Embed this dashboard

Copy the code below into any HTML page or WordPress Custom HTML block. The iframe resizes itself to fit its content.

Free to embed under CC BY 4.0 with attribution to Axis Intelligence Research.
Axis Intelligence Research · Industrial Tech

AI in Manufacturing Statistics 2026

Co-author Sophie Winslow · updated October 26, 2026 · source & dataset
11 CHARTS · 4 TABLES · 386 SOURCED ROWS · CC BY 4.0
EU
EU adoption
17.27%
EU manufacturers using any AI, 2025 · 10.57% in 2024
FFAS
Factory-Floor AI Share
20.2%
AI-using EU manufacturers applying it to production
US
US adoption
22.9%
US manufacturers using AI, Sep 21 – Oct 4, 2026
1:5
Scaling gap
4.2 : 1
Use cases creating value per use case scaled across sites
+11%
Robots installed
603,307
Industrial robots installed worldwide, 2025 · stock 5,079,078
EurostatICT usage in enterprises 2025isoc_eb_ain2 · NACE C
U.S. Census BureauBTOS release Oct 22, 2026sector 31–33 · NAICS 311–339
Intl. Federation of RoboticsWorld Robotics 2026released Sep 24, 2026
Deloitte · SiemensAI in Manufacturing 2026 · TCOD 2024company research
WEF · NAM · RockwellLighthouses · Q2 Outlook · SoSM2026 editions
01Office AI vs shop-floor AIWhere European factory AI actually runs
From all EU manufacturers to AI on the line · 2025
Only 3.49% of all manufacturers use AI in production
No AIAI elsewhere in the businessAI in production processes200 cells · 1 cell = 0.5% of manufacturers, rounded
UniverseEU manufacturing enterprises, 10+ persons employed
100%
Any AI technologyUp 6.70 points in one year
17.27%
AI for marketing or salesThe largest single purpose
5.25%
AI in production processesFourth of seven purposes
3.49%
Formula5.25 ÷ 3.49 = 1.50 — EU manufacturers apply AI to marketing and sales 1.5 times as often as to production. Each cell in the grid is 0.5% of the manufacturing population.
ReadingThe adoption wave reached the commercial office first. A licence and a browser are a full deployment in sales; on the line, AI needs a camera, an edge device, a PLC connection and an acceptance test on cycle time.
SourceEurostat, isoc_eb_ain2, NACE section C, 2025 survey year, data updated June 15, 2026. Ratio computed by Axis Intelligence Research.
What AI-using EU manufacturers apply AI to · share of AI users · 2025
Production ranks 4th of 7
Formulashare = purpose rate ÷ 17.27 × 100. Marketing: 5.25 ÷ 17.27 = 30.4%, matching Eurostat's own published 30.41%. Firms can report several purposes.
ReadingLogistics is the outlier: only 6.7% of AI-using manufacturers point it at moving material, the lowest of the seven purposes Eurostat measures.
SourceEurostat, isoc_eb_ain2, 2025. Shares re-based to AI users by Axis Intelligence Research.
EU manufacturers using AI · percent2021–2025
Eurostat, isoc_eb_ain2, NACE C, enterprises with 10+ persons employed. No 2022 observation is published.
AI technologies in EU manufacturing · percent2025
Eurostat, share of all EU manufacturers using each technology. Language tools lead; machine-facing technologies trail.
Language AI against physical AI · EU manufacturers · 2025
Text mining is 5.81× autonomous-machine AI
Text mining · written language
9.42%
The single most used AI technology in EU manufacturing
5.81×multiple
Autonomous robots, vehicles, drones
1.62%
AI that lets machines move by observing their surroundings
Formula9.42 ÷ 1.62 = 5.81. Meter segments are 1 point each on a 0–10% scale. Image recognition, the basis of vision inspection, sits at 3.14%.
ReadingPhysical AI is the smallest technology category in the survey, which is consistent with how the robot industry itself grades it: vision is industry-grade, general-purpose physical AI is not yet.
SourceEurostat, isoc_eb_ain2, 2025; IFR World Robotics 2026 Executive Summary. Multiple by Axis Intelligence Research.
02Factory-Floor AI ShareAxis Intelligence Research metric · FFAS · 27 countries
How FFAS is built · recomputable from two Eurostat indicators
EU reading 20.2%
Factory-Floor AI Share (FFAS)
E_AI_PPP · AI in productionE_AI_TANY · any AI× 100 =3.4917.27× 100 =20.2%
The share of AI-using manufacturers that point AI at production processes. Read it together with production-AI penetration: a high share on a small base, as in Romania (41.0% on 3.24% adoption), describes very few plants.
Leader
Belgium
10.18% of manufacturers run AI in production · FFAS 25.6%
EU27
3.49%
Production-AI penetration · FFAS 20.2%
Broad, shallow
Germany
24.38% use AI, 4.00% in production · FFAS 16.4%
FormulaFFAS = E_AI_PPP ÷ E_AI_TANY × 100, NACE section C, enterprises with 10+ persons employed, 2025 survey year, per country.
ReadingFFAS separates the AI a factory sells with from the AI it makes things with. That distinction is invisible in every headline adoption rate.
SourceEurostat isoc_eb_ain2 inputs; calculation by Axis Intelligence Research. Finland is unscored: Eurostat did not publish its 2025 values.
Country league · any AI (bar) against AI in production (dot) · 2025
Sorted by production-AI penetration
Manufacturers using any AIUsing AI in production processesScale: bar 0–40%, dot 0–11%
FormulaBar length is E_AI_TANY ÷ 40; dot position is E_AI_PPP ÷ 11, a finer scale that keeps small production rates legible next to the adoption bar.
ReadingNordic factories adopt broadly but sit low on FFAS: Sweden 15.3%, Norway 10.8%. Austria and Belgium combine high adoption with high production depth.
SourceEurostat isoc_eb_ain2, 2024–2025. FFAS and changes computed by Axis Intelligence Research.
Full FFAS table · 26 EU countries, EU27 and Norway · 2025
EU adoption up +6.70 pp on 2024
CountryAny AI 2025Any AI 2024Change (pp)AI in productionFFAS
EU27 reading17.27%10.57%+6.703.49%20.2%
FFAS10% → 41%
FFAS against production-AI penetration · EU countries2025
Axis Intelligence Research from Eurostat isoc_eb_ain2. Upper right: deep and broad factory AI. The dashed line marks the EU reading of 20.2%.
03United StatesCensus Bureau BTOS · release of October 22, 2026
US manufacturing against all US businesses · AI use in the prior two weeks
Gap narrowed from 4.6 to 1.7 points
Nov 3–16, 2025
12.7%Manufacturing
17.3%All businesses
Sep 21 – Oct 4, 2026
22.9%Manufacturing
24.6%All businesses
Manufacturing, eleven months
+10.2 pp
1.8 times the starting level, against +7.3 points for all businesses. 27.0% of manufacturers expect to use AI within six months, an intent gap of 4.1 points.
Formula22.9 − 12.7 = 10.2 · 22.9 ÷ 12.7 = 1.80 · gap 24.6 − 22.9 = 1.7. Standard error on the latest manufacturing reading: 1.35 points.
ReadingA two-week question counts a planner drafting in a chatbot the same as a vision cell rejecting parts at line speed. Read the climb as breadth of use, and the subsector split below for where it is concentrated.
SourceU.S. Census Bureau, BTOS national and sector response estimates, question 7 (AI in any business function), release of October 22, 2026. Calculations by Axis Intelligence Research.
US manufacturers using AI · percent of businessesNov 2025 – Oct 2026
U.S. Census Bureau, BTOS sector 31–33. Points are labelled by the start of each two-week reference period; alternate periods shown, all periods in the dataset.
AI use by US manufacturing subsector · estimate with ±1 standard error
Machinery 13.0% → 27.2%
Sep 21 – Oct 4, 2026±1 standard errorNov 3–16, 2025Scale 0–45%
FormulaShown only where standard error < 0.25 × estimate; nine of 21 subsectors qualify. Change column is the latest reading minus November 2025.
ReadingFood manufacturers nearly tripled their AI use and machinery makers more than doubled it. Chemicals barely moved because they started first. Fabricated metal remains the slowest measurable subsector at 14.8%.
SourceU.S. Census Bureau, BTOS subsector response and standard-error estimates, release of October 22, 2026. Filter by Axis Intelligence Research.
04Robots and physical AIIFR World Robotics 2026
Installed, 2025
603,307
Industrial robots worldwide, up 11% · first year above 600,000
Operational stock
5,079,078
Robots at work in factories, end-2025, up 9%
Market value
US$18.6bn
Value of 2025 installations, up 11%
Humanoids sold
~7,000
Full-size units in 2025 · 174 manufacturers identified
Robot installations by market · units2025
IFR World Robotics 2026. China 354,226 (59% of the world); the US passed Japan by 2,209 units.
Installations by customer industry · world2025
IFR World Robotics 2026 Executive Summary. Electronics led automotive by about 3,800 units.
How the robot industry grades factory AI · IFR maturity reading
Cobots: 11% of installations
Industry-grade today

Machine vision for detection and robot guidance is established in production.

Increasingly used

Predictive maintenance, process optimization and AI-based inspection are spreading through industry.

Not yet industry-grade

General-purpose physical AI and learning from simulation and experience. Most humanoids remain prototypes or pilots.

Formula~7,000 ÷ 603,307: about one humanoid sold for every 86 conventional industrial robots installed in 2025.
ReadingIFR counts physical machines and excludes software bots and RPA, so five million robots are not five million AI deployments. The maturity grading matches Eurostat: vision is on the line; general-purpose physical AI is not.
SourceIFR World Robotics 2026 Executive Summary and press-conference presentation, September 24, 2026. Ratio by Axis Intelligence Research.
Robot density vs AI in production · EU countriesr = 0.12
IFR robot density per 10,000 manufacturing employees (2023, World Robotics 2024) against Eurostat production-AI rates (2025). Shown side by side, years not merged.
Installations by region · units2025
IFR World Robotics 2026. Asia 76% of installations; Europe down 14%, Americas up 14% to a record.
Automation depth against AI depth · Germany and Belgium
Correlation across 12 countries 0.12
Germany
429
Robots per 10,000 employees (2023) · 4.00% of manufacturers with AI in production (2025)
1.92×density
Belgium · IFR: with Luxembourg
224
Robots per 10,000 employees (2023) · 10.18% with AI in production (2025), 2.55× Germany
Formula429 ÷ 224 = 1.92 · 10.18 ÷ 4.00 = 2.55 · Pearson r across 12 EU countries = 0.12. Meters show production-AI rate on a 0–10% scale.
ReadingRobots and factory AI are being bought as two separate investments. In Germany, 83.6% of AI-using manufacturers do not apply AI to production processes.
SourceIFR public density chart (2023); Eurostat isoc_eb_ain2 (2025). Calculations by Axis Intelligence Research.
05Pilot to plant networkWhy factory AI stalls between site one and site two
Value proven, value replicated · manufacturers engaged with AI
Scaling Gap Ratio 4.2 : 1
Report measurable value
84%
of 140+ manufacturers surveyed by Deloitte
Use cases scaled across sites
~20%
About one in five, consistently across sites or enterprise-wide
Physical AI / closed loop
18%
Respondents using AI that acts directly on equipment
Formula84 ÷ 20 = 4.2. Quality leads adoption at 62%, production 57%, logistics 49%; 79% cite operational disruption as a risk.
ReadingThe pilot works because an integrator instrumented one asset. The rollout stalls because asset 37 is a different model on a different control system. The ratio measures that brownfield distance.
SourceDeloitte, AI in Manufacturing 2026, survey of more than 140 manufacturing organizations. Ratio by Axis Intelligence Research.
Why EU manufacturers that considered AI did not adopt it · 2025
Only 15.8% found AI not useful
Formulashare = reason rate ÷ 11.76 × 100, where 11.76% of EU manufacturers ever considered AI. Firms can cite several reasons.
ReadingThe obstacles are capability and installed base, not doubt about value. Incompatibility with existing equipment, at 44.6%, is the brownfield problem in a single line.
SourceEurostat isoc_eb_ain2, NACE C, 2025. Re-based by Axis Intelligence Research.
06The cost baselineDowntime and what the best plants report
Cost of one hour of unplanned downtime · US$log scale
Siemens, True Cost of Downtime 2024 (company research; 181 interviews extrapolated to Fortune Global 500 plants). An automotive hour costs about 64 times an FMCG hour.
Downtime ledger · Fortune Global 500
~$1.4tn
Lost a year to unplanned downtime
11% of revenue
27 h
Unplanned downtime per large plant per month
39 h in 2019
$2.3M
One lost hour, large automotive plant
Highest in the study
$36,000
One lost hour, fast-moving consumer goods
63.9× lower
Global Lighthouse Network · site results reported through the World Economic Forum
Network size 238 sites
SiteCountryReported resultValueRecognized
Hitachi Vantara, NormanUnited StatesOrder-to-ship lead time−77%June 2026
Rockwell AutomationSingaporeUnits per person-hour+43%June 2026
Taiji GroupChinaOEE · production cost+13% · −10%June 2026
CIMC Reefer ContainersChinaConversion cost · lead time−24% · −32%June 2026
Midea WuhuChinaDelivery lead time · 35% of 113 use cases AI-driven−39%January 2026
Unilever HefeiChinaLead time · 65% of 31 use cases AI-driven−75%January 2026
New sites in 202623 in January · 16 in June94% of transformations combine several technology domains
FormulaSite results as reported by each Lighthouse and published by the Forum; no averaging across sites. OEE multiplies availability, performance and quality.
ReadingLighthouses describe the ceiling, not the median plant. Midea's 35% is the useful detail: even at a top site, two in three digital use cases are not AI.
SourceWorld Economic Forum releases of January 15 and June 22, 2026.
07Workforce and readinessTasks are changing faster than headcount
~96%
AI-using US firms with no headcount change in six months
Minneapolis Fed · Census
55.2%
US manufacturers training frontline workers on AI
NAM Q2 2026
41.43%
US manufacturers providing no AI training
NAM Q2 2026
34%
Operations manufacturers consider AI-augmented today
Rockwell 2026
43%
Collected plant data used effectively
Rockwell 2026
46%
Manufacturers with a cyber incident in the past year
Rockwell 2026
FormulaShares as published by each source. NAM: 215 responses, fielded May 12–28, 2026. Rockwell: 1,560 respondents in 17 countries.
ReadingThe data gap and the training gap point the same way. Plants keep more data than they use, and four in ten US manufacturers have not yet trained the people who would act on a model's output.
SourceFederal Reserve Bank of Minneapolis from Census BTOS; National Association of Manufacturers; Rockwell Automation (company research).