The short answer: China’s industrial-robot output rose 29.1% year on year in the first three quarters of 2026. The IFR’s World Robotics 2026 reports that China installed 354,000 industrial robots in 2025, or 59% of global installations. But output, installations, smart-factory counts, and robot density measure different things. For a 2027 automation budget, start with cycle time, utilisation, changeover, yield, and payback at the workstation — not the headline growth rate.
This report is for manufacturing leaders, plant managers, and procurement teams evaluating automation, line upgrades, and manufacturing AI. Data cut-off: 25 October 2026. Sources are the National Bureau of Statistics (NBS) via Xinhua, the International Federation of Robotics (IFR), MIIT/Xinhua, and the Shanghai municipal government. This is public-source research, not a first-party survey, and it does not quote robot or system-integration prices. Market totals and a single factory’s return are separate questions.
Output growth is not installation growth. Installations are not utilisation. Smart-factory counts are not AI maturity.
What changed in China’s industrial robot market in 2026?
Output rose 29.1% year on year in January–September 2026, while China installed 354,000 robots in 2025, 59% of the global total. The growth is concentrated in electronics, automotive, and smart-manufacturing upgrades, but it does not mean every factory is expanding.
Start with two different measurements. NBS data reported by Xinhua show industrial-robot output up 29.1% in the first three quarters of 2026, with high-tech manufacturing up 14.3% and equipment manufacturing up 10.2%. Output measures domestic production. It is not the number of robots installed in the same period, and it does not show whether those robots are fully utilised.
The IFR’s World Robotics 2026 measures 2025 installations: China installed 354,000 units, up 20% year on year and 59% of the global total. Global installations were about 603,000, up 11%. The IFR forecasts about 655,000 units globally in 2026 and 806,000 by 2029. These are industry statistics and forecasts, not one factory’s order book or payback.
| Indicator | Value | Period / basis | What it measures |
|---|---|---|---|
| Industrial-robot output | +29.1% year on year | Jan–Sep 2026, China | Domestic production; not installations or utilisation |
| High-tech manufacturing value-added | +14.3% year on year | Jan–Sep 2026, China | Broad high-tech expansion; not attributable to robots alone |
| China installations | 354,000 units, +20%, 59% of global | 2025, IFR statistics | New installations; not the installed base or payback |
| Chinese-brand installations | 195,000 units, 55% domestic share | 2025, IFR statistics | Domestic supply share; not a full high-end replacement |
| Electronics installations | 96,400 units, +16%, 27% of China | 2025, IFR statistics | Electronics is a major customer industry |
| Smart factories (basic / excellence / leading) | 56,000+ / 500+ / 15 | H1 2026, MIIT/Xinhua | Graded build-out; not AI adoption or ROI |
| Shanghai 2028 targets | 500 advanced factories, 2,000 AI scenarios, 600 robots/10,000 employees, 8,000 humanoid robots | 2028 policy target, Shanghai government | Policy ambition; not completed installations |
Different units and periods: output, installations, installed base, smart-factory counts, and robot density cannot be added together or used interchangeably. Shanghai figures are 2028 targets; IFR figures are 2025 actuals; NBS figures are Jan–Sep 2026 year-on-year changes.
Does output growth mean every factory is earning a return?
No. Output measures supply. Payback depends on utilisation, cycle time, changeover, yield, and maintenance; the same robot on a low-load line can lose money.
A robot cell is not “robot price divided by labour cost”. It includes the robot, end effector, fixtures, safety fencing, sensors, PLC, programming, shutdown and line modification, spares, maintenance, and training. If orders are insufficient, changeovers are frequent, or cycle time does not match, fixed costs are spread over too few good parts.
Industry output can show supply capacity, localisation, and demand momentum, but it cannot prove that a specific factory’s project will pay back. The 29.1% output figure is for Jan–Sep 2026; the 20% installation figure is for 2025. Treating them as one “2026 installation growth” number is a category error.
| Number | What it measures | What it cannot prove |
|---|---|---|
| Output +29.1% | Domestic production year on year | Every company is buying, installing, or using robots |
| 354,000 installations | New installations in 2025 | High utilisation or completed payback |
| Installed base / density | Stock or robots per 10,000 employees | Flexibility, yield, or AI capability |
| Smart-factory count | Graded assessment result | Every scenario is intelligent |
| Shanghai 2028 targets | Policy target | Already completed in 2026 |
Editorial analysis. Procurement should use the plant’s own cycle time, labour cost, yield, changeover, downtime, and maintenance history, not the national total.
Which industries are driving demand?
Electronics, automotive and parts, batteries, and general industry are the main demand sources. In 2025, electronics installed 96,400 units, 27% of China’s total and up 16% year on year.
The IFR reports that electronics installed 96,400 industrial robots in China in 2025, up 16% and 27% of the national total. This is linked to investment cycles in electronics manufacturing, device replacement, and export demand, but the IFR does not attribute the growth to a single customer or technology.
Automotive and parts remain the most mature robot markets, with clear cycle times and safety rules in welding, painting, handling, and assembly. Batteries, solar, appliances, metalworking, and logistics are also adding automation, but process differences are large; an electronics cycle time cannot simply be copied to discrete assembly or chemical processing.
- Electronics/semiconductors: precision, cleanliness, inspection, and fast changeover; robots often work with vision, AOI, and MES.
- Automotive and parts: cycle time, welding/painting, and safety certification; long projects and integrator capability matter.
- Batteries/new energy: fast process iteration, flexibility, and safety; expansion cycles are volatile.
- General industry/SMEs: machine tending, palletising, grinding, and welding cells are more common; prove payback at one station first.
What do smart-factory counts and robot density actually measure?
They measure digital and automation coverage, not AI capability. In H1 2026 China had 56,000+ basic, 500+ excellence, and 15 leading smart factories.
Smart factories are a graded programme: basic, advanced, excellence, and leading levels have different requirements. A grade shows that a company has completed a certain level of digital, networked, or intelligent build-out. It does not mean every line uses robots, and it does not prove that AI is profitable. MIIT/Xinhua reported 56,000+ basic, 500+ excellence, and 15 leading factories as of H1 2026.
Robot density usually means industrial robots per 10,000 employees. It is useful for comparing regions or industries at a high level, but it does not show utilisation, process difficulty, quality, or payback. Shanghai’s 2028 target is 600 robots per 10,000 employees, alongside 500 advanced factories, 2,000 AI scenarios, and 8,000 humanoid robots. These are targets, not 2026 achievements.
| Concept | Good for answering | Not for answering |
|---|---|---|
| Robot density | Rough regional/industry automation comparison | Whether this plant should buy, or will pay back |
| Smart-factory grade | Stage of digital/networked/intelligent build-out | AI model accuracy or investment return |
| AI scenario count | Policy ambition or application breadth | Whether scenarios run reliably or replicate |
| Humanoid-robot target | Policy direction and industry development | Current general labour replacement |
Editorial judgement. Local policy targets are demand signals, not supplier delivery promises; buyers still need to validate their own process, cycle time, and safety requirements.
What does this mean for 2027 automation and AI budgets?
Budget the full line, not one machine: robot, fixtures, safety, programming, integration, downtime, maintenance, and data. Fix the bottleneck first; plant-wide upgrades come later.
Industry growth can affect supplier lead times, prices, and talent supply, but a 2027 budget should not simply grow by 29.1%. Split the investment into three layers:
- Automation equipment: robots, special machines, conveyors, inspection; prove payback at one station using cycle time, yield, labour, and safety.
- Data and systems: PLC/SCADA, MES, ERP, WMS, and equipment connectivity; solve traceability, scheduling, work orders, and quality closure.
- AI applications: visual inspection, predictive maintenance, knowledge bases, and agents; require data and acceptance metrics before buying a model or platform.
| TCO item | What to ask the supplier | Common omission |
|---|---|---|
| Robot + end effector | Payload, reach, accuracy, cycle time, spares | Grippers, suction, welding torch priced separately? |
| Safety and compliance | Fencing, light curtains, e-stops, risk assessment, certification | Safety retrofit downtime and acceptance responsibility |
| Vision/sensors/PLC | Camera, lighting, algorithms, communication protocols | Shop-floor lighting, occlusion, false-reject rate |
| Integration and programming | Cycle-time validation, commissioning, trial run, training | Changeover programs and operator hours |
| Shutdown and lost output | Retrofit window, ramp-up, backup plan | Who bears the cost of lost production |
| Maintenance and energy | Service interval, spare prices, remote support | Response after the line stops |
| Data/MES integration | Interfaces, traceability, reports, permissions | Data silos and secondary development |
No robot or integration price is quoted here: configuration, brand, payload, accuracy, and site modification vary too much. Require a workstation-level TCO and acceptance plan, not a robot-only unit price.
What should manufacturers of different sizes buy first?
Large manufacturers should replicate proven cells and connect data; mid-sized firms should validate one bottleneck station; smaller firms should audit processes and compare integrators, not chase humanoids.
| Company type | First move | Avoid first |
|---|---|---|
| Multi-site/large manufacturer | Common data definitions, replicate proven cells, MES/ERP integration, framework agreements | Every plant buying an incompatible system |
| Mid-sized manufacturer | POC at one bottleneck station; prove cycle time, yield, labour, and maintenance payback | Full-line robots at once; buying humanoids first |
| Small manufacturer | Process audit, standard workcells, lease/instalments, integrator comparison, safety review | Using “machine replacement” slogans instead of order and process validation |
| Export-oriented manufacturer | Design automation with traceability, lead time, energy, and carbon data | Buying equipment without data and after-sales capability |
Editorial guidance, not investment advice. Industry, order structure, and safety requirements vary; use plant data and a supplier’s site assessment.
What remains uncertain?
Output and installation measures differ, full-year 2026 data is not out, and evidence on robot prices, utilisation, and SME payback is incomplete. Shanghai’s 2028 targets are targets, not results.
What this article can establish is the direction and scale in public statistics, not the return of every segment, company, or workstation. Indicators to watch include full-year 2026 output and installations, utilisation, robot prices and lead times, electronics and automotive investment cycles, acceptance data for AI inspection and maintenance, and SME automation penetration.
Public reports also use different counts for the “advanced” smart-factory tier. This article does not use that number; it uses only the basic, excellence, and leading counts that can be cross-checked. An industry report should inform direction, not replace a plant-level feasibility study.
Where does Lingyi fit?
Lingyi does not build robots. We integrate ERP/MES/OA and manufacturing data, and provide AI inspection, knowledge-base, and agent foundations around the equipment.
Robots provide motion, but orders, scheduling, work orders, quality, equipment, and traceability still need systems. Lingyi’s LingMOE manufacturing platform covers ERP, MES, OA, and manufacturing data; LingHarness can serve as a private AI foundation for knowledge bases, inspection, equipment documentation, and agent integration. We do not sell robot hardware and do not promise an automation payback period.
For a 2027 automation or AI budget, list the workstation, cycle time, yield, downtime, and data interfaces before deciding whether to buy equipment, software, or integration. Related reading: manufacturing AI in 2026, AI inspection cost and ROI, and ERP/MES/OA integration cost. To scope a budget, talk to us.
FAQ
Does 29.1% output growth mean demand is booming everywhere?
No. The 29.1% is China’s Jan–Sep 2026 output growth, a supply-side measure. China installed 354,000 robots in 2025, up 20%. Output, installations, and installed base are different, and a national total cannot tell you a single plant’s orders, utilisation, or payback.
Is 354,000 the global or China installation number?
China. The IFR’s World Robotics 2026 reports 354,000 industrial robots installed in China in 2025, about 59% of the global total of 603,000. This is a 2025 actual, not a 2026 forecast.
Do all 56,000 smart factories use robots?
No. The 56,000+ is the cumulative basic-tier count. The basic tier focuses on digital upgrades and automation of key processes, not full-line robotisation or proven AI returns. The excellence and leading tiers are much smaller.
Are Shanghai’s 8,000 humanoid robots already in factories?
No. The 8,000 is a 2028 policy target, alongside 500 advanced smart factories and 2,000 AI scenarios. It signals policy direction, not current installations or replicable commercial cases.
Should a small manufacturer buy robots now?
First check whether one station has stable orders, enough utilisation, acceptable changeover time, and the safety conditions. If cycle time and payback work, consider a standard workcell, leasing, or instalments. Do not buy a full line because of national output growth or humanoid hype.
How should automation and AI budgets be split?
Separate equipment, data/systems, and AI. Equipment needs a workstation TCO and safety case; systems need interfaces, traceability, and downtime planning; AI needs data and acceptance metrics. If the data foundation is missing, fix PLC/MES/equipment connectivity before buying a large model.
Sources and method
This is public-source research, not interviews, surveys, or equipment testing. Data cut-off: 25 October 2026. Market totals, installation counts, smart-factory counts, and policy targets come from different institutions and periods; each is labelled in the text. The article makes no promise about a single company’s payback, robot prices, or delivery time. Public reports use different counts for the “advanced” smart-factory tier; this article does not use that number.
- Xinhua/National Bureau of Statistics, China’s industrial-robot output up 29.1% in the first three quarters of 2026 (19 Oct 2026): news.cn. Output growth, not installations.
- International Federation of Robotics, World Robotics 2026 report (24 Sep 2026): ifr.org. China installed 354,000 units in 2025, +20%, 59% of global; domestic brands 195,000, 55% share; electronics 96,400, 27%.
- MIIT/Xinhua, smart-factory development in the 15th Five-Year Plan opening year (18 Sep 2026): miit.gov.cn. 56,000+ basic, 500+ excellence, 15 leading factories as of H1 2026.
- Shanghai Municipal Government, next-generation intelligent manufacturing action plan (16 Oct 2026): english.shanghai.gov.cn. 2028 targets: 500 advanced factories, 2,000 AI scenarios, 600 robots/10,000 employees, 8,000 humanoid robots.
The Chinese version is at 2026 中国工业机器人观察.