-duration tasks. This framework addresses previous VLA limitations in physical reasoning and rule learning, offering improved robustness in dynamic settings and enabling generalization under few-shot or zero-shot conditions.

Counterpoint Research forecasts that global humanoid robot shipments will surpass 50,000 units in 2026, a 210% year-over-year growth. Over the next five years, it is anticipated that service and industrial applications will displace entertainment and research categories as the primary drivers of demand within the humanoid robotics sector.

Concurrently, the competitive landscape for vendors will shift from a focus on raw model performance and production capacity toward establishing closed-loop capabilities. This includes advanced model development, vertical-scenario implementation, robust data system architecture, and the capacity for rapid model iteration.