Executive summary – what changed and why it matters

China’s EV-driven manufacturing and supply-chain leverage is turning humanoids from demos into repeatable operational pilots.

  • Substantive shift: Humanoid companies in China have moved beyond proof-of-concept spectacles to scaled pilot deployments in manufacturing, logistics, retail, and rehabilitation.
  • Scale indicators: Industry reports suggest roughly 13,000 humanoid units shipped globally last year, with Chinese firms capturing over 80% of installations in 2025.
  • Persistent bottlenecks: Software maturity, real-world data scarcity, safety certification, and emerging regulations remain the primary constraints on broader adoption.

Evidence of scale: shipments, deployments, and funding

Multiple market analyses indicate that 2025 marked a turning point for humanoid robots in China. Counterpoint Research data show around 16,000 global installations, with Chinese vendors accounting for more than four-fifths of that volume. The top five global firms—led by AgiBot with an estimated 30% share and Unitree at roughly 26%—collectively held about 73% of the market. Industry projections from government-affiliated bodies suggest the domestic humanoid market could grow from approximately ¥2.8 billion in 2024 to over ¥5 billion in 2025 and approach ¥20 billion by 2026, though longer-term forecasts vary.

Behind these numbers are rapidly closing hardware iteration cycles. Unitree’s Series C funding round valued the company near $3 billion, while AgiBot’s recent financing exceeded $300 million, underscoring investor conviction in volume-driven strategies. At the same time, policy targets set in 2023 aim to elevate China’s humanoid development landscape by 2025, fostering integrated supply chains and international collaboration.

How EV supply-chain ecosystems accelerate hardware iteration

China’s leadership in electric-vehicle manufacturing has seeded a rich ecosystem of high-performance motors, Li-ion battery packs, and perception sensors—critical building blocks for humanoid platforms. Contract manufacturers originally honed in smartphones and EV components now offer turnkey prototyping and small-batch runs that compress R&D-to-production cycles from months to weeks. This dense supplier network lets firms test new actuator designs, vision modules, and chassis configurations on a cadence unmatched by many Western rivals.

Sources familiar with several Chinese humanoid startups note that leveraging off-the-shelf EV-grade components lowers bills of materials by up to 30% compared with bespoke robotics parts. The ability to swap in automotive-grade lidar arrays or custom battery modules in each hardware iteration drives down unit costs and supports early volume discounts in sectors like warehousing and assembly.

Software and data remain the bottleneck

Despite these hardware advances, autonomy and manipulation capabilities lag. Developers rely heavily on large simulation environments and vision-language models running on Nvidia Orin platforms, while domestic AI-chip startups are still maturing. Real-world embodied data—essential for refining world models and edge-case handling—is scarce. Simulated training can accelerate initial development, but it cannot fully capture the variability of on-floor conditions.

Analysts point to predictable failure modes in grasping, dynamic balance recovery, and adaptive planning. Without large-scale field data pipelines, improvements in core autonomy frameworks advance incrementally. This creates a widening gap between mechanical reliability and software intelligence—a gap that will need to close before humanoids can safely navigate open or unstructured environments at scale.

China’s positioning in the global humanoid race

China’s approach leverages a rapid prototype-manufacture-trial-iterate feedback loop, built on decades of EV-industry scale. In contrast, U.S. and Japanese players have emphasized deeper software stacks, advanced control theory, and stringent safety validations. Companies like Tesla, Figure, Agility and Japan’s leading robotics firms are investing heavily in autonomy research and certification processes, aiming for commercial volume by 2027–28.

Meanwhile, Hyundai in partnership with Boston Dynamics is scaling production of articulated mobile platforms, and European entrants are exploring customized service-robot deployments. Yet China’s local cost base, government subsidies, and dense supplier networks offer a time-limited lead in early commercial pilots, especially in logistics, manufacturing and retail fulfillment centers across the Asia-Pacific region.

Regulatory and safety considerations

As pilots expand into real-world operations, safety certifications and data-governance frameworks will become crucial. Industry observers warn that a single high-profile accident could trigger public backlash and export restrictions. Today’s regulatory landscape remains fragmented: domestic standards are in early draft stages, and export markets have varying requirements for liability, privacy and electromagnetic compliance.

In response, Chinese authorities are reportedly drafting national guidelines for humanoid testing corridors, incident reporting protocols, and performance benchmarks. Similar efforts are emerging in the EU and North America, where policymakers are debating liability frameworks for embodied agents and privacy rules for sensor data collection in public spaces.

Implications for stakeholders

  • Enterprises: Pilots in controlled environments—such as warehouses, factory floors and medical-rehab centers—are likely to deliver the first clear ROI signals, favoring vendors that can demonstrate transparent metrics on uptime, task consistency and maintenance cycles.
  • System integrators: Incentives will grow around robust system-level safety architectures, edge-compute redundancy and scalable data-collection pipelines to accelerate software refinement from field operations.
  • Investors: Capital flows appear to be tilting toward firms that combine manufacturing scale with emerging software-and-data strategies—differentiating between pure hardware plays and platform approaches that control the long-term data moat.
  • Regulators and policymakers: A move toward harmonized certification corridors and standardized reporting requirements is expected as deployments scale, particularly to preempt ad-hoc restrictions following any operational incidents.

Looking ahead

China’s supply-chain edge has accelerated the transition of humanoids from demonstrations into operational pilots faster than many anticipated. The near-term battleground will remain in repetitive, predictable tasks under controlled conditions. Longer-term leadership will hinge on closing the software-and-data gap while sustaining safe, certifiable deployments.

Key signals to watch include Q2–Q4 roadmap announcements from China’s Big 5 humanoid firms, progress in domestic AI-chip performance and any large-scale production ramps from U.S. anchors such as Tesla. These developments will indicate whether China’s early volume advantage can translate into a durable ecosystem or if Western autonomy specialists will reclaim share as software maturity and certification regimes advance.