A factory that knows a bearing will fail before it does is not science fiction. It is sensors, a gateway, a model and a team that can secure all three. The skills behind industrial IoT, from shop-floor protocols to digital twins and edge AI, the EU rules arriving from 2026 to 2028, and how to build a portfolio.
What is industrial IoT, and how is it different from consumer IoT?
Industrial IoT (IIoT) applies connected sensors, networks and data to machines, production lines, energy networks and logistics. The technology looks similar to a smart home; the constraints are not.
In a plant, data moves through a stack described from the floor up: sensors and actuators, the PLCs (programmable logic controllers) that drive machines, SCADA supervision, the MES that runs production, and the ERP that runs the business. Industrial IoT adds connectivity and analytics across this stack, ideally without touching the control loops that keep machines safe: IT/OT convergence, information technology meeting operational technology.
| Question | Consumer IoT | Industrial IoT |
|---|---|---|
| What happens if it fails? | Inconvenience, a support call | Stopped production, damaged equipment, safety risk |
| Typical life of the device | A few years | Machines can stay in service for decades, so new sensors must fit old equipment |
| Protocols | Wi-Fi, BLE, Zigbee, cloud APIs | OPC UA, MQTT and fieldbus protocols, plus LoRaWAN or NB-IoT for wide sites |
| Where data is processed | Mostly in the cloud | Often at the edge, next to the machine, for latency and confidentiality |
| Who you work with | Product and app teams | Maintenance, quality, production and safety engineers, and the IT department |
Industry 4.0 is the umbrella term for this digital transformation of production. The European Commission now frames the next step as Industry 5.0, which builds on Industry 4.0 around three priorities: human-centricity, sustainability and resilience. In that vision, IIoT projects are judged on energy, worker safety and resilience as well as productivity.
How far have French and European industries gone?
Eurostat’s latest business survey on IoT, for 2021, shows how much room is left, especially in France.
| Indicator, manufacturing enterprises with 10+ staff (2021) | France | EU 27 |
|---|---|---|
| Used IoT at all | 21.1% | 29.6% |
| Used IoT for production processes or logistics | 6.0% | 12.0% |
| Used IoT for condition-based maintenance | 4.0% | 8.3% |
The EU has set targets for the infrastructure behind this. Under the Digital Decade, the Commission aims for 75% of European businesses to use cloud-edge technologies and for 10,000 climate-neutral and highly secure edge nodes to be deployed by 2030. Its Edge Observatory report counts about 1,836 nodes in 2024, up from 498 in 2022. The Commission also writes that it expects 80% of all data to be processed soon in smart devices closer to the user, at the edge, rather than in large data centres.
For a student, these gaps mean that many industrial IoT projects in France are still ahead rather than behind.
Which industrial IoT and Industry 4.0 skills matter most?
Six skill families come up again and again in IIoT projects. You do not need to be an expert in all six, but a junior who can work across them fits easily into a project team.
| Skill family | What you actually do | Tools to learn | Portfolio proof |
|---|---|---|---|
| 1. Sensing and connectivity | Pick sensors, wire them to boards, choose the radio for range and battery | ESP32, Raspberry Pi, LoRaWAN, NB-IoT, BLE | A battery-powered vibration or temperature node sending data over LoRa |
| 2. OT integration | Read data from machines without disturbing control, map it to a common model | OPC UA, MQTT, Node-RED, a PLC simulator | Bridge simulated PLC data to an MQTT broker with a clear data model |
| 3. Edge computing and edge AI | Filter, aggregate and score data next to the machine | Docker on a gateway, TensorFlow Lite, model quantisation | An anomaly model running on a Raspberry Pi, with measured latency |
| 4. Data pipelines and predictive maintenance | Clean time series, label failures, predict remaining useful life | Python, pandas, Airflow, InfluxDB, scikit-learn | A pipeline that raises a maintenance alert before a simulated failure |
| 5. Digital twins | Keep a live virtual model of an asset or a line and test changes on it first | Simulation in Python, Grafana, 3D or process models | A mini twin of a conveyor fed by live sensor data |
| 6. Security and compliance | Secure devices and updates, segment networks, document risks | TLS, certificates, OTA updates, IEC 62443 concepts, threat modelling | A threat model and an update plan for your own project |
Digital twins: more than a 3D model
A digital twin is a virtual replica of an asset, process or system kept synchronised with it through sensor data. Its value lies in what it enables: spotting drift, testing a setting without stopping the line, or training a model on rare simulated failures. It takes modelling, data engineering to keep the twin fed, and software skills to make it usable by maintenance teams. Start small: one machine, three sensors, one question.
Edge AI: why the model moves next to the machine
The Commission’s research programme on cloud, edge and IoT explains the logic: processing needs to be closer to where data is produced, to address latency, security, privacy and environmental concerns. On a production line, a vibration anomaly may need to be flagged within milliseconds, and some sites prefer to keep raw data on the premises. Edge AI skills are about small, robust models: compressing them, measuring inference time on a constrained board and planning updates in the field. If machine learning is your entry point, compare with our guide to machine learning master’s programmes in France.
Shop-floor protocols: OPC UA and MQTT
OPC UA gives machines a common, structured way to describe and share their data; MQTT is a lightweight publish and subscribe protocol that suits many sensors and unreliable links. Many IIoT architectures combine them: OPC UA close to the machines, MQTT towards the edge and cloud. Practise explaining when to use which.
Which EU rules shape Industry 4.0 projects from 2026 to 2028?
Five texts now frame how connected machines and products are designed, sold and run in Europe. Engineers who can turn them into design requirements are useful in any IIoT team.
| Rule | Applies from | What changes for IIoT teams |
|---|---|---|
| Data Act, Regulation (EU) 2023/2854 | 12 September 2025 | Users of connected products, including industrial machinery, gain access to the data they generate; products must be designed to allow data sharing |
| Cyber Resilience Act, Regulation (EU) 2024/2847 | Reporting from 11 September 2026; main obligations from 11 December 2027 | Security by design for products with digital elements, vulnerability handling during the support period, reporting through an ENISA platform, CE marking |
| Machinery Regulation (EU) 2023/1230 | Machinery placed on the market from 20 January 2027 | New provisions for AI-powered safety functions and for the cyber-safety of safety control systems and compliance-relevant software |
| NIS 2 Directive (EU) 2022/2555 | Transposition deadline 17 October 2024 | Critical product manufacturing is among the 18 sectors whose medium and large entities must manage cyber risk and report incidents |
| AI Act | 2 August 2028 for AI in regulated products (Annex I) | High-risk rules apply to AI used as a safety component of products such as machinery, with the link to the Machinery Regulation clarified to avoid duplication |
What the Cyber Resilience Act means in practice
According to the Commission’s summary, manufacturers must set a support period for each product, state its end date (month and year) at the time of purchase, and handle vulnerabilities effectively throughout it. When they become aware of an actively exploited vulnerability or a severe incident, they must send an early warning within 24 hours, a notification within 72 hours, and a final report within 14 days of a fix becoming available for an exploited vulnerability, or within one month for a severe incident. Some important and critical product categories need third-party assessment by a notified body.
For an IIoT engineer, this turns into concrete work: an up-to-date inventory of the software components in every firmware image, signed and tested over-the-air updates, vulnerability monitoring, and documentation a market surveillance authority can read. The cybersecurity side of NIS 2 and the AI Act is covered in more depth in NIS 2 and AI Act skills for students.
How can you build industrial IoT skills step by step?
A five-project ladder that a student can complete in a year, each step reusing the previous one.
- 1Instrument one machinePut two or three sensors on a motor, a pump or a fan, send readings over MQTT and graph them.
- 2Bridge OT and ITSimulate a PLC, expose its values through OPC UA, and forward them to your broker with a clean data model and timestamps.
- 3Move intelligence to the edgeRun an anomaly detection model on the gateway, measure inference time and memory, and send only events to the cloud.
- 4Predict and actBuild a pipeline that estimates remaining useful life or flags drift, and triggers a maintenance ticket. Explain the false alarms.
- 5Twin it and secure itFeed a simple digital twin with live data, then write the threat model, the update plan and the support period you would declare under the CRA.
Document each step with a README and a short demo video, then test the skills on real equipment during a placement; work-study in France explains how a salaried contract can fund the final year. The jobs these skills lead to, and what they pay, are covered in our IoT engineer career guide for France.
Where does AI2’s MSc fit for industrial IoT?
AI2, the Applied Institute of Artificial Intelligence, teaches its MSc Applied AI & IoT 100% in English (B2 level, no French, GRE or GMAT required) at the Hub Nikola Tesla campus in Écouen, 22 minutes from Paris Gare du Nord by Transilien line H. Students work on real hardware (ESP32, LoRa, Raspberry Pi), and the career directions it names include digital-twin engineer and IoT and edge-AI engineer.
| Industry 4.0 skill family | Where it is taught at AI2 | Hours |
|---|---|---|
| Industrial information systems | Y1.05 Industrial IS needs analysis: sensors, PLC, SCADA, MES and ERP, BPMN and UML | 35 h |
| Industrial AI and digital twins | Y1.08 AI for Industry (predictive maintenance, visual quality control, digital twins, IoT and edge); Y1.13 IS prototyping and simulation with a mini digital twin | 24.5 h + 42 h |
| OT integration and device security | M2.3 Security and multi-sensor integration: MQTT and OPC UA, remote LoRa link | 21 h |
| Device life cycle | M3.3 Device management and infrastructure as code: provisioning, OTA updates, Ansible, Terraform | 14 h |
| Predictive maintenance and edge AI | M5.2 Statistics and machine learning on IoT data: anomaly detection, remaining useful life, edge deployment with TensorFlow Lite | 28 h |
| Risk and regulation | M1.2 Protocols, risks, budget and regulation: OWASP IoT Top 10, GDPR, NIS2, Radio Equipment Directive | 21 h |
Year 2 follows the five competency blocks of RNCP40554 « Architecte Internet des objets (IoT) », certified by DATAKOO, AI2’s parent company, at level 7. Each block closes with a defence before a jury, and the programme ends with a six-month internship and a professional thesis on an innovative IoT solution. The final year awards the AI2 diploma and prepares this professional certification recognised by the French State at master’s level (EQF 7). How the IoT MSc compares with other English-taught options is in our guide to master’s in IoT in France in English.
| Format | Duration | Tuition 2027 |
|---|---|---|
| Full MSc | Two years: Year 1 applied AI core (500 h), Year 2 IoT (420 h), then 6-month internship | Year 1 €12,500, from €10,500 after scholarship; Year 2 €13,000 |
| Direct entry into Year 2 | 18 months: 12 months on campus, then 6-month internship | €13,000, from €11,800 after scholarship |
Intakes are planned in February, May and September 2027; the priority application deadline for February is 30 November 2026. Budget for visa resources of at least €877.50 a month on top of tuition (Service-Public F2231). Full fee rules are on AI2 tuition fees 2027, and the curriculum module by module is on the MSc Applied AI & IoT programme page.
Frequently asked questions
What is the difference between IoT and industrial IoT?
Industrial IoT applies connected sensors, networks and analytics to machines, production lines, energy and logistics. It must fit equipment that runs for decades, use protocols such as OPC UA, often process data at the edge, and meet stricter safety and security requirements.
What skills are needed for Industry 4.0?
The core skills are sensing and connectivity, integration of operational technology through OPC UA and MQTT, edge computing and edge AI, data pipelines for predictive maintenance, digital twins, and security by design. Communication with maintenance and production teams matters as much as code.
What is a digital twin in manufacturing?
A digital twin is a virtual replica of a machine, line or process that stays synchronised with it through sensor data. Engineers use it to detect drift, test settings without stopping production and train models on simulated failures.
Why does edge AI matter in industrial IoT?
Running models next to the machine cuts latency, keeps sensitive data on site and saves bandwidth. The EU targets 10,000 secure edge nodes by 2030; about 1,836 were deployed in 2024, according to the European Commission.
How does the Cyber Resilience Act affect industrial IoT products?
Manufacturers must design products with digital elements securely, declare a support period, handle vulnerabilities and report actively exploited ones within 24 hours. Reporting has applied since 11 September 2026 and the main obligations apply from 11 December 2027.
Is industrial IoT used much in France?
Eurostat’s latest data, for 2021, show 21.1% of French manufacturers with 10 or more staff using IoT, against 29.6% in the EU, and only 6.0% using it for production or logistics. There was still plenty of room for growth.
Can I learn industrial IoT in English in France?
Yes. AI2’s MSc Applied AI and IoT is taught 100% in English near Paris, with ESP32, LoRa and Raspberry Pi hardware, modules on OPC UA, MQTT, predictive maintenance, edge AI and digital twins, and intakes in February, May and September 2027.
Sources consulted on 29 October 2026: Eurostat: Internet of Things by NACE Rev. 2 activity (isoc_eb_iotn2, updated 11 December 2025) · European Commission: Cloud computing and Digital Decade cloud-edge targets · European Commission: Edge Observatory dashboard · European Commission: Investing in Cloud, Edge and the Internet of Things · European Commission: Industry 5.0 · European Commission: Cyber Resilience Act · European Commission: CRA summary of main provisions · European Commission: Data Act · European Commission: Machinery legislation · European Commission: NIS2 Directive · European Commission: AI Act regulatory framework · France Compétences: RNCP40554, Architecte Internet des objets (IoT) · Service-Public F2231: student visa resources (verified 1 August 2026) · AI2: MSc Applied AI & IoT · AI2: tuition fees 2027
Apply online: AI2’s admissions team reviews your transcripts, holds an online interview and confirms your entry route, tuition and any scholarship in your admission letter. Priority deadline for February 2027: 30 November 2026.
