You source components for a humanoid robot program. Last week in Beijing, 373 exhibitors and 557,000 attendees filled four halls at the World Robot Conference, and somewhere in that noise are the signals that will determine your BOM costs, lead times, and supplier options through 2027. Most trade show recaps read like press releases. This guide doesn't. It extracts what actually changed — and what you should do about it.
We attended with a buyer's eye: which demos were backed by deployments, which component makers have real capacity, and where the pricing pressure lands next. Each section ends with an action item. If a signal doesn't change a sourcing decision, it's noise.
The headline: WRC 2026 was the year the industry stopped performing and started delivering. 40,000+ humanoid robots shipped in H1 2026 — 97% of them from China — and the component supply base that made it possible just branded itself in public.
The Numbers Behind the Show
The Shift: Demos Out, Deployments In
The Component Makers' Coming-Out Party
Who Showed Real Deployments (and Who Didn't)
What This Means for Your Sourcing Strategy
The Post-WRC Supplier Checklist
Decision Framework: Act Now or Wait
Frequently Asked Questions
Before reading any trend into a trade show, check the base rates. WRC 2026's scale tells you this was an industry event, not a science fair:
| Metric | WRC 2026 | vs. Prior Year |
|---|---|---|
| Exhibiting companies | 373 | — |
| Debut products | 311 | +21% |
| On-site attendance | 557,000 | +106% |
| Livestream viewers | 110 million | +94% |
| Short-video topic views | 7.3 billion | +62% |
Two programming decisions mattered more than the attendance records. First, the show ran its first-ever Procurement Day: 49 Chinese central state-owned enterprises arrived with 12 published application scenarios — power grid inspection, mining, oil and gas, emergency response — with real budgets attached. Second, the exhibition layout reserved a dedicated core-components hall, which drew disproportionate foot traffic from engineers and procurement staff: exactly the people who sign specs.
The choreographed dance routines that used to headline robot shows? Largely absent. In their place: parcels sorted under interference, industrial cells palletized live, hotel laundry cycles run end-to-end. The delivery data explains the mood:
China shipped 40,000+ humanoid robots in H1 2026 — 97% of the global total
China's robot industry booked RMB 165.5 billion (~$23.6B) in revenue, up 24.5% year-over-year, after crossing RMB 300 billion for full-year 2025
Agibot has brought 15,000+ robots off its lines cumulatively; UBTech's U1 companion humanoid hit 13,361 pre-orders by end of June
Galbot won a RMB 236 million (~$33.7M) embodied-intelligence equipment contract from CATL
Unitree listed on the STAR Market on opening day, briefly touching RMB 444.9 billion (~$63.5B) in market capitalization
Nobody on the floor was declaring victory, though.
"Operational efficiency is still below human level, and insufficient generalization remains the two biggest pain points for real-world deployment."
— Wang Xingxing, Founder, Unitree
That quote captures the buyer's situation precisely. When thousands of units are moving to paying customers, the questions change shape: nobody assembling 10,000 units asks whether a hand can do a card trick. They ask what the cycle life is, what the lead time is, and what happens when unit 500 fails in the field.
For a decade, robot trade shows followed a simple layout logic: whole machines on the stage, parts in the catalog. WRC 2026 broke it. Chinese tech media summarized the shift bluntly: this year's component exhibitors were collectively "taking the robot apart."
LinkerBot — dexterous hand specialist, exhibiting its LinkerCraft fully automated production line alongside a dual-arm Linker Hand workstation
PaXini — tactile sensor maker with its own major booth and demo scenarios
Wolong Electric — motor giant exhibiting as a standalone robotics brand
Luoyang Hongyuan — the precision bearing supplier behind Unitree and EngineAI joint modules
Core component localization in China has passed 75%, with component prices down roughly 30% over the past twelve months
Leaderdrive now holds 60%+ of China's harmonic reducer market — a category that was import-dominated as recently as 2023
Landian Touch supplies 72.6% of the six-axis force sensors going into Chinese humanoids
Dexterous hand startups alone raised RMB 8 billion+ (~$1.14B) in the first seven months of 2026, up more than 15x year-over-year
The dexterous hand category is projected to sell 70,200 units this year against 19,200 in 2025
💡 What these exhibitors were selling is the deeper signal: not parts for a specific machine, but standard modules any robot can use. The barrier to building a robot has shifted from assembling a supply chain to qualifying the right modules — which concentrates engineering risk in validation, not discovery.
The show's live task environments forced machines to compete on stability rather than spectacle. The pattern across every serious exhibit: a demo backed by a deployment, a deployment backed by a contract. Choreography with no purchase order behind it drew, appropriately, no crowds.
Galbot: Breakfast-cooking long-horizon task ran under deliberate interference, and its new biped Galbot ET-1 learned to play tennis live at the booth — both driven by the "AstraBrain" embodied model that won the CATL contract
UBTech: Cruzr Y1/S2 industrial humanoids ran unattended parts feeding and depalletizing in simulated factory conditions, while the companion U1's 13,361 pre-orders made the consumer case
Xingdong Jiyuan: Its M7 handled the full parcel flow — recognition, grasping, flipping, induction — a system already live in multiple logistics centers
Daxiao Robotics: Instant-retail fulfillment in high-density narrow aisles with sub-1N force control, unattended hotel laundry from collection to folding, and outdoor patrol quadrupeds on 24/7 duty
Unitree: The GD01 crewed mecha — 3 meters tall, switching between biped and quadruped in 2.8 seconds — is positioned for cultural-tourism revenue and is in volume production
Note the pattern for supplier evaluation: every credible exhibitor could name the customer behind the demo. When you qualify a component supplier, ask the same question — who else runs this module in production, and can you talk to them?
Reading the show as a buyer, five structural changes demand action. Each one moves a line in your BOM or your risk register:
Forty-nine central SOEs publishing procurement scenarios means near-term Chinese component capacity has a hungry home-market customer with multi-year budgets. The demand stack now runs SOE programs first, domestic OEM production second, contracted export partners third.
With localization above 75% and prices falling ~30% annually, any Western quote more than ~1.5x the Chinese catalog level now needs justification. Use the gap to negotiate — or to switch.
Finding a supplier is now trivial; validating one is not. Concentrate your engineering effort on cycle-life data, batch consistency, firmware roadmaps, and warranty terms — the failure modes that surface at unit 500, not at the demo bench.
In a market where demand triples annually, the most expensive line in your BOM is waiting. A module that costs 20% more but ships from a US warehouse in days rather than weeks from Shenzhen is usually the better engineering decision.
Sample-tested modules through 2026 are worth converting into contracted allocation now, before the SOE programs' delivery schedules harden into capacity commitments. The buyers who qualify second sources this year will be the ones still shipping next year.
Whether you met suppliers at the show or are qualifying them remotely afterward, run every candidate through the same checklist. This is the process our team uses before a module enters the RobX catalog:
Monthly production capacity stated and verifiable (ask for line photos or audit reports, not brochures)
Component self-development rate known — vertical integration predicts price stability
Explicit allocation policy for export customers — where do you sit in the demand stack?
Lead time for samples vs. production quantities, in writing
Cycle-life test data available (not just peak specs)
Batch consistency data across at least 3 production lots
Named production customers you can reference
Firmware roadmap and update policy documented
English documentation — SDK, datasheet, integration guide
Warranty terms with defined response times
Spare-parts availability and field-replacement policy
Local (US-based) technical support or a partner who provides it
Warning
If a supplier cannot name a production customer, has no cycle-life data, and offers only Chinese-language documentation, treat them as a prototype shop — fine for R&D units, wrong for a 10,000-unit program. The gap between the two is where late-stage program failures come from.
Not every program needs to move immediately. Use this matrix to match your situation to an action:
| Your Situation | Decision | Action |
|---|---|---|
| Scaling to production in 2026-2027 | ACT NOW | Convert sample-tested modules into contracted allocation before Q4. Dual-source every critical category. |
| Prototype stage, production in 2027+ | PREPARE | Sample now, benchmark against Chinese reference pricing, keep two qualified candidates per category. |
| Currently buying Western components at 2x+ Chinese pricing | REVIEW | Re-anchor your pricing. The arbitration is no longer between suppliers; it's between a reference price and a legacy quote. |
| R&D only, single units | MONITOR | No urgency. Buy from distributors with local stock; re-evaluate when volumes approach double digits. |
If you source components at volume, yes — but go with an agenda. WRC 2027 runs August 18, 2027 in Beijing E-Town. The core-components hall is where your time belongs: it's the densest concentration of Chinese module makers you can meet in three days anywhere. Book supplier meetings before you fly.
No, but the window is closing. The SOE procurement programs announced at the show will harden into capacity commitments over the next two quarters. Overseas buyers who move before Q4 2026 will get contracted allocation; those who move later will get whatever remains.
The premise is outdated. The same components now ship in 97% of the world's humanoid robots — the quality question has been settled by volume, not by argument. The real quality variance is between individual suppliers, which is exactly what the Section 6 checklist is for.
This is why dual-sourcing and local inventory matter. The module itself is increasingly fungible; the risk sits in logistics and policy. A US-based distributor with California stock absorbs both — you pay a margin for buffer stock and domestic warranty, and for most scaling programs that trade is worth it.
Where the price gap and consolidation are largest: harmonic reducers (Leaderdrive at 60%+ domestic share), force sensors (Landian Touch at 72.6%), and dexterous hands (LinkerBot at 80%+ of high-DOF units). These are also the categories where allocation pressure will hit first.
The bottleneck moved to delivery. 40,000+ Chinese humanoid units shipped in H1 2026 — 97% of the global total — with real contracts behind the demos.
Components took the stage. For the first time in the show's history, component makers exhibited as independent brands — modules are now a standalone procurement category.
Localization passed 75%, prices fell ~30%. Chinese catalog pricing is now the global reference. Leaderdrive at 60%+ in harmonic reducers, Landian Touch at 72.6% in force sensors.
The first Procurement Day changed the demand stack. 49 central SOEs with published scenarios mean overseas buyers now queue behind domestic programs for capacity.
Buyers must act before Q4. Re-anchor pricing, dual-source by category, run the supplier checklist, and lock allocation while first sources are still shipping.
RobX Robotics supplies Linker Hand dexterous hands and CrabNova joint actuators from leading Chinese manufacturers, with US-based PhD-level technical support, California sample inventory, and transparent lead times.